3 L .2533:
Say-5.033”

.‘o‘. 3.2.... II!»

2...... 1...
6.33%... 5,...
..‘..:.-1 %.\~u..r
135...”... 2132qu .
II ”I a 15-. It‘ll

M233.

U»
i . is 5...
IV. I}... 5!...
.“W\.\ . 5L:
7 4:) .2. .
i an 0.5....
[hr

‘
aﬁtﬁv..1
In. 5

flag”.

 

.11....
it}. is... A. I
xv h! a 3511...):
11.253!!!
tall)
10::
I. ..

 

.IvH...w... .

y -333... .. ,.

.nuuv....§....}ev.a i .2 1
.. “31!.“ 3;... “£394...

m V' ‘ 4 '
. .. ,-
.uliA-N’ﬁ

l('\. “I

 

LIBRARY
Michigan State

I'm:u IAUA:‘e .
UlllVb‘lblly

 

 

 

This is to certify that the
dissertation entitled

Motivation to Change in Sex Offenders:
Exploring Sex Offender Types

presented by

LaTanya A. Carter

has been accepted towards fulﬁllment
of the requirements for the

PhD. degree in Psychologx

 

 

%/%e/7

/ Major Profess’or's Signature
////2:./2

Date

MSU is an Afﬁrmative Action/Equal Opportunity Employer

_ -.-.—-a-o-o-l-s---.-.—

-.—.—---.—.--.-.-.-.- —

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:lProj/Aoc&Pres/ClRC/Dateoue.indd

 

MOTIVATION TO CHANGE IN SEX OFF ENDERS:
EXPLORING SEX OFFENDER TYPES

By

LaTanya Adelia Carter, MA.

A DISSERTATION
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY

Psychology

2009

 

ABSTRACT

MOTIVATION TO CHANGE IN SEX OF FENDERS:
EXPLORIN G SEX OFFENDER TYPES

By
LaTanya Adelia Carter, MA.

The current study examined two issues: First, this study examined how variations
in psychopathology yielded clusters or types of sex offenders. Second, this study
examined the clinical relevance of these clusters focusing on motivation to change and
response to treatment. These issues were evaluated using the archival data of 228
convicted adult male sex offenders (81% with child victims; 19% with adult victims) who
were assessed and treated at a local outpatient treatment facility while on probation or
parole. Cluster analyses were conducted using both agglomerative and iterative methods,
resulting in the formation of four clusters. The ﬁrst cluster (N =86) displayed no clinical
levels of psychopathology; even their level of antisocial behaviors and attitudes was not
in clinical ranges. The second cluster (N=69) was characterized by antisocial behaviors
and attitudes as well as hostility and mistrust of others. The third cluster’s (N=3 6)
psychopathology was deﬁned by antisocial behaviors and a cunning, manipulative
interpersonal style but not hostility or mistrust. The ﬁnal cluster (N =3 7) had the most
severe psychopathology, including elevated antisocial behaviors, depression, anxiety,
paranoia or mistrust, and symptom exaggeration. Demographically these clusters
differed in age of the offender (p<.001) and parenthood status (p<.05) but not in race,
education, marital or employment status, or a range of offense and victim characteristics.

Next, this study analyzed motivation to change as a latent construct. Using

structural equation modeling, the four subscales of the motivation to change

 

 

questionnaire served as the indicators predicting motivation to change. Results indicated
a non-signiﬁcant chi-square and good ﬁt indices (Chi-square=2.97; df=2; p=0.23;
GFI=O.99; CFI=1.00; RMSEA=0.05; BICﬁ6.40). Also, each of the four indicator paths
were signiﬁcant and in the expect direction. These results indicate that motivation to
change does not occur in stages but rather can be better represented as unitary processes.
To validate motivation to change as a predictor of treatment response, structural equation
modeling was used to ﬁt the latent construct as predictors of post-treatment response
controlling for pre-treatment functioning. Motivation to change was signiﬁcantly
associated with treatment response (Chi-square=15. 1; df=8; p=0.06; GFI=O.98; CFI=O.99;
RMSEA=0.06; BIC=85.71) but the effect size was modest (beta <.30).

Finally, differences were evaluated between the clusters with regard to clinical
variables including motivation to change, treatment attendance and completion, and a
number of clinician-rated psychological variables assessed before and after treatment.
Clusters differed on treatment attendance (p<.01) and motivation to change (p<.05) but
not other variables. These ﬁndings provide partial validation to the clusters.

These results suggest that variations in psychopathology may result in the
formation of distinct clusters of sex offenders. These offender clusters may not best be
deﬁned by their offense characteristics. However, the clinical differences between these
clusters suggests that, regardless of their offense characteristics, the type and severity of
psychopathology that these offenders display could have particular relevance for their
motivation to change and attendance in treatment. A key next step is to evaluate whether
those offenders with low motivation to change may beneﬁt from motivational

interviewing prior to treatment assignment.

TABLE OF CONTENTS

LIST OF TABLES ................................................................................... vi
LIST OF FIGURES ................................................................................ viii
INTRODUCTION AND LITERATURE REVIEW ............................................. 1
Sex Offending as a Societal Problem ...................................................... 1
MOTIVATION TO CHANGE IN SEX OFF ENDERS ......................................... 7
Sex Offenders’ Heterogeneity: Psychopathology, Motivation, and Treatment
Responses ...................................................................................... 7
Paralleling the Sex Offending and Substance Abuse Literatures to Conceptualize
Mot1vat10n12
Theories of Sex Offending ................................................................ 18
Treatment Implications of Studying Motivation to Change .................................. 26
Conclusion .................................................................................. 27
PSYCHOPATHOLOGY AND SEX OFFENDER HETEROGENEITY ................... 28
Overall Prevalence Rates of Psychopathology in Sex Offenders .................... 28
Prior Attempts to Cluster Sex Offenders Based on Psychopathology ............... 29
Treatment Implications of Studying Psychopathology Types ........................ 42
Conclusion .................................................................................. 43
AIMS OF PRESENT STUDY ..................................................................... 44
Expected Psychopathology Clusters ..................................................... 44
Descriptive Variables ...................................................................... 45
Hypotheses .................................................................................. 46
METHODS .............................................................................................. 48
Deﬁnitions .................................................................................. 48
Archival Data ............................................................................... 48
Data Collection and Preparation ......................................................... 52
Data Quality Assurance ................................................................... 52
Final Sample Size, Missing Data, and Invalid Protocols ............................. 55
Measures .................................................................................... 57
Data Analyses .............................................................................. 66

iv

RESULTS ............................................................................................ 68
Hypothesis 1: Sex Offenders will form at least ﬁve homogeneous clusters based

on their psychopathology traits ........................................................... 68
Hypothesis 2: The clusters will differ on offense characteristics such as victim
Hypothesis 3: Motivation to change will form into one construct which will
predict both treatment attendance and completion for se offenders ................. 87
Hypothesis 4: Clinically, these clusters will vary by their level of motivation to
change and treatment variables (e.g., attendance, completion) ...................... 89
Post-Hoc Analyses ......................................................................... 91
CONCLUSIONS AND DISCUSSION ........................................................... 93
Comparison of Current Clusters to Previous Clustering Attempts .................. 93
Integration of Cluster Analyses and Treatment Implications ........................ 97
Motivation to Change as Latent Construct ............................................ 103
Motivation to Change as a Predictor of Treatment Responses ..................... 105
Treatment Implications .................................................................. 107
Limitations ................................................................................. 108
Future Research Directions .............................................................. 1 11
Conclusions ................................................................................ l 13
APPENDICES ..................................................................................... 114
Appendix A: Tables for Statistical Results ............................................ 114
Appendix B: Figures for Statistical Results ........................................... 145
Appendix C: Figures and Tables for Introduction .................................... 151
Appendix D: Additional Research Information on TTM ........................... 154
Appendix E: Psychopathology and Victim Status .................................... 156
Appendix F: The URICA and Sex Offenders ......................................... 158
Appendix G: Intervention Strategies Used in CBT with Sex Offenders .......... 160
Appendix H: Motivational Interviewing ............................................... 163
Appendix I: Sampling Procedures ...................................................... 167
Appendix J: Description of MMPI Scales Used in Present Study .................. 170
Appendix K: Cluster Analysis: Background .......................................... 178

Appendix L: Post-Hoc Analyses for Predictors of Motivation to Change. . . . . ...182

REFERENCES .................................................................................... 187

 

 

LIST OF TABLES

Table 1: Reliability Statistics for All Variables in the Dataset .............................. 114
Table 2: Descriptive Statistics: All Variables (Before Imputations) ........................ 116
Table 3: Distribution of Categorical Variables ................................................ 1 19
Table 4a: Inter-Correlations of Offense Characteristics ...................................... 120
Table 4b: Correlations between URICA Subscales and Offense Characteristics ......... 121
Table 40: Correlations between MMPI Scales and Offense Characteristics ............... 122
Table 5a: Inter-Correlations of Clinical Variables ............................................. 123
Table 5b: Correlations between Clinical Variables and Offense Characteristics ......... 124
Table 5c: Correlations between URICA Subscales and Clinical Variables ............... 125

Table 5d: Correlations between MMPI Scales and Clinical Variables ...................... 126
Table 6a: Inter-Correlations of URICA Subscales ............................................ 127
Table 6b: Inter-Correlations of MMPI Subscales ............................................. 128
Table 6c: Correlations between URICA Subscales and MMPI Scales ..................... 129
Table 7: Scale Reliability Statistics for URICA ............................................... 129
Table 8a: Inter-Item Correlations for URICA Pre-Contemplation Subscale ............... 130
Table 8b: Inter-Item Correlations for URICA Contemplation Subscale ................... 130
Table 8c: Inter-Item Correlations for URICA Action Subscale ............................. 130
Table 8d: Inter-Item Correlations for URICA Maintenance Subscale ..................... 130
Table 9a: Scale Reliability Statistics for Pre-Treatment Variables ......................... 131

Table 9b: Scale Reliability Statistics for Post-Treatment Variables ........................ 131

Table 10a: Initial Iterative Cluster Analysis with Scale Sc (Three-Cluster Solution). . ..132

vi

Table 10b: Initial Iterative Cluster Analysis without Scale Sc (Three-Cluster
Solution) ............................................................................................ 132

Table 11a: Initial Iterative Cluster Analysis with Scale Sc (Four-Cluster Solution). . 1 33

Table 11b: Initial Iterative Cluster Analysis without Scale Sc (F our-Cluster
Solution) ............................................................................................. 133

Table 12a: Initial Iterative Cluster Analysis with Scale Sc (F ive-Cluster Solution). . ....l34

Table 12b: Initial Iterative Cluster Analysis without Scale Sc (F ive-Cluster

Solution) ............................................................................................ 1 34
Table 13a: Means for Three-Cluster Solution (Agglomerative Hierarchical) ............. 135
Table 13b: Means for Four-Cluster Solution (Agglomerative Hierarchical) .............. 135
Table 13c: Means for Nine-Cluster Solution (Agglomerative Hierarchical) .............. 136
Table 14: Final Iterative Cluster Analysis (Four-Cluster Solution) ......................... 137
Table 15: Case Description of Cluster Assignment ........................................... 138
Table 16a: MANOVA for Cluster Membership and MMPI ................................. 141
Table 16b: Between-Subject Effects for MANOVA in Table 16a ........................... 141

Table 17: Summary of Cross-Validation Analyses for Demographic, Offense, and
Clinical Variables .................................................................................. 142

Table 18a: Hierarchical Regression for Cluster Membership and Motivation to Change
Predicting Post-Treatment Scores ............................................................... 144

Table 18b: Hierarchical Regression for Cluster Membership and Motivation to Change
Predicting Post-Denial Scores .................................................................... 144

Table 18c: Hierarchical Regression for Cluster Membership and Motivation to Change
Predicting Clinical Change ....................................................................... 144

Table 18d: Hierarchical Regression for Cluster Membership and Motivation to Change
Predicting Denial Change ........................................................................ 144

vii

 

LIST OF FIGURES

Figure 1: Diagram of Agglomerative Hierarchical Cluster Analysis (N=228) ............ 145
Figure 2: Diagram of Agglomerative Hierarchical Cluster Analysis for Random Half of
Dataset (N=114) .................................................................................... 146
Figure 3: Diagram of Agglomerative Hierarchical Cluster Analysis for Other Random
Half of Dataset (N=114) .......................................................................... 147
Figure 4: Line Graph of the Four Psychopathology Clusters and Sample Mean ......... 148
Figure 5: Case Migration from Hierarchical Solution to Final Iterative Solution. . . . . 149

Figure 6: SEM Analysis Testing Motivation to Change as a Single Latent Construct. ..150

Figure 7: SEM Analysis Validating Motivation to Change Using Treatment
Variables ............................................................................................ 151

viii

 

INTRODUCTION AND LITERATURE REVIEW
S at Offending as a Societal Problem

Prevalence of sex oﬂending. Sexual offending as used here refers to direct
physical sexual assault of a child or adult woman by an adult man. It is a major societal
problem not only because of the high incidence rates but also because of the sustained
physical and psychological injury it causes the victims (e. g., Sommers et al., 2006).
Statistical reports indicate that 18% of adult women are sexually assaulted during some
part of their lives (15% completed assault; 3% attempted assault; Tjaden & Thoennes,
1998). However, these numbers are probably underestimates due to under-reporting
(Tjaden & Thoennes, 1998). For children under the age of 18, estimates of sexual assault
rates have reached as high as 44% (both attempts and completions; Greenfeld, 1997). Of
this 44%, 15% were under the age of 12 and 29% were between the ages of 12 and 17
(Greenfeld, 1997). Evidences of sustained physical and psychological injury are
numerous in the literature (e. g., Ahrens, Dean, Rozee, & McKenzie, 2008; Sarkar &
Sarkar, 2006; Sommers, 2007; Sommers, Schaefer, Zink, Huston, & Hillard, 2001;
Sommers et al., 2006). Thus, the problem is important.

The present study focused on male sex offenders. 1 An understanding of their
characteristics may aid in treatment planning and reducing re-offense rates, assuming that
current sexual re-offense rates can continue to decrease. Although researchers have been
striving toward this goal for many decades (e.g., Bowman, 1938; Smith, 1924), it has not

yet been achieved and new, innovative research is necessary to keep the ﬁeld moving

 

' Because females compose only l-2% of all sex offenders and are, therefore, understudied in the literature
(Vandiver, 2002), the sample for the present study consists of only male sex offenders. Consequently, the
literature review which follows describes studies with samples of male sex offenders only. Additional
operational deﬁnitions of several of the concepts being used in this document can be found in the Methods
section.

.sAl’.

 

 

forward in this direction. In the United States there are more than 386,000 registered sex
offenders (Adams, 2002). This estimate is increased from previous estimates which
indicated that 234,000 sex offenders were under the supervision of correction
departments around the country (Greenfeld, 1997). Of these sex offenders, 60% were
under supervision in the community (i.e., on parole or probation), with the other 40%
either in prison or acting under their own recognizance in the community (Greenfeld,
1997). These statistics imply that at least 60% of convicted sex offenders were in a

position to re-offend against a member of the general public if they were not successfully

 

rehabilitated or deterred. However, as with other crimes, not all sex offenders re-offend, i
which may suggest potentially important clinical differences among them. This variation
among offenders may provide clues valuable to new interventions that may further reduce
chances of re-offending.

Recidivism2 among sex oﬂenders. Sex offender re-offense rates are an ongoing
concern in American communities. Although from one perspective recidivism may be
“low” (see below), recidivism remains important to try to reduce further given the harm
that these offenses cause (Sarkar & Sarkar, 2006). The present study assumes that factors
which have not been empirically tested in sex offenders such as motivation to change
may be key to understanding the variation in sex offenders’ recidivism rates.

New laws have been developed and implemented in order to protect the public
from sex offenders in the community (e. g., Megan’s Law, 1996; Levenson & Cotter,

2005). Convicted adult sex offenders in the community must publicly register their

 

2 Recidivism refers to the number of offenders who re—offended among a total number of offenders in a
given time period. These re-offenses could be violent, in that they involve physical assault or nonviolent
(e. g., theft). They could also be sexual, in that the offenders committed another sexually violent crime or
nonsexual (e.g., domestic assault). In the present study, sexual offenses are conceptualized as a special
type of violent offense.

 

name, address, crime(s), aliases, and photographs so that the public can access them at
any time (Levenson & Cotter, 2005). Sex offenders cannot live close to schools nor have
jobs involving children. They are mandated to treatment and, in many ways, are
monitored more closely than other violent, non-incarcerated offenders (Levenson &
Cotter, 2005). With these policies in place, a recent study revealed that, after an average
follow-up period of ﬁve years, sex offenders’ general recidivism rate (i.e., including
violent and nonviolent, sexual and nonsexual offenses) was 28.2%, about the same as a
27.2% general recidivism rate for other types of criminals (Craig, Browne, Beech, &

Stringer, 2004). When considering sexual recidivism speciﬁcally (i.e., sexual offenses

 

only), sex offenders’ rate was 11.8%, compared to a 0% sexual recidivism rate for i
nonsexual, nonviolent offenders. Though these rates may appear low (i.e., the majority
of the sex offenders in this follow-up study did not recidivate), the actions of the other
sex offenders reduce the physical and mental health of their victims and their victim’s
families as well as the health and safety of the surrounding community. Examining the
variation in sex offenders’ behaviors and the factors associated with it may assist in
understanding how to further reduce this recidivism rate, assuming this goal is attainable.
The present study was not able to examine recidivism as an outcome because those data
were not available. However, future studies should consider recidivism as a treatment
outcome variable. Instead, the present study focused on sex offenders’ short term
response to treatment through measures of their attendance, engagement in treatment, and
measures of pre- and post-psychological functioning.

Sex oﬂenders ’ response to treatment. In light of these ﬁgures and the devastation

of even a single sexual offense to victims and society (Levenson & Cotter, 2005),

 

treatment of sex offenders has been a long-standing interest. Observers are continuing to
call for improvement to programs and more treatment response studies in order to gain
information that may help target those sex offenders who have a worse treatment
prognosis and higher recidivism rate (Ward & Gannon, 2006). However, these observers
are assuming that improving treatment programs or developing new innovative programs
are the keys to targeting the seemingly treatment-resistant sex offenders. The present
study does not accept that assumption. Rather, the present study assumes that not all sex
offenders will respond successfully to treatment (regardless of its design) and that having

knowledge of the internal mechanisms of the individual offender (e.g., motivation to

~‘D‘_-‘

change, psychopathology) may provide information about how each person may respond

 

to any treatment approach. This assumption emerges from a more idiographic approach
than has been used in the literature previously and attempts to embrace the variation and
heterogeneity of sex offenders’ response to treatment rather than assuming that, if the
treatment is designed well enough, all sex offenders should respond equally well to it.
Several treatment options are available to sex offenders, including
psychopharrnacology (Hill, Briken, Kraus, Strohm, & Berner, 2003), chemical castration
(Miller, 2003), motivation-enhancement therapy (Miller, Zweben, DiClemente, &
Rychtarik, 1992), and novel rehabilitation approaches (Ward & Gannon, 2006).
However, cognitive-behavioral therapy (CBT) has had the most success in rehabilitating
many sex offenders (Beech & Fisher, 2002). As with traditional CBT, CBT with sex
offenders focuses on maladaptive co gnitions and behaviors; however, there are a few
modiﬁcations that tailor the treatment to this group. The most tailored and

comprehensive CBT treatment programs for sex offenders, which usually take a group

 

therapy format (Barker & Beech, 1993), typically include elements of each of the
following: 1) behavior therapy to reduce inappropriate and increase appropriate sexual
arousal; 2) prosocial skills training and development; 3) restructuring cognitive
distortions and enhancement of empathy for the victim; and 4) relapse prevention (Abel,
Osborn, Anthony, & Gardos, 1992). (Details describing how each of these strategies is
implemented are provided in Appendix G).

Using these techniques with sex offenders has shown some promise in reducing
recidivism rates. For example, in two studies, while recidivism rates (for sexual and
nonsexual offenses) for untreated sex offenders range from 22% to 27% (Marques, Day,
Nelson, & West, 1994; McGrath, Hoke, & Vojtisek, 1998), the recidivism rates for sex
offenders treated with cognitive-behavioral therapy ranged from 8% to 19% (Marques et
al., 1994; McGrath et al., 1998) over a ﬁve— to six-year follow-up period. However the
apparently promising ﬁndings of both of these studies are deceptive: the sex offenders
were not randomly assigned to treatment or control groups. Rather, the groups comprised
offenders who volunteered for treatment and those who refused (Marques et al., 1994;
McGrath et al., 1998). Therefore, the treatrnent’s real effects apart ﬁ'om pre-existing
offender characteristics (such as motivation or degree of psychopathology) were unclear.
Those sex offenders who sought treatment may have been more intrinsically motivated to
stop offending than those who did not seek out treatment.

Appropriately addressing the issue of motivation in sex offenders would require a
study of a group of sex offenders who did not volunteer for treatment but, rather, were
assigned to a treatment program, regardless of their preference. Having a representative

sample of offenders in a study may help in the identiﬁcation of the types of offenders

 

who are actually ready for participation in a specialized sex offender treatment program
and, thus, ready for release to the community versus those who may need to remain in
prison for longer periods or be assigned to motivation-enhancing treatment. Therefore
motivation becomes one of two main foci in the present study (psychopathology being
the other focus). Before continuing it is necessary to clarify the conceptualization of

motivation to change as used in the present study.

 

 

MOTIVATION TO CHANGE IN SEX OFFENDERS

Motivation to change may be one factor important to clients’ participation and

success in treatment (Brogan, Prochaska & Prochaska, 1999). For this study, motivation

to change speciﬁcally refers to sex offenders’ self-reported intrinsic desire to terminate
their maladaptive sexual behaviors (modiﬁed from Blanchard, Morgenstem, Morgan,
Labouvie, & Bux, 2003). In studies on sex offenders, motivation to change has been
largely overlooked. It has too often been equated with the very different concept of
motivation for treatment. Treatment motivation is confounded by external motivating
factors such as early release from prison if the treatment program in completed
successfully (Lanyon, 2001). Therefore, while these two concepts are likely correlated,
they are not the same. Equating them may lead to misinterpretations of sex offenders’
intentions in treatment. In order for this research to move forward, it is important to
question the assumption that all sex offenders are intrinsically unmotivated to change
(Kear-Cowell & Pollack, 1997). Variation in intrinsic motivation Grereafter,
“motivation”) has been postulated as one reason why treatment responses are variable
with this population (i.e., some succeed well in treatment and others do not; Tierney &

McCabe, 2002). However, little is known clinically about variations in or determinants

of sex offenders’ actual desire to change their maladaptive sexual behaviors (Drapeau et

al., 2003; Terry & Mitchell, 2001).

Sex Offenders’ Heterogeneity: Psychopathology, Motivation, and Treatment Responses

Heterogeneity in Psychopathology and Motivation. Sex offenders may be

heterogeneous in a variety of areas, including psychopathology (Ahlmeyer et al., 2003;

IF .1: .3,—

 

 

Schlank, 1995).3 To examine the heterogeneity of sex offenders, one study identiﬁed
several different sex offender offense “pathways” (i.e., the interacting internal and
external factors which may inﬂuence their choices in committing an offense; Hudson et
al., 1999).4 Though they did not measure psychopathology per se, these authors did
describe internal factors of the offender (e.g., affect, cognitions) and found they were
related to different sexual assault styles (e.g., impulsive, planned) and to different levels
of motivation to change. A unique feature of this model as it applies to my study is that ~
these authors directly mention motivation to change as an integral part of the post-offense

thought process and make assumptions about variation in motivation to change in relation

 

to variation in the offense process. 1
The offense “pathways” model was derived from qualitative data in a method

known as grounded theory (Hudson et al., 1999). In their study Hudson and colleagues

qualitatively coded the verbal accounts of the offenses of 86 incarcerated sex offenders

(sex offenders against adults [N=14] and sex offenders against children [N=72]) in order

to derive offense types composed of affect, cognition, and behavior.5 The results

indicated eight types—three major and ﬁve minor types. Because 75% of the sampled

sex offenders were categorized in the three major types (Hudson et al., 1999), only those

types are described here.

 

3 In the present study, psychopathology refers to patterns of psychological disturbances oﬁen reﬂected in
maladaptive behavior (Hall & Hirschman, 1991).

4 Though the authors describe their model in terms of “pathways” which implies that any sex offender
could use any pathway to offend, the description of these “pathways” actually sounds more like a “type”
which implies that certain sex offenders offend in a speciﬁc manner only. The present study assumes that
the authors “pathways” are actually “types” of sexual offenders.

5 These methods were not clustering techniques. The descriptions of the offense were coded on six points
relating to a theoretical sexual offense chain which was developed in a previous study (Ward, Louden,
Hudson, & Marshall, 1995). After coding the descriptions, they matched the resulting offense proﬁles to
the theoretical stages on conceptual rather than empirical grounds.

 

The ﬁrst major type, comprised of 33% of the sample, was characterized by the
offender having positive affect prior to and during the offense (e.g., feeling excited,
powerﬁil, and/or loved; Hudson et al., 1999). The offender then devises a speciﬁc plan
for the assault (e. g., getting the children away from their mother; planning to tie up the
woman) and expresses an overt desire to have sex with the victim. During and after the
sexual assault, the offender has a distorted belief in the assault being consensual sex (e. g.,
“she was my partner”) and creates a positive reconstruction or evaluation of the offense
after it happens (e. g., “she wanted to have sex”). Finally, the offender appears to express
low motivation to change possibly because he has justiﬁed the offense as something both
he and the victim desired. Of the sex offenders characterized in this type, 82% were sex
offenders against children and 18% were sex offenders against adults, which is similar to
the base rate for the sample (prior paragraph). Based on these offenders’ affective
presentations, level of cognitive distortions, and lack of motivation to change, the authors
characterized this group as “appetitively-driven” to offend (i.e., they had an explicit
desire to offend and found ways to do so; Hudson et al., 1999).

i. The second major type, comprised of 16% of the total sample, was characterized
by the offender experiencing negative affect prior to the offense (e. g., depression,
loneliness) and then devising a speciﬁc plan for the assault (Hudson et al., 1999). During
and aﬁer the offense, the offender has an “energized,” positive affect (e.g., gratiﬁcation,
power, excitement). Also, rather than conceptualizing the offense as being consensual
and mutual, the offender focuses on having met his emotional and/or physical needs and
does not consider the victim’s feelings during or immediately after the offense. The

offender has a negative evaluation of the offense after it happens (e.g., disgust with

himself, regret, empathy for victim) and an increased motivation to change. Of the
offenders characterized in this type, 86% were sex offenders against children and 14%
were sex offenders against adults (Hudson et al., 1999), again, similar to the base rate for
the sample.

The third and ﬁnal major type, comprised of 24% of the total sample, was
characterized as having negative affect (e.g., depression, anger) throughout the entire
offense process (i.e., starting before and ending after the offense; Hudson et al., 1999). i
They do not experience any positive affect during the offense, unlike the ﬁrst and second

types of offenders. Also unlike the ﬁrst and second types, this type was characterized by

 

impulsive offending without the degree of planning in the other two groups. That is, sex
offenders in this type were more likely to commit sexual assault while committing
another offense such as larceny rather than explicit planning to seek out and victimize
someone (Hudson et al., 1999). Finally, this type involves the offender negatively
evaluating the offense after it happens and expressing increased motivation to change the
offending behavior. Of the offenders in this category, 67% were sex offenders against
children and 33% were sex offenders against adults (Hudson et al., 1999). When
considering the base rates for the sample, these results indicate that sex offenders against
adults are slightly more likely to be categorized in this pathway than are sex offender
against children.

The major strength of this model as it pertains to the current study is that it
incorporates the offenders’ expressed motivation to change as a part of the offending
process or cycle. It connects motivation to change to other offense processes (e. g., affect,

impulsivity) and does so using empirical means (Hudson et al., 1999). This model is

10

particularly relevant to the current study because it suggests that motivation to change
may differ depending on the psychological characteristics of the sex offender. The
results of this study provide some support for the hypothesis that psychopathology may
have a relationship with motivation to change.

Heterogeneity in Motivation and Treatment Responses. Motivation to change
may also be related to treatment response. Tierney and McCabe (2002) hypothesized that
treatment participation/attendance, treatment behavior (e.g., implementation of relapse
prevention strategies), treatment completion, and recidivism are each associated with
motivation to change. Other authors have made similar assertions. Looman and
colleagues (2005) reported that offenders with low motivation to change were more likely
to withdraw from treatment. They also reported that motivation to change is associated
with other treatment behaviors, including attendance and participation in treatment, use
of relapse prevention strategies, and recidivism. While these assertions are largely based
on anecdotal evidence, one formal study provided initial empirical support of a
relationship between motivation to change and treatment responses (Barrett, Wilson, &
Long, 2003). In a sample of incarcerated sex offenders in a specialized treatment
program, two features of motivation to change (i.e., admission of guilt and accepting
responsibility) were both associated with more successful treatment responses (Barrett et
al., 2003). However, one limitation of this study is that motivation to change and
treatment response were both coded by the sex offenders’ therapists. Thus, the data
source for both variables was confounded: therapists may have been biased toward seeing

more motivation in clients who they believed improved.

11

 

Thus, taking this information together, one may make two speculations. The ﬁrst
speculation is that variations in internal factors of the offenders such as psychopathology
may be lawfully organized into patterns, as was indicated in the Hudson et a1. (1999)
study. One could also speculate that the factors associated with variation in offense
patterns include variation in motivation to change. Variation in motivation to change
may, in turn, be related to different treatment responses in sex offenders.

Paralleling the Sex Offending and Substance Abuse Literatures to Conceptualize
Motivation

The conceptualization for motivation to change in the current study drew upon the
long-standing efforts to develop treatments for addictive behaviors such as alcohol or
drug use (Blanchard et al., 2003). Carrying the substance abuse model forward to sex
offending admittedly carries some hazards. For example, one potential concern of
paralleling sex offending to substance abuse is that, if overemphasized, sexual offending
can be viewed as an addictive behavior out of the offenders’ control. This is generally not
a viable model for sexual offending. For one thing, it could imply that it absolves the sex
offender of responsibility for his actions, something which would be troubling to say the
least. Further, the motivations behind sex offending may be quite different ﬁ'om those
involved in substance abuse. Therefore, the present study does not conceptualize sexual
offending as an addictive behavior. Rather this analogy from the addiction literature is
simply used as a way to begin to access the concept of motivation to change in a sample
of sex offenders, in the absence of alternative and perhaps more appropriate measures
that might be more directly applicable to sex offending. The current motivation to change

measure is thus seen as a ﬁrst step attempt.

12

 

The present study was not the ﬁrst to draw upon the substance abuse model of
motivation to change (e.g., Mann, 1996 as cited in Ginsburg, Mann, Rotgers, & Weekes,
2002). Unfortunately, the researchers who ﬁrst made the connection between these two
bodies of literature did so as a result of conceptualizing sexual offending as an addictive
behavior (George & Marlatt, 1989). As sex offending is no longer largely conceptualized
as an addictive behavior, current rationale for drawing upon the substance use model of
change is the similar rate of relapse in substance abusers and sex offenders as well as the
difficulty these populations have in beginning and maintaining change (Ginsburg et al.,
2002). However, these two bodies of literature differ substantially in the
conceptualization of their respective maladaptive behaviors in that sex offenders are not
psychologically or physiologically dependent on sex offending like substance abusers
may be dependent on drugs or alcohol. Further, in substance abuse, often the victim who
suffers the most is the abuser himself. However, in sex offending, the victims are
innocent people in the community. Because of these differences between the behaviors,
the sources of motivation to change these behaviors will likely differ substantially
depending on which behavior is being discussed. Thus, this parallel between the
literatures is quite limited and several concerns can be raised. More discussion will be
given to this topic in the Conclusions and Discussion section. The current study uses the
work in the substance abuse literature simply as a starting point in the absence of
alternatives. It is hoped that further research and understanding of motivation in sex
offenders will lead to a better conceptualization of this construct.

With these caveats in mind, what can be drawn from the substance use literature

regarding motivation to change? Substance use researchers long ago expanded their focus

l3

from the substance-abusing behaviors themselves to methods of changing those
behaviors, with a particular focus on the psychological components associated with that
change process (Blanchard et al., 2003). Among the key developments from this research
was the Trans-Theoretical Model of Behavior Change (TTM; Prochaska & DiClemente,
1982)

Parallels between the substance abuse literature and the sexual offending literature
have been drawn mostly in the form of relapse prevention models which focus on the
behavioral interventions that may result in a cessation of sexual offending (e. g., Laws,
2003; Price, 1999). However, these models may not explicitly incorporate strategies to
increase motivation to change among offenders (e. g., Price, 1999). More recently the
parallels between these two bodies of literature have expanded to include models of
motivation to change, including the TTM (Tierney & McCabe, 2002, 2005). Although
the motivation to change model needs some conceptual changes and, of course, empirical
support to successfully apply to sexual offending, it appears to provide a suitable
foundation for the conceptualization of sexual offending as a behavior which is difficult
to change and for which motivation would be an important component in changing,
analogous to substance abuse.

The Stages of Change from the IT M and Sex Oﬂenders. The TTM6, as applied to
sex offenders, served as the guiding model of motivation to change for the current study.
The central constructs of this model are the Stages of Change, Processes of Change, and

Decisional Balance.7 The focus in this study was on the Stages of Change which,

 

6 Figure A1 in Appendix A illustrates the theoretical ﬂow of this model, the details of which are beyond the
scope of this study.

7 Detailed descriptions of the Processes of Change and Decisional Balance as well as additional background
information on the origins of the Trans-Theoretical Model can be found in Appendix B.

14

according to the TTM, include the stages Pre-Contemplation, Contemplation,
Preparation, Action, and Maintenance.

Tierney and McCabe (2005) attempted to modify the TTM to apply it to sexually
offensive behavior consistent with the suggestions of previous researchers (e.g., Kear-
Cowell & Pollack, 1997). These modiﬁcations are followed here. They suggested that
the Pre-Contemplation process represents sex offenders’ degree of strong denial or
minimization of their maladaptive sexual behaviors (Kear-Colwell & Pollock, 1997;
Tierney & McCabe, 2005). The Contemplation process involves the sex offender
recognizing his sexual behavior as maladaptive and acknowledges limited desire to want
to change that behavior (Tierney & McCabe, 2005). Given the difﬁculty that sex
offenders have in changing their behaviors (Tierney & McCabe, 2005), conceivably this
process will occur to some extent throughout the course of treatment (Prochaska et al.,
1992). The Preparation process may lead sex offenders to think more critically about
their behavior and make a concrete decision to change their behavior (Tierney &
McCabe, 2005). However, in this conception the offenders do not start making concrete
behavior changes until their Action processes are activated (Kear-Colwell & Pollock,
1997; Tierney & McCabe, 2005). Lastly, the Maintenance process involves the
individual continuing to follow his relapse prevention strategies and strengthen his
internal motivation to remain changed (Tierney & McCabe, 2005).

In light of newer data, the original TTM was re—conceptualized in such a way that
these “stages” of change are no longer linear and distinct but are cyclical or parallel
processes that organize around a uniﬁed change process (Prochaska, DiClemente, &

Norcross, 1992). Therefore, they are conceptualized as co-occurring processes rather

15

than distinct stages (Prochaska etal., 1992). That is, a sex offender would not “ﬁnish”
the Pre-Contemplation stage and then begin the Contemplation stage, for example.
Rather, he is more or less committal to change at any given time in his life, depending on
the circumstances (e. g., currently in treatment; Prochaska et al., 1992). However, further
research needs to be conducted to determine if these processes can be understood as
reﬂecting a single latent motivation construct in sex offender populations. Also,
necessary is research examining the factors which may affect the change process such as,
for example, affective presentation and psychopathology. Or, perhaps motivation exists
outside of these factors and is not affected by these other processes within the offender.
Because motivation to change is currently understudied, using an exploratory approach
with broad concepts and questions may be useful in honing in on which factors, if any,
motivation comprises.

The present study thus empirically tested whether motivation to change is a latent
construct and whether it is associated with sex offenders’ treatment response. Because
motivation to change has not been validated in sex offender populations, this step is
necessary to move the ﬁeld closer to understanding motivation to change and thus
eliminate the need to borrow from the substance abuse literature in conceptualizing
motivation. Notably, the current measure used to assess motivation to change (i.e.,
URICA) has several limitations in its research and clinical application (which is further
discussed in the Results and Discussion sections). However, until more information and
empirical research draws conclusions about motivation to change, thus leading to the
development of an improved measure, then the use of this measure is viewed as a

necessary ﬁrst step.

16

Although additional exploration of the factor structure of Stages as reﬂecting
unitary or parallel processes needs to occur, some evidence from the substance abuse
literature suggests that the Stages of Change are valid predictors of behavior change at
least in some clinical groups. The research supporting this claim shows correlations
between the stage of motivation and the mechanisms involved in changing the behavior,
prior attempts to change, conﬁdence in one’s ability to change, and relapse potential
(Blanchard et al., 2003). These ﬁndings suggest that exploring the relationship between
motivation to change and treatment behaviors may be an important next step in sex
offender research. Further, if measures of motivation to change reﬂect one uniﬁed latent
construct or motivational proﬁle (rather than four separate processes or stages), then it
has some validity in that it predicts several features of treatment response.

While the suggestions of Tierney and McCabe (2005) appear relevant to sex
offenders, there are two key gaps in the application of this concept of motivation to sex
offenders. First, as already implied, the construct of motivation to change has not been
validated in sex offender populations and may be better understood as a single latent
variable. A factor analysis using a sample of substance users showed that Pre-
Contemplation loaded on one factor alone while the other three processes (i.e.,
Contemplation, Action, Maintenance) loaded together on another factor (Blanchard et al.,
2003). These factors were validated on treatment behaviors, indicating that those with
high Pre-Contemplation scores had higher treatment attrition rates (Blanchard et al.,
2003). These results provide preliminary support to the hypothesis that the “stages” of

change may actually be processes that functionas two constructs (i.e., Pre-Contemplation

l7

then Contemplation, Action and Maintenance) instead of four separate constructs. This
needs to be evaluated in sex offenders.

Second and more crucially, the model assumed that all sex offenders may
experience the change process in the same manner. For example, no distinctions have
been made about the application of the TTM to different types of sex offenders (e. g.,
those with and without severe psychological disturbances). Indeed, even if the processes
necessary to affect change are the same in all types of sex offenders (which is doubtﬁrl
given the variation that has already been noted in sex offenders’ responses to treatment
and other areas), the difﬁculties which may arise during this process may vary
signiﬁcantly based on offenders’ internal psychological characteristics. Therefore, in the
present study, one may hypothesize that at the beginning of treatment those offenders
who have certain psychopathology compositions will have different levels of motivation.
For example, if their psychopathology includes psychosis, then motivation may be low
because of the lack of insight into behaviors that is often associated with thought
disorders (APA, 2000) which may cause the offender to think that he has no maladaptive
behaviors to change. Or, if their psychopathology includes prominent depression
symptoms, then motivation may be low because of the pessimistic attitudes associated
with depression (APA, 2000) which may lead the offender to think that he cannot change
his behavior.

Theories of Sex Offending

Because little empirical evidence exists about the factors affecting motivation to

change in sex offenders (Tierney & McCabe, 2002), it is helpful to consider the literature

on theories of sex offending for clues about what characteristics may be important in

18

inﬂuencing motivation to change. When trying to understand the important sources of
variation among sex offenders which may inﬂuence treatment response and recidivism,
of course it is not enough to only examine motivation to change (Looman et al., 2005).
Instead, knowledge of the offender’s psychological proﬁle needs to be gathered as well
(Looman et al., 2005). For the purposes of this study, of particular relevance are theories
which address the function of psychopathology in sexual offending.

Excluded theories. The following discussion of theories of sex offending is
therefore selective. Many theories of sex offending are not germane to the concern about
motivation to change because they either do not address the psychological factors which
may be important to the change process, address only one aspect of psychopathology
(e.g., affective disorders), or mention psychopathology only in passing. These include:
Murphy, Coleman, Haynes, and Stalgartis’ (1979) Integrative Theory of Rape;
Finkelhor’s (1984) Four-Factory Theory of Child Sexual Abuse; Marshall and Barbaree’s
(1990) Integrated Theory of Sexual Deviancy; Hall and Hirschman’s (1992)
Quadripartite Theory of Child Molestation; Prentky and Burgess’ (1993) Biological
Theory of Repetitive Sexual Aggression; Schwartz and Masters’ (1993)
Psychodynamic/Trauma—Based Theory of Sexual Addiction; Lalumiere, Chalmers,
Quinsey, and Seto’s (1996) Mate Deprivation Hypothesis of Sexual Coercion; Ward and
Hudson’s (1998) Meta-theoretical Framework of Sexual Offending; and Ward and
Beech’s (2006) Integrated Theory of Sexual Offending, among others. While each of
these excluded theories could provide valuable information about the process of sexual
offending, such as the role of attachment (e. g., Schwartz & Masters, 1993), biological

factors (e.g., Prentky & Burgess, 1.993), and evolutionary considerations (e.g., Lalurniere

19

et al., 1996), a full discussion of all of these theories is far beyond the scope of this paper.
Instead, two theories which appear most relevant to the present study are discussed.

Included theories. Two theories were chosen as they relate to understanding the
mechanism associated variation in sex offenders and whether psychopathology may be
one possible source of that variation. The ﬁrst theory, the quadripartite model, considers
how psychopathology (among other factors) may function in a sequential progression to
produce sexually offensive behaviors (Hall & Hirschman, 1991). The second, more
recently developed theory, the pathways model, considers the ways in which dysfunction
in four psychological factors (i.e., intimacy deﬁcits, sexual arousal, emotional regulation,
and cognitive distortions) interact to produce ﬁve sexual offense pathways (Ward &
Siegert, 2002). Each theory is now considered further.

Hall & Hirschman ’s (1991) Quadripartite Model. A well-cited theory of sexual
aggression (against women) proposed that sexual violence has its origins in four
etiological factors within the individual: physical arousal, cognitive distortions, affective
dyscontrol, and dysﬁrnctional personality which interact in one model (Hall &
Hirschman, 1991). Physical sexual arousal is theorized to begin the process of sex
offending in this model. Sexual arousal of course could be harmless or harmful
depending on the context of the situation. However, for those with a propensity toward
maladaptive sexual behaviors, once sexual arousal is activated, they may have difﬁculty
inhibiting that response or developing a more socially-appropriate and less harmful way

of handling that response (Barbaree & Marshall, 1991).8 This physical sexual arousal, in

 

8 It should be noted that those offenders with a propensity toward maladaptive sexual behaviors have a
history of being sexually abuse and/or connecting their sexual desires to aggressive behaviors (e. g., rape
fantasies; Hall & Hirschman, 1991). Thus, this theory is not proposing that all men have the potential to
rape or have difﬁculty controlling their sexual desires.

20

some individuals, is proposed to interact with cognitive distortions about women (e. g.,
rape-supportive beliefs such as “women are sexual objects”) which may justify the
offense behavior (Hall & Hirschman, 1991). These could also be distortions about the
probability of being arrested (i.e., “I will not be caught”). As the cognitive distortions
become active, affective control, which involves the ability to regulate internal emotions,
particularly negative emotions, decreases. When negative emotions such as frustration or
anger from being sexually rejected by a woman (for example) become potent enough to
suppress appropriate emotional reactions to physically harming someone (e.g., guilt,
shame), sexual violence is more likely to occur (Hall & Hirschman, 1991). Irnportantly,
psychopathology is conceptualized as interacting with each of the three previously
mentioned components. Hall and Hirschman (1991) hypothesized that early
developmental experiences with parents may have contributed to the development of
antisocial personality characteristics in sex offenders. These characteristics vary by
offender depending on the type of childhood experiences. More so than the other factors,
psychopathology is hypothesized to be centrally causal in the operation of sexual
offending (Hall & Hirschman, 1991). This theory suggests that psychopathology may be
one source of variation in sex offenders’ behaviors.

One advantage of this model is that it highlights the manner in which different
aspects of psychopathology (e. g., emotions, cognitions) may work together in the offense
process. Having a theory which not only deconstructs the factors involved in the offense
process but also attempts to present them in a sequential manner allows researchers and
clinicians to hypothesize about where in the offense process motivation to change may be

most negatively affected. For example, one may hypothesize that the initial cognitive

21

 

distortions have a more negative inﬂuence on motivation to change than affective control
because they allow the offender to justify his actions against the victim (e. g., “She was
playing hard to get”). Though this hypothesis was not tested in the current study, a
researcher could develop a study to test the relationship between cognitive distortions and
motivation to change while a clinician could target cognitive distortions during treatment
in an effort to increase motivation to change.

One critique of this model is it focuses on state-like or temporary factors such as
physiological and emotional arousal and does not consider enough of the trait-like or
predisposing stable factors (e. g., personality traits) which may be inﬂuencing the
offensive behaVior (Ward & Hudson, 1998). Hall and Hirschman (1991) do consider
personality factors and psychopathology but they do not fully explain the manner in
which these factors may be inﬂuencing offensive behavior (Ward & Hudson, 1998).
However, the components of offending included in this model can be used to justify
exploring a connection between psychopathology and motivation to change, as
mentioned.

Ward & Siegert ’s (2002) Pathways Model. A subsequent theory was developed
through combining several well-cited theories of sex offending of the early 19903. Ward
and Siegert (2002) sought to create a comprehensive model of sex offending against
children through combining the strengths of each several theories and, in doing so, hoped
to compensate for each theory’s individual weaknesses. The resulting theory presented
ﬁve pathways in which four psychological mechanisms were featured: intimacy deﬁcits,
sexual arousal, emotional regulation, and cognitive distortions. Unlike the Hudson et al.

(1999) study discussed previously, Ward and Siegert (2002) are proposing pathways (as

22

opposed to typologies) and assume that any offender could use any of these pathways in
the commission of a sexual offense. These authors further assume that once an offender
has used a particular pathway, he is not guaranteed to only use that pathway in the
commission of another offense. The authors propose that all four of these factors are
working together in the offense process but the prominence of one factor over the others
determines the offender’s reason for offending. In the ﬁrst four sexual offense pathways,
one of each of these factors is prominent over the others and is theorized to inﬂuence the
resulting sexual offense. The ﬁfth pathway involves having balance (i.e., equal
deﬁciencies) among the four mechanisms (Ward & Siegert, 2002). The authors do not
operationalize when one of these mechanisms is problematic and when it is not nor do
they offer ideas of how these mechanisms would be measured. Thus, it is unclear if
having “normal” levels of these mechanisms means having levels which are comparable
to the general population or to nonsexual offenders.

The ﬁrst pathway is characterized by intimacy deﬁcits primarily (Ward & Siegert,
2002). These intimacy deﬁcits could be due to an insecure attachment and social
ineptness. Thus it was proposed that sexual offending against children may take place
because these offenders do not know how to socially engage with adults and form
appropriate sexual relationships with adults. The authors proposed that offenders who
use this pathway may offend when they feel the need for intimacy (emotional or
physically) but their lack of social skills interferes with their ability to form a healthy
adult relationship where those needs can be met (Ward & Siegert, 2002).

The second pathway involves having inappropriate sexual arousal or having

sexual arousal in situations that the general population would not usually experience

23

arousal. The authors proposed that early childhood sexual abuse caused difﬁculty in sex
offenders being able to distinguish when it is appropriate to be sexually aroused. The
offenders who use this pathway may seek out sexual partners when they (i.e., the sex
offenders) are in a negative emotional state (e. g., anger), may chose inappropriate sexual
partners (e. g., children), or may engage in risky sexual behaviors without the consent of
their partner (Ward & Siegert, 2002).

The third sexual offense pathway which Ward and Siegert (2002) proposed
involved problems with emotional regulation. These offenders may have problems
identifying emotions, self-regulating their emotions or using social supports to help
process difﬁcult emotions, and difﬁcult controlling emotional outbursts such as anger.
For offenders who use this pathway, sexual offending is caused by deﬁcits in emotional
and behavioral control or using sex as a coping mechanism.

The fourth pathway involves having cognitive distortions which may lead to
sexual assault (Ward & Siegert, 2002). These offenders may not have inappropriate
sexual fantasies but may have antisocial attitudes which justify criminal behaviors. They
may also endorse rape-supportive beliefs (e. g., children are sex objects) which
speciﬁcally justify sexual aggression. The ﬁnal pathway was described as a “multiple
dysfunctions” pathway because it involved the sex offender having deﬁcits in each of the
four previously mentioned psychological characteristics. The authors proposed that these
offenders were the most severely disturbed and had a variety of problems with insight
and judgment as a result of the multiple deﬁcits (Ward & Siegert, 2002). As a result,

their pathway or motives for offending may be less predictable or easily categorized.

24

Similar to the Hall and Hirschman (1991) theory, the relevance of this theory to
the present study is that it considers that manner in which different psychological factors
may be working within the offender and leading to variations in behaviors—helping
characterize variation. One limitation of the Ward and Siegert (2002) theory is that it
was developed only for application to sex offenders against children. One possible
explanation is that they did not think that it would. be possible to create an overarching
theory for sex offenders against adults (Brown, 2005). However, it would be preferable
to consider the full range of offender variation.

Summary. Several other theories of sex offending could have been discussed (see
above cited theories). The two theories which were emphasized addressed variation in
sex offenders’ psychological factors or psychopathology. These theories provided some
support to the assumption of the present study that clusters of sex offenders may be
formed as a result of variation in psychopathology. The ﬁrst theory highlighted the role
of psychopathology to the sex offending process (Hall and Hirschman, 1991), while the
second theory describes differential pathways for sex offending based on psychological
factors (Ward & Siegert, 2002). Though no one theory was available which could fully
link psychopathology to motivation to change, the importance of these theories as they
relate to the present study rests in their explanation of the variation in sex offenders and
the implications that psychopathology may be one source of that variation. Further,
though these theories do not represent a majority position in the sex offending literature,
together they suggest that sex offenders may form subtypes based on psychological

characteristics and those types may have differential meaning for motivation to change.

25

Treatment Implications of Studying Motivation to Change

The treatment implications for the present study were part of what motivated the
study. The study was intended to evaluate whether using motivation to change as a
central feature in the rehabilitation process of sex offenders might be beneﬁcial. For
example, one of several methods which could be used for this purpose is motivational
interviewing, which is a client-centered approach to therapy with the goal of enhancing
intrinsic motivation to change in clients who are not successfully engaging in traditional
treatment approaches (Arkowitz & Miller, 2008). These clients may appear to be
resistant to treatment but, while some of them are resistant, others of them are assumed to
be ambivalent about changing and in need of encouragement to continue in the change
process (Arkowitz & Miller, 2008). The premise of the present study was that correctly
identifying the psychopathological status of the sample of sex offenders would be
valuable in understanding their motivation and predicting their response to treatment.
One advantage of using motivational interviewing as opposed to other motivation-
enhancing techniques is that it can be incorporated into any number of established
treatment programs, included CBT (Farbring & Johnson, 2008). However, because of the
heterogeneity that manifests in sex offender populations (Schlank, 1995), motivational
interviewing (or at least the same type of motivation-enhancing techniques) may not
work as well with all sex offenders. The present study sought to determine whether
different clusters of offenders may have distinct motivational stances based on their
psychopathology. (Detailed descriptions of motivational interviewing can be found in

Appendix H).

26

Conclusion

The preceding review highlighted the importance of variation in sex offenders’
criminal and clinical presentations, which may be related to their variation in treatment
responses. One potential source of variation in treatment responses is motivation to
change. Motivation is an under-explored aspect of sex offender populations which may
have important implications for treatment. Another potential source of variation in
treatment responses is the variation in psychological characteristics or psychopathology
that appears evident in sex offender samples. Despite the number of theories which are
available to explain sexual offending, few of them sufﬁciently explicitly address the
inﬂuence of psychopathology on the offense process and still fewer explicitly address
motivation to change. Because of this, both of these factors need to be evaluated in more
depth to better understand their ﬁrnctioning and potential treatment implications. A
typology approach which can appreciate the heterogeneity of this population may be

appropriate for such a task.

27

PSYCHOPATHLOGY AND SEX OFFENDER HETEROGENEITY

As already noted, psychopathology may be an important part of the variation
which seems to reappear in sex offenders and some of the sex offender research provides
evidence to this effect. Several prior attempts to explore psychopathology heterogeneity
set the stage for the present study and are now discussed. Hall and Hirschman (1991)
attempted to integrate it into their offense process model and other studies have used it to
examine subtypes of offenders with implications for treatment (discussed below). Sex
offenders tend to have heterogeneous psychopathological proﬁles (Schlank, 1995).
Therefore studying them in subtypes based on their psychopathology may be an
appropriate method to use in order to understand how sex offender pathology and
motivation to change may function.
Overall Prevalence Rates of Psychopathology in Sex Offenders

Sex offenders in general often experience some types of psychopathology even
more frequently than other types of offenders. Compared to nonsexual offenders, sex
offenders have higher prevalence rates on the DSM-IV personality disorders Schizoid,
Schizotypal, Narcissistic, Obsessive-Compulsive, Dependent, and Avoidant (Ahlmeyer,
Kleinsasser, Stoner, & Retzlaff, 2003). These rates vary from 7% (Schizotypal
Personality Disorder; 3% in nonsexual offenders) to 37% (Avoidant Personality Disorder;
25% in nonsexual offenders). For Antisocial Personality Disorder, which is commonly
associated with criminal behavior (APA, 2000), sex offenders have rates of 24% while
nonsexual offenders have rates of 29% (Ahlmeyer et al., 2003). This result is not

surprising considering that the Ahlmeyer et a1. (2003) sample of nonsexual offenders

28

included violent offenders which the authors report sometimes have higher rates of ASPD
than nonviolent offenders.

Further, in incarcerated sex offenders, researchers have found rates of 35%, 22%,
and 15% for current clinical levels of depression, anxiety, and co-morbid depression and
anxiety, respectively (Stinson et al., 2005). These rates, though not necessarily higher
than in other offenders, are higher than expected in the general, non-clinical population,
in which one would ﬁnd 12-month rates of 6.7% prevalence for adults (males and
females) with depression and 18.1% prevalence for adults with anxiety (Kessler et al.,
2005). In terms of more severe psychopathology, sex offenders have rates of 4% for
thought disorders (compared to 1% in nonsexual offenders); for delusional disorder,
sexual and nonsexual offenders had equal rates (2%; Ahlmeyer et al., 2003). Given the
rates of these emotional and personality disorders in this population, it would appear that
behavioral differences, including those related to motivation to change and treatment
response, may be related to the type and level of psychopathology evident in the offender.
Prior Attempts to Cluster Sex Offenders Based on Their Psychopathology

Psychopathology may be a key factor in characterizing and understanding the
observed differences in sex offenders’ offense patterns and responses to treatment
(Drapeau et al., 2003). Because of the different emotional and behavioral patterns
associated with each type of psychopathology (e. g., Lee et al., 2001), one might
hypothesize that sex offenders may cluster into meaningful subtypes based on their
psychopathology proﬁles. Indeed, in the early 19903, ﬁve studies were published which
used cluster analysis with sex offenders to pursue this possibility (Duthie & McIvor,

1990; Kalichman et al., 1992; Kalichman, Shealy, & Craig, 1989; Shealy et al., 1990;

29

Schlank, 1995). Each of these studies sought to create homogenous clusters of sex
offender subtypes based on their psychopathology characteristics (as measured by the
MMPI). The ﬁndings from each of the four clustering studies which used the MMPI as
their measure of psychopathology will be presented followed by a brief description of the
Schlank (1995) study which did not use the MMPI as its clustering measure. Then the
overall strengths and weaknesses of these studies will be presented in the Interpretation
of Results section below.

Kalichman et al. (1989). Kalichman and colleagues (1989) analyzed the
psychopathology proﬁles of 120 incarcerated sex offenders against adult women. Using
the MMPI t-scores as their measure of psychopathology, the authors found a ﬁve-cluster
solution best ﬁt their data. They interpreted their clusters based on the number of t-score
elevations above 70. The ﬁrst cluster of individuals (33% of total sample) had all scales
within normal clinical ranges (i.e., t-scores were below 70). They were likely to have
known their victims and to have committed the sexual assault during the commission of
another crime. They also had the lowest maladaptive sexual arousal scores (i.e., they
were not sexually aroused by thoughts or images of sexual assault). The authors
concluded that this cluster of offenders had sexual assault as a secondary motivation in
their criminal activity.

Elevations in depression, antisocial behaviors, paranoia, and schizophrenic
symptoms characterized the next cluster of offenders (10% of total sample; Kalichman et
al., 1989). The authors found this cluster to be associated with cognitive distortions and
propensities toward rape. The authors concluded that this cluster of offenders had rape as

a primary motive and were likely to not know their victims. The third cluster (33% of

30

total sample) showed a single elevation on the antisocial behavior scale, with all other
scales within normal limits. Like the ﬁrst cluster, this group of offenders was likely to
commit their sexual assault during the commission of another crime and show low levels
of maladaptive sexual arousal.

Individuals in cluster four (18% of total sample) displayed elevations in antisocial
behavior, schizophrenia symptoms, and hypomania symptoms (Kalichman et al., 1989).
This cluster showed the widest range of maladaptive sexual interests and behaviors,
including knowing their victim, thinking more frequently about rape, showing more
maladaptive sexual interests, and having more disturbed thought processes. Similar
patterns were found in cluster ﬁve, but were more severe. The ﬁfth cluster (8% of total
sample), which showed the most severe psychopathology, had exaggerated response
styles, generalized anxiety, and schizophrenia symptoms. This cluster of offenders also
had the highest levels of substance use, became sexually aroused to thoughts of rape, and
reported several maladaptive sexual thoughts and behaviors.

Duthie and McIvor ( I 990). Duthie and McIvor (1990) evaluated the
psychopathology clusters of 90 convicted, pre-sentence sex offenders against children.
Instead of using the MMPI t-scores as in the other reviewed studies, they used the raw
scores from each of the MMPI subscales in their cluster analyses and then described
elevations of the clusters in terms of t-scores greater than 70. In their preliminary
analyses, hypomania symptoms did not contribute to the analyses and, thus, were
removed from the subsequent clustering. In the end, they found an eight-cluster solution

to best ﬁt the data (Duthie & McIvor, 1990).

31

The ﬁrst cluster (9% of total sample) was characterized by depression, antisocial
characteristics, and generalized anxiety (Duthie & McIvor, 1990). Heightened
depression and antisocial characteristics, together, is a common pairing in sex offenders
(Anderson & Kunce, 1979; Erickson, Luxenberg, Walbeck, & Seely, 1987), though it
was present in only 9% of the Duthie and McIvor (1990) sample. In previous research,
this pairing has been associated with intoxication during the offense and physical
penetration of the victim (Anderson & Kunce, 1979). In the Duthie and McIvor (1990)
study, offenders in this cluster were more likely to have female child victims (as opposed
to males or both sexes as child victims) and 25% of these offenders had a prior criminal
history of sexually offending against children.

The second cluster (23% of total sample) was the most frequently occurring type
of cluster and was characterized by antisocial characteristics and low levels of masculine
characteristics (Duthie & McIvor, 1990). This pairing is also common in sex offender
populations (Duthie & McIvor, 1990; Erickson et al., 1987). In Duthie and McIvor
(1990), offenders in this cluster offended primarily against female children (52%) and
fewer had prior sexual offense convictions (14%). Cluster three (9% of the total sample)
was characterized by hypochondriasis symptoms, antisocial characteristics, hysteria
symptoms, generalized anxiety, and schizophrenia symptoms (Duthie & McIvor, 1990).
This proﬁle is unique to this study. It was associated with having female victims (78% of
this cluster), being intoxicated at the time of the offense (40%), and a low rate of prior
convictions for sexual offense (10%).

Cluster four (7% of total sample) was characterized by antisocial characteristics,

paranoia, and schizophrenia symptoms (Duthie & McIvor, 1990). Offenders in this

32

cluster were also characterized by an exaggerated response style. The elevations of these
characteristics together are frequent in sex offenders (Anderson & Kunce, 1979; Erickson
et al., 1987) and, in previous research, were associated with intoxication during the time
of offense, prior convictions, and juvenile delinquency (Anderson & Kunce, 1979). In
the Duthie and McIvor (1990) study, 33% of this cluster had prior convictions or arrests
for sexually offending against children and 50% offended exclusively against female
children. The other 50% offended against both male and female children; none of them
were intoxicated during the offense.

In the ﬁfth cluster (18% of total sample), only antisocial characteristics were
elevated; all other psychopathology was sub-clinical (Duthie & McIvor, 1990), similar to
cluster three of the Kalichman et al. (1989) study. Thirty-one percent of these offenders
have prior convictions (Duthie & McIvor, 1990). No other characteristics distinguished
this cluster of offenders from others. Eight percent of the total sample was included in
the sixth cluster which was characterized by sub-clinical levels of hysteria and antisocial
characteristics (Duthie & McIvor, 1990). This cluster was also characterized by a
repressive response style and under-reporting psychopathology. Fourteen percent of the
offenders in this cluster had prior convictions for sexual and/or nonsexual crimes.

The last two clusters were both characterized, in part, by schizophrenia symptoms
(Duthie & McIvor, 1990). In addition to these symptoms, cluster seven (9% of total
sample) was also characterized by low masculine characteristics and heightened
symptoms of social introversion. Interestingly, this was the only cluster in this study
which did not include elevated antisocial characteristics. Ten percent of the offenders in

this cluster had prior convictions; 90% offended primarily against female children. In

33

addition to the schizophrenia symptoms, cluster eight (13% of total sample) was
characterized by depression, antisocial characteristics, generalized anxiety, and an
exaggerated response style (Duthie & McIvor, 1990). Eighty-three percent of these
offenders had prior convictions, 58% were intoxicated at the time of the offense, and 73%
offended primarily against female children (Duthie & McIvor, 1990).

Shealy et al. (1990). Shealy et al. (1990) examined 90 incarcerated sex offenders
against children using the MMPI. Though Shealy and colleagues used t-scores instead of
raw scores like the Duthie and McIvor (1990) study, they did not require their t-score
elevations to surpass 70 in order to be interpretable in this study. Therefore, elevations
were interpreted in relation to each other in one proﬁle rather than in comparison to the
overall clinical cut-off of the measure. This point is important to note in terms of how
these clusters are compared to the cluster proﬁles of other studies.

Of the four resulting clusters, the ﬁrst group (50% of total sample) was
characterized by antisocial characteristics and hypomania symptoms, though these
elevations were sub-clinical (Shealy et al., 1990). This cluster was associated with high
self-esteem, antisocial behavior, and impulsivity but not maladaptive sexual behavior.
The second cluster (19% of total sample) was characterized by clinical levels of paranoia;
four other scales (i.e., hypochondrias, depression, hysteria, and antisocial behaviors) were
sub-clinical but had nearly equal t-scores. Individuals in this cluster had low sexual and
psychological disturbances but were resentful, suspicious, and guarded (Shealy et al.,
1990).

The third cluster (18% of total sample) showed elevations on antisocial

characteristics, paranoia, and schizophrenic symptoms, all of which were associated with

34

hostility and poor judgment (Shealy et al., 1990). This cluster was also associated with
increased anxiety, anger, obsessions with sex, and disturbing sexual thoughts. The ﬁnal
cluster (13% of total sample) had the greatest levels of psychopathology in exaggerated
response styles, paranoia, generalized anxiety, and schizophrenic symptoms (Shealy et
al., 1990). This cluster had higher levels of anxiety and anger, was most likely to need
mental health services, and was associated with sexually disturbing behaviors.

Kalichman et al. (1992). In this study, MMPI psychopathology t-scores of 110
outpatient sex offenders against children were examined; a ﬁve-cluster solution ﬁt the
data the best (Kalichman et al., 1992). They again based their interpretations of the
clusters on the number of t-score elevations above 70. Two of the clusters had all scales
within normal clinical ranges (i.e., t-scores were below 70). One of these clusters was
associated with primarily offending against female children (14% of total sample). The
other was associated with a tendency to deny or minimize symptoms (35% of total
sample). The third cluster (26% of total sample) in this study had elevations in antisocial
behavior and low masculine characteristics, which was associated with normal
psychosexual functioning and offending against male children. The traits of this cluster
parallel those found in the ﬁrst cluster of Kalichman et al. (1989).

The last two clusters were associated with severe psychological disturbances, with
clinical elevations on nearly all scales (Kalichman et al., 1992). One of these clusters
(19% of total sample), which was characterized by exaggerated response styles,
depression, antisocial characteristics, low masculine characteristics, paranoia, generalized
anxiety, schizophrenic symptoms, and social introversion, was associated with severe

psychological distress, cognitive disturbances, and low sexual functioning. The last

35

cluster, which was characterized by antisocial behavior, low masculine characteristics,
paranoia, schizophrenic symptoms, and hypomania symptoms, was suggestive of
frequent aggression and impulsivity (10% of total sample). They also had extreme levels
of unusual sexual interests (e.g., fetishes, sadomasochism) and the highest levels of
sexual aggression of all ﬁve clusters. The authors likened this proﬁle to cluster ﬁve of
the Kalichman et al. (1989) study.

Schlank(1995). This study is different from the rest of these studies because it
did not use the MMPI to cluster; it used the Multiphasic Sex Inventory (MSI; Nichols &
Molinder, 1984). The MSI is a measure of sex offender psychopathology (e.g., cognitive
distortions, rumination of sex offending) and has been used to validate the clusters in
three of the four previously described clustering studies (i.e., Kalichman et al., 1989,
1992; Shealy et al., 1990). Because the proposed study has chosen to study
psychopathology more broadly in sex offenders (as measured by the MMPI) rather than
focusing on sexual psychopathology speciﬁcally, the details of the sexual
psychopathology clusters do not seem particularly relevant to this discussion.9 However,
Schlank (1995) has relevance in this discussion because it using a mixed sample of sex
offenders (i.e., sex offenders against adults and sex offenders against children) as
opposed to a sample of only one type of offender as with the other cited studies. This
study provides an alternate way of thinking about sampling procedures when clustering
sex offenders.

Schlank (1995) clustered a mixed sample of 164 incarcerated sex offenders based

on their psychopathology on the MSI. The sample included 80 sex offenders against

 

9 The reasons for using psychopathology more broadly-deﬁned (rather than sexual psychopathology
speciﬁcally) is provided in the Methods section below.

36

adults, 81 sex offenders against children, and 3 sex offenders with both genders as
victims. Seven different psychopathology clusters emerged from the sample (Schlank,
1995). Of these seven clusters of offenders, four were distinguished ﬁom the other
clusters by their victims (i.e., offenders against adult females; incest offenders against
female children; non-incest offenders against female children; mixed set of offenders
against multiple types of victims). Because three of the seven clusters were not deﬁned
by victims, the author concluded that there was a type of sex offender taxonomy which
existed independently of victim status (Schlank, 1995). Thus, studies which examine
only one type of offender may be overlooking these important subtypes.

Interpretation of results of prior cluster analytic studies. The following chart

summarizes the results of the reviewed studies:

 

 

Study N Sex Offender Clusters MMPI Scales
Kalichman 120 Incarcerated, 5 Cl: normal limits
et al. Adult victims C2: D, Pd, Pa, Sc
(1989) C3: Pd
C4: Pd, Sc, Ma
C5: F, Pt, Sc
Duthie & 90 Outpatient, 8 C1: D, Pd, Pt
McIvor Child victims C2: Pd, Mf
(1990) C3: Hy, Pd, Hs, Pt, Sc
C4: Pd, Pa, Sc
C5: Pd
C6: Hy, Pd, K, L
C7: Sc, Mf, Si
C8: Sc, D, Pd, Pt, F
Shealy et 90 Incarcerated, 4 Cl: Pd, Ma
al. ( 1991) Child victims C2: Pa, Hy, D, Hs, Pd

C3: Pd, Pa, Sc
C4: F, Pa, Pt, Sc

Kalichman l 10 Outpatient, 5 Cl: normal limits
et al. Child victims C2: normal limits
(1992) C3: Pd, Mf

C4: F, D, Pd, Mf, Pa, Pt, Sc, Si
C5: Pd, Mf, Pa, Sc, Ma
Schlank 164 Incarcerated, 7 n/a
(1995) Mixed victims

 

37

Overall, it seems that the ﬁndings from these studies produce somewhat
consistent results in supporting the utility of the MMPI in clustering techniques with sex
offenders. One study yielded a four-cluster solution (Shealy et al., 1990), two studies
yielded a ﬁve-cluster solution (i.e., Kalichman et al., 1989, 1992), and one yielded an
eight-cluster solution (Duthie & McIvor, 1990). While none of the clusters in any of
these studies had the exact elevations on the exact same traits, conceptually, these studies
had somewhat similar ﬁndings regarding the types of psychopathology clusters which
were evident in their samples. The conceptual parallels between these studies are
particularly interesting considering that one used a sample of sex offenders against adults
(Kalichman et al., 1989) and the others used samples of sex offenders against children
(Duthie & McIvor, 1990; Kalichman et al., 1992; Shealy et al., 1990). Perhaps these
similar clusters are reﬂective of a higher-order conceptual similarity among sex offenders
which varies by individual psychopathology traits of the offender. Also, the overlap in
results across these studies, as well as the results from the Schlank (1995) study, suggests
that sex offenders against adults and sex offenders against children may be more similar
than different in their psychopathology proﬁles.

Concerning the nature of the clusters, some of the MMPI elevations were
consistent for all of the studies reviewed: paranoia, depression, antisocial behaviors,
schizophrenic symptoms, and generalized anxiety. The consistency in elevation of the
schizophrenia scale across these studies (regardless of the victim status as adult or child)
was unexpected given that it is most commonly elevated scale (along with the antisocial
deviance scale) in sex offenders against adults but not sex offenders against children

(Butcher & Williams, 2000; Megargee, 2006). Also, of the four MMPI clustering studies

38

reviewed, hypomania symptoms were elevated in two (i.e., Kalichman et al., 1989, 1992).
The consistent presence of each of these scales across these four studies suggests that
they may be prominent features of sex offenders’ psychopathologies.

The consistent ﬁndings regarding the presence of MMPI scales of paranoia,
depression, antisocial characteristics, and impulsivity symptoms align with those in a
more recent study (non-cluster analysis) which found that these four factors were the
most commonly elevated factors in criminal populations (Megargee et al., 1999). These
results also parallel those of previous research assessing the MMPI two-point code
elevations of sex offenders (Anderson & Kunce, 1979; Erickson et al., 1987). In a non-
cluster analytic study of 403 convicted, incarcerated sex offenders (mixed sample of sex
offenders against children and sex offenders against adults), antisocial characteristics
were elevated in seven of the eight code type pairings (Erickson et al., 1987). Antisocial
characteristics were paired with depression, paranoia, and hypomania, consistent with
Megargee et al. (1999), as well as with masculine/feminine characteristics, generalized
anxiety, and schizophrenia symptoms, which parallel the ﬁndings of the four MMPI
clustering studies.

In addition to the consistency among the traits of interest in these studies, this
review also suggests that psychopathology is consistently related to certain offense
characteristics. The relationship to the victim (Kalichman et al., 1989), criminal history
(Duthie & McIvor, 1990), physical penetration of the victim (Anderson & Kunce, 1979),
gender of the victim (Duthie & McIvor, 1990), and intoxication during the offense

(Anderson & Kunce, 1979) all appear to be important offense characteristics which may

39

distinguish psychopathology clusters from one another and create a richer description of
the offenders in each cluster.

Based on this information, the four MMPI clustering studies reviewed here
suggest the importance of seven psychopathology characteristics (i.e., depression,
antisocial characteristics, masculine/feminine characteristics, paranoia symptoms,
generalized anxiety, schizophrenia symptoms, and hypomania symptoms) in sex offender
samples and expand on the understanding of these characteristics through clustering
offenders into meaningful subtypes.

However, these clustering studies have several limitations. One concerns their
sample composition. Each of these studies used one type of sex offender in their
samples—either sex offenders against adults (Kalichman et al., 1989) or sex offenders
against children (Duthie & McIvor, 1990; Kalichman et al., 1992; Shealy et al., 1990)——
making their samples only selectively represented.'0 The authors justify the use of only
select portions of the sex offender population on the grounds of needing to explore the
heterogeneity within each sex offender group. Indeed, the results from each study
indicate signiﬁcant variation in psychopathology even within one type of offender.
However, the review of these studies suggests that the overlap in psychopathology
between these two types of offenders may be more prominent than the differences
between them. Thus, at best, the reviewed clustering studies provided information on
some of the unique psychopathology features of each sex offender type and then
replicated each other in terms of general features of psychopathology within the sex

offender population. Knowing the unique psychopathology features of subtypes of

 

'0 Additional information on how psychopathology (i.e., individual disorders such as depression or
Antisocial Personality Disorder) differs by victim status in sex offenders can be found in Appendix C.

40

offenders may be useful for some research questions, but in exploratory studies
examining general patterns of psychopathology, combining the offender types has the
advantage of exploring the overlap in psychopathology among sex offenders generally
and devising implications for a wider range of offenders.

Another weakness of these studies is that they do not provide implications for the
clinical meaning or application of the results. Each of the clusters which emerged in
these studies was compared based on the offense characteristics and speciﬁc types of
sexual psychopathology (e. g., rumination on sex). However, only one of these studies
examined the clinical relevance of the emerging clusters. Kalichman and colleagues
(1989) conducted a follow-up study to examine differences in treatment attendance
among the ﬁve clusters (Kalichman, Shealy, & Craig, 1990). In this study, they used a
sub-sample of the original 120 sex offenders against adults from the 1989 study for a
total of 55 offenders who participated in a treatment program while incarcerated
(Kalichman et al., 1990). Results indicated that those offenders in clusters four and ﬁve
(i.e., those with more severe psychopathology) were more likely to attend treatment while
those in cluster one (i.e., those with the least severe psychopathology) had the lowest
rates of attendance (Kalichman et al., 1990). The authors interpreted these ﬁndings as
indicating that those who were more severely disturbed were more motivated to attend
treatment because they were under more distress and needed relief. Thus
psychopathology clusters are related to treatment behaviors and, thus, may have clinical
relevance for sex offenders. However, no study assessed motivation to change or
treatment completion, which may also differ by psychopathology clusters and provide

even more clinical information about the function of psychopathology in this population.

41

A ﬁnal critique of these studies concerns their sample sizes. The studies started
with small samples and, when they ﬁnished their clustering, the result was a small
number of offenders in each cluster. For example, Duthie and McIvor (1990) in
particular had cluster sizes of six or eight offenders because they had a total sample size
of 90 and found eight clusters. These small ntunbers in the clusters may be one reason
why Duthie and McIvor (1990) found cluster types that did not overlap with the other
studies. Having this small number of offenders per cluster may make it difficult to draw
reasonable conclusions about the psychopathology subtypes in their samples. Having a
larger sample size may increase the number of offenders per cluster and strengthen the
internal validity of the study.

Summary. This discussion has highlighted previous attempts at clustering
psychopathology in sex offender samples. While these studies have shown that clusters
of psychopathology types can be formed and replicated to a certain extent in this
population, problems with insufﬁcient representation of sex offender types, clinical
correlates, and sample size need to be improved upon to validate and frilly evaluate the
utility and reliability of a clustering approach. The present study sought to correct for
these limitations in the exploration of heterogeneity in sex offenders and its implications
for treatment approaches and outcomes.

Treatment Implications of Studying Psychopathology Types

Though speculative, it is valuable to consider possible clinical implications here.
It has been postulated that variations in sex offenders’ clinical presentations may
differentially affect responses to the same treatment protocol (Ahlmeyer et al., 2003).

Sex offenders who withdraw from treatment typically have more severe types of

42

psychopathology (e.g., personality disorders, psychosis) than sex offenders who complete
treatment (Looman, Dickie, & Abracen, 2005). Other research has found that the clusters
of sex offenders with the most severe psychopathology were more likely to attend
treatment regularly than the offenders with less severe psychopathology (Kalichman et
al., 1990). In either case, these ﬁndings contradict the assumption of Ward and Gannon
(2006) that new treatment approaches applied to all sex offenders may be important to
further reducing recidivism rates and instead suggest that different treatment approaches
may be necessary for different types of psychopathology clusters. In addition, if it is the
case that heterogeneous motivation and psychopathology both affect treatment responses,
perhaps there is an unidentiﬁed relationship between these two factors—or perhaps
looking at them together as opposed to separately would provide a more complete picture
of the manner in which internal psychological factors interact to produce certain
behaviors and outcomes in sex offenders.
Conclusion

This literature review has attempted to illustrate that motivation to change is an
important feature of sex offenders which could have direct inﬂuences on treatment
responses. However, because of the heterogeneity of sex offenders, different
psychopathologies could lead to different genuine or disingenuous motives for changing
behavior. Therefore, clustering this population into psychopathology subtypes and
descriptively evaluating these clusters may provide key information about the types of
sex offenders who may be motivated to change. Finally, comparing these clusters by
motivation to change and other clinical variables may provide information about

treatment response and new intervention strategies.

43

AIMS OF THE PRESENT STUDY

The present study examined how variation in sex offenders may be expressed
through psychopathology, motivation to change, and treatment responses. To address
this goal, the present study examined the archival data of non-incarcerated male sex
offenders beginning an outpatient treatment program. This study utilized a cluster
analytic approach to identify psychopathological types of offenders and then validated
those clusters with both offense characteristics and clinical variables, including
motivation to change. It was hypothesized that psychopathology clusters (which were
intended to capture the heterogeneity within this sample of sex offenders) would be
associated with different levels of motivation to change based on the previously reviewed
theoretical and empirical research.
Expected Psychopathology Clusters

Based on the cluster analyses previously conducted with sex offenders, it was
expected that four or ﬁve clusters will form in the present study since three of the four
reviewed MMPI clustering studies yielded either a four— or ﬁve-cluster solution. The
expected clusters include: (I) offenders who have psychopathology scores within normal
clinical ranges. Their motivation to change scores will likely be representative of
offenders who are beginning treatment (i.e., motivation will likely be low). (2) Those
that have elevations on measures of self-denial and minimization as well as typical
criminal offender psychopathology (e.g., antisocial characteristics, impulsivity). These
individuals may have lower motivation to change because they deny the seriousness of
their actions and, thus, better justify their offending behaviors (Kear-Cowell & Pollack,

1997). (3) Those who have a higher sensitivity to guilt and, therefore, may feel more

44

motivated after offending to change their behaviors (Hudson et al., 1999). Their
psychopathologies may include elevations in anxiety and depression as these pathologies
can be associated with insecurities, worry, and shame (APA, 2000). (4) Some sex
offenders have genuine and extensive psychological disturbances such as schizophrenic
symptoms (i.e., poor judgment, social isolation), disturbed thought patterns, paranoia, and
impulsivity. These offenders may have low motivation to change because they are not
attuned to the meaning of their behaviors. (5) If a ﬁfth cluster emerges, it is expected to
be of those who are highly psychologically disturbed who have elevations on multiple
traits associated with antisocial behavior (e. g., exaggerated response styles, antisocial
characteristics, overly masculine tendencies, impulsivity). These individuals may be the
least motivated to change as they may be driven by external gains (Lee at al., 2001 ).
Descriptive Variables

Offense characteristics. The ﬁrst offense characteristic which was used to
describe these clusters was victim status. As the literature review indicated, sex
offenders against adults and sex offenders against children may have different types of
psychopathology characteristics, though these distinctions may not be as prominent as is
sometimes stated in the literature. It is therefore hypothesized that the clusters of the
present study may differ by victim status. The other offense characteristics which were
used in this study were included based on the results of previous clustering studies and
the information available in the criminal ﬁles of this study’s participants. Those
additional characteristics included the type of offense perpetrated, whether the offense
involved physical penetration of the victim, the gender of the victim, the victim’s age, the

relationship of the victim to the offender, whether the offender was intoxicated at the

45

time of the offense, whether the offender had prior offenses, whether the offender is
currently on probation or parole, and the number of months the offender spent in jail or
prison for the sexual offense.

Clinical characteristics. Motivation to change was the primary theoretical
comparison variable. Treatment attendance and completion also served as comparison
variables. These are presumably related; Tierney and McCabe (2002) postulate that if
sex offenders are motivated to change then it would be reﬂected in actions such as
treatment attendance and completion. The speculation about the relationship between
motivation to change and treatment behavior is supported by limited empirical research
ﬁnding that lower levels of motivation to treatment were associated with higher attrition
rates (Beyko & Wong, 2005). A number of clinician-rated psychological variables
assessed before and after treatment were also included in the analyses. Those variables
were risk of committing another sexual offense, psychopathology, maturity, denial, and
impulsivity. It was expected that the clusters of individuals with higher levels of
motivation to change would attend more treatment sessions and complete the treatment
program requirements.

Hypotheses

The following hypotheses were developed:

1. Hypothesis 1 : Sex offenders will form into at least homogenous clusters based

on their psychopathology traits.

2. Hypothesis 2: The resulting psychopathology clusters of Hypothesis 1 will

differ on offense characteristics such as victim status, suggesting validity in

cluster structure.

46

. Hypothesis 3: Motivation to change will form into one latent construct which
will predict both treatment attendance and completion for sex offenders.
. Hypothesis 4: Clusters will vary by their level of motivation to change and

treatment variables (i.e., attendance, completion).

47

METHODS

Definitions

The archival data used in this study consisted of the clinical and criminal records
of 228 non-incarcerated adult (18 or older) male sex offenders on probation or parole for
having been convicted of a sexually violent act. As explained later, the ﬁnal 228 ﬁles
used in the analyses were the result after data cleaning and evaluation of data quality on
an initial set of 274 ﬁles. Based on information obtained through the Michigan Penal
Code (Michigan Legislature, 1931), for the purposes of this study, sexually violent acts
included coerced or forced vaginal/anal sex, oral sex, and/or inappropriate fondling of the
breasts, buttocks, or genitals. For this study only those convicted of these types of
contact offenses were included in the analyses and, in accordance, the term sexual
oﬂending referred to coerced or forced sexually violent acts with an adult woman or
child. ” In this study, sex oﬂenders against adults referred to men convicted ofa
sexually violent act against a victim aged 16 or older while sex offenders against children
referred to men convicted of a sexually violent act against an individual under the age of
16.12
Archival Data

This study used archival data which had not been analyzed previously. Data for
clients who entered a treatment program between 2000 and 2005 and who completed all

of the necessary intake assessment measures were used. To explore the possibility of

 

” Non-contact sexually violent acts will be conceptualized as those in which the offender does not make
direct physical contact with the victim and/or no one specific victim can be identiﬁed. Examples may
include possession of child pornography, exhibitionism, or voyeurism. These types of offenses will not be
considered in this study unless they are committed in conjunction with a contact sexual offense.

’2 The age of 16 was chosen to distinguish children from adults because the legal age at which an individual
can consent to sexual activity in Michigan is 16 years of age (Eaton County Prosecuting Attorney, 2006).

48

using this data for the purposes of this study, a data sampling procedure was conducted
and a random sample of client ﬁles were analyzed for content. This information was
discussed in detail in the proposal and has now been moved to Appendix I. All
information taken from the clients’ records and included in the database for analysis in
this study were de—identiﬁed and cannot be linked back to any particular offender.

The ﬁles for this study were taken from a mid-Michigan treatment rehabilitation
program for offenders. This state-sponsored center regularly evaluated and treated adult
individuals after they were convicted of a crime, released from prison, and placed on
probation or parole. When the clients arrived at the treatment facility, the clinic’s intake
workers collected information about the sexual crime the client most recently committed
and for which they were being treated. These crimes may or may not have been their ﬁrst
sexual offense. This program conducted court-mandated outpatient group therapy with
offenders of differing crimes who were on either probation or parole and were required to
participate in treatment as a condition of their sentence. Some of these offenders were
recently released ﬁom jail/prison while others were not sentenced to jail/prison (i.e., went
directly to probation). The clinic placed offenders into a group for treatment based on the
crime that they most recently committed and were sentenced to treatment for committing.
At this facility there were groups for sexual offenders, domestic assaulters, perpetrators
of property crimes, and others; only the sex offender group records are used here. The
facilitators for these groups varied in their credentials and training. Group facilitators
were Master’s level counselors, Ph.D. level clinical psychologists, and Master’s and

Ph.D. level social workers. All facilitators were unaffiliated with the present study.

49

Prior to treatment, the group facilitator interviewed clients and administered a pre-
treatment assessment. The assessment battery differed for each facilitator and each
group, but was the same for all the sex offenders. To the sex offenders, the facilitators
administered two questionnaires assessing substance use, the Brief Symptom Inventory
(Derogatis & Melisaratos, 1983), the University of Rhode Island Change Assessment
Scale (McConnaughy et a1, 1983), the Sex Offender Maturity Index (an unpublished
manuscript developed by the director of the treatment facility), and the MMPI-2 (Butcher
et al., 1989). These assessments usually lasted 2-4 hours because the group facilitator
read each question on the questionnaire to the client. Once the assessment was complete,
the client began to attend group sessions regularly, though he may have joined a group
that had already met for several sessions. He was required to continue his treatment until
he had attended the program’s required number of sessions and completed all related
program requirements. Poor attendance or cooperation in the groups could have resulted
in a report being sent to the client’s probation or parole ofﬁcer and/or removal from the
group. Thus, clients had obvious extrinsic motivation to attend the groups—but not to
achieve particular goals from the treatment.

The exact treatment approach taken in each type of group therapy (i.e., therapy in
the domestic violence group, assaultive offenders group) was dependent on the counselor
and the nature of the group. For the sex offenders group (which did not separate sex
offenders into different therapy groups based on the degree or nature of their sexual
offense), the facilitator used a non-validated Maturity Model. There were eight domains
of this model including Healthy Growth, Responsibly Sexuality, Productivity,

Socialization, Nurturing, Wisdom, Critical Thinking, and Purposeful Living. In addition

50

to the material relevant to these domains, the facilitator also covered speciﬁc treatment
material from the Michigan Department of Corrections, which was often cognitive-
behavioral in nature (e.g., cognitive distortions, affective control). When clients
completed the required 30-sessions (even if they were not completed consecutively), they
wrote a brief summary (typically a paragraph) of the lessons and tools they learned
during the program, which became a permanent part of their treatment record. They also
took a ﬁnal exam testing their knowledge of the information presented to them during the
course of the treatment. They then became a “graduate” of the program and were
required to return to the facility once per month for individual meetings with the group
facilitator for the remainder of their probation or parole.

In order for his records to be included in the present study, an individual must
have been a contact sex offender. That is, he must have been convicted of criminal
sexual conduct (CSC) in the ﬁrst, second, third, or fourth degree at least once and his
crime must have included physical contact with the victim. (For additional information
and details associated with each of these convictions, see The Michigan Penal Code, Act
328 of 1931). At least one of these sexual offenses must have occurred after the age of
18; those with only a juvenile record of sexual offense were excluded. It was expected
that some clients would have convictions other than CSC on their criminal records; this
was not an exclusion criterion. However, because each category of offender (e. g. sex
offender, domestic assaulter) who entered the treatment program answered a different set
of questionnaires, only the clinical ﬁles of those offenders who enrolled in sex offender
group therapy and completed all of the associated measures were used in this study.

Therefore, the clinical records used in this study were from persons who were most

51

recently convicted of a CSC offense and either had no criminal history (i.e., the present
crime was their ﬁrst conviction) or had a history of CSC and/or non-CSC (e. g. breaking
and entering, domestic assault) crimes. Clinical records of persons who previously
completed this particular treatment pro gram or started the program but did not ﬁnish
were included in the study as long as all of the necessary measures for the current study
were completed at the same time period. Further, because the majority of sex offenders
who participated in this treatment program were males, records of female offenders were
excluded.
Data Collection and Preparation

The data used in this study were collected between 2000 and 2005 during the
intake procedures described above. Because of the use of archival data and the lack of
research consent, de-identiﬁed data were utilized per appropriate IRB requirements. The
section that follows (Data Quality Assurance) describes the main process to getting the
data and their source.
Data Quality Assurance

Anonymity. Data ﬁles were stored in locked ﬁling cabinets in a locked office of a
locked suite. In this lab, the ﬁles were de-identiﬁed by two undergraduate research
assistants and assigned identiﬁcation numbers speciﬁc to the present study for use by the
PI.

Data entry training To assure that the data were coded accurately, two training
sessions were held. The ﬁrst served to familiarize the assistants with the information
they would need to enter into the database. During this session, the assistants were

shown blank copies of the forms which they would use to gather the necessary data

52

information. The questions on the forms were explained in connection with the SPSS
database where they would enter the participants’ responses.

During ﬁrst training session the assistants were also given the coding schemes for
the two variables in the study which required subjective coding: Offense Type and
Relationship to Victim. Though offense type (i.e., CSC 1, 2, 3, or 4) is an objective
coding, it was not consistently provided as such in the offenders’ records. Thus, at times
the assistants would have to provide a code based on the legal deﬁnitions of CSC crimes
and the description of the crime provided in the participants’ ﬁles. Similarly, Relationship
to Victim had a subjective coding scheme which required additional training. The
assistants were required to read descriptions of the criminal offense and make a judgment
of how (or if) the offender knew the victim. The rest of the variables were either taken
from a questionnaire in the participants’ ﬁles which was already numerically coded (i.e.,
education, pre- and post-treatment variables, URICA), dichotomized as yes/present or
no/absent, or given a speciﬁc numerical scheme to use to code the variables (i.e., race,
marital status). The assistants were given a detailed spreadsheet of the variables to be
found, how they were to be numerically coded, examples of which types of responses
would fall into each category, and how missing data should be recorded.

Rating Reliability. After the ﬁrst training, the assistants began data entry with
entering the data used for evaluating keystroke and rater entry reliability. Each assistant
entered 20 randomly selected ﬁles into a separate database in order for reliability to be
checked (see Table 1). To check their ﬁles, the assistants made copies of the ﬁles and
used white out or black marker to de-identify the data. The primary investigator then

checked the responses in the database to the responses in the ﬁles and noted the errors of

53

each assistant. After the 20 reliability ﬁles were entered and the reliability calculated
(i.e., kappas and ICCs) using SPSS 14.0, a second training was held to clarify any
problems in the coding schemes and correct the keystroke errors between the two
assistants.

The variables with the lowest reliabilities (i.e., below .75) were Marital Status and
Relationship to Victim. After analyzing the types of errors being made, it appeared that
one coder would rely on the offenders’ report of their marital status at intake while the
other coder would rely on the clinician’s psychological report for that information. For
example, an offender may self-report that he is single at intake but the psychologist may
report that the offender is divorced. These two sources of information resulted in two
different codes and lower reliability (see Table 1). At the re-training, the assistants were
instructed to use the psychologist’s report for the demographic and criminal information.

As for the Relationship to Victim variable, originally it included nine categories
to capture the possible relationship that the offender could have had with his victim (i.e.,
biological child, stepchild, signiﬁcant other, friend, coworker, acquaintance, relative,
stranger, and other). However, after running inter-rater reliability for this variable, it
seemed that nine categories were too many and may have been the cause of the error in
calculating the reliability the ﬁrst time (see Table 1). Thus a new six-code scheme was
developed in an attempt to increase the reliability. The changes in the scheme resulted in
an inter-rater reliability (ICC) of 0.93.

For the remaining errors in the other variables, inspection and discussion revealed
them to be keystroke errors that did not require further training or calibration. The

assistants were told which variables had discrepancies between them and were instructed

54

Ctl

r61

,L
cit

to correct their own errors. Once this was complete, the reliability was run a second time
on all of the errors. The ﬁnal reliability (ICC) ranged from 093-100 (see Table 1).

Data entry error checks. Of the 306 ﬁles available, 284 were usable (i.e., all of

 

the necessary measures were present). Each assistant entered all of the data for all 284
ﬁles, providing two complete datasets to double check errors. Once all of the data were
entered, an error-checking SPSS syntax was used which outputs the cases and variables
which had discrepancies between the two entries. The assistants used this output to
correct their errors. Sixty-three errors were found (< 1%). Once the errors were
corrected, the syntax was run a second time. A total of 14 errors were found in the
database. Again, the assistants corrected their own errors. A third and ﬁnal error check
revealed one error in the database. The assistants ﬁxed this ﬁnal error, making the
database ready for imputations and free of keystroke errors.
Final Sample Size, Missing Data, and Invalid Protocols

Overview. Not all of the ﬁles which had the necessary measures for inclusion in
the study had these measures completed in their entirety. Thus, it was important that
missing data be handled in some manner. The amount of missing data varied by variable
(range: 2.92-25.18%) and can be found in Table 2. Ten participants had more than 20%
of all of their data missing, which was beyond the cut-off decided at the time of this
study’s proposal approval. These 10 records were removed from the dataset. An
additional 46 cases were removed after it was determined that the MMPI proﬁle was
invalid due to an excessively elevated validity scale. This resulted in the ﬁnal N=228.

Missing data. The MMPI was the only continuous variable with no missing data.

For the remaining variables missing data were imputed using an EM estimation method

55

in SPSS. The EM estimation method was chosen because it is a maximum likelihood
estimation which makes fewer assumptions of the data than other imputation processes
(e. g., multiple regressions) and assumes that the values are missing at random. The
variables chosen for imputation were all continuous variables which had missing values
(i.e., months in prison, number of group sessions, pre-treatment subscale, post-treatment
subscale, pre-denial score, post—denial score, Pre-Contemplation subscale, Contemplation
subscale, Action subscale, and Maintenance subscale). It should be noted that the
Motivation to Change score as well as the Clinical Change and Denial Change scores
were calculated after their scale items were imputed. The variables not imputed were
categorical variables or demographic variables (i.e., victim’s age and offender’s age).

Correlations among the previously listed variables were compared pre- and post-
imputation. The average absolute value change in the correlations among the variables
after imputations was 0.006 (range: 0-0.06). Because the change in the correlations was
very small, it was concluded that the imputation worked as intended and preserved the
data structure. The imputed variables were used in the analyses.

Determining validity of MMPI-2 proﬁles. Similar to the way that an outlier skews
statistical results, having invalid MMPI proﬁles in the cluster analysis could result in
clusters forming around the elevated validity scores of a few offenders which would not
be representative of the data (Hair & Black, 2000). Thus, caution was used in
determining which ﬁles should be included in the analyses and which could be removed.
Criminal sample t-score means and recommended validity cut-off scores for scales F, L,
and K are listed in Table C1 of Appendix C. This information for scales F, L, and K was

used to determine the validity of the MMPI responses in the present study. Forty-six

56

MMPI proﬁles had elevations above the cut-off score on one of the validity scales and
those records were removed.13 The following chart summarizes the invalid cases and

their t-scores:

 

 

Scale T- Cases Scale T- Cases Scale T- Cases
score score score
F 101 4 L 83 9 K 87 1
104 8 87 7
107 3 91 4
1 10 2
1 13 1
1 19 1
120 6

 

 

 

Measures

The measures used in this study were part of the general assessment battery used
during the intake procedure at the treatment facility. The records used in this study were
taken from clients who had completed the program approximately two to three years
prior to data retrieval and analysis for the current study. No additional measures were
added for the purposes of this study. Instead, the primary investigator used the measures
available to her in order to operationalize the concepts in this study and answer the

research questions. All information used in this study was de-identiﬁed.

 

'3 To test the effect that these invalid cases may have had on the results of this study if left in the analyses,
a dichotomous covariate was created in which a score of zero was given to cases with all of their validity
scores within normal ranges and a score of one was given to cases with a validity score above the cut-off
range. This covariate was used in the MANOVA analyses which cross-validated the clusters against
continuous offense characteristics and treatment variables. After rurming these preliminary analyses,
results indicated that variables in the analyses were insigniﬁcant regardless of the presence of the covariate,
which suggested that the clusters did not differ on these variables. The covariate itself was signiﬁcant for
the variables Pro-Contemplation (p<.05), Contemplation (p<.01), and Motivation to Change (p<.05). These
results suggested that the invalidate proﬁles accounted for more variance in those variables than did
psychopathology. An analysis was also run after removing those cases with t-score elevations one standard
deviation above the cut-off score (14 cases total) but leaving the other cases in the dataset and including a
covariate representing the invalid cases (i.e., scores above the cut-off but below one standard deviation
above the cut-off). The same pattern of results emerged. Further, authors have indicated that using
covariates in clustering is counterproductive because it involves controlling for the heterogeneity which the
researchers is attempting to be unveiled (Luke, 2005). Therefore, the most cautious approach which could
be taken was to remove the invalid profiles from the analyses.

57

Clustering Variable: Psychgpathologv

Minnesota Multiphasic Personality Inventory, Second Edition (MMPI-2; Butcher
et al., 1989). This measure was designed to assess the psychopathology and personality
patterns of individuals. It is used frequently in forensic settings as an indicator of
psychological functioning (Megargee, 2006). The measure includes 554 True/False items
which create 8 subscales measuring the validity of the respondents’ answers (see Table
C1 of Appendix C), 10 basic psychopathology subscales (see Table C2 of Appendix C),
15 content scales, and several other supplementary content scales. Two validity scales
(i.e., F and K) and 7 clinical scales (i.e., D, Pd, Mf, Pa, Pt, Sc, Ma) were used in the
present study, each of which is described in more detail in Appendix J. The MMPI-2’s
reliability and validity statistics have been well-established in the literature in both
forensic and general populations (Butcher & Williams, 2000).

Me for use of MMPI-2 in present study. It was fortuitous that the MMPI
was available in the archival data and is one of the reasons why this pool of archival data
was selected for use in the present study. The MMPI-2 is an appropriate measure of
psychopathology in this study not only because of the success of previous studies in
clustering with this measure in sex offender populations (previously discussed) but also
because of the broad range of psychological problems which it assesses. The limitation
of using this measure with sex offenders is that it does not include any measures of
inappropriate sexual thoughts or behaviors which may be associated with sexual
offending (Butcher & Williams, 2000). Though the schizophrenia scale has questions
related to maladaptive sexual behavior, these behaviors are only one of the six concepts

which the schizophrenia scale is proposed to assess. Thus ﬁnding an elevated

58

schizophrenia scale in a sex offender may be a reﬂection of inappropriate sexual interests
but it could also signify drug use, extreme social isolation, or a thought disorder.

A different measure which could have been used was the Multiphasic Sex
Inventory (Nichols & Molinder, 1984). This measure assesses cognitive distortions about
sexual assault, atypical sexual interests, sexual dysfunctions, and social desirability as it
relates to sexual knowledge (Nichols & Molinder, 1984). While this measure has shown
acceptable reliability statistics in sex offenders (.71-.80; Nichols & Molinder, 1984) and
would presumably yield interesting ﬁndings, it was not used in the present study. This
measure is too focused on sexual psychopathology for the purposes of this study, which
had the goal of understanding general psychopathology in sex offenders. Since little is
known empirically about motivation to change in sex offenders, it was intended to keep
the constructs used in this study broad, and, thus, able to be built upon in future research,
consistent with suggestions from Tierney and McCabe (2002). The Multiphasic Sex
Inventory does not ﬁt with that goal of the study.

Another option instead of the MMPI clinical scales could have been specially
derived scales ﬁ'om the MMPI which would have been speciﬁc to non-sexual criminals
(e.g., Megargee, 1994) or sexual offenders (e.g., Langevin, Wright, & Handy, 1990a,
1990b). Unfortunately, the predictive and test-retest reliabilities as well as the successful
replication of these scales have been inconsistent (Schlank, 1995), which severely limited
their utility in the present study. Thus, the MMPI-2 was viewed as the best available
choice in terms of clinical utility and statistical validity for the overall purposes of this
study.

Comparison variables

59

1. Motivation to change

University of Rhode Island Change Assessment Scale (URICA; McConnaughy et
al., 1983). Inspired by the TTM, McConnaughy and colleagues (1983) developed the
University of Rhode Island Change Assessment Scale (URICA; also referred to as the
Stages of Change Questionnaire) to measure the extent to which an individual is
motivated or ready to change a maladaptive behavior. The URICA was another measure
which the treatment facility included in their intake procedures and which was
appropriate to include in this study given that it is one of the few objective measures of
motivation to change available.

In completing this questionnaire, clients answer 32 questions about the extent to
which they feel their problem affects them; they mark their answers by using a 4-point
scale ranging from “Not at all” to “Extremely.” During the administration of this
measure for this study, the target problem was their most recent CSC offense. The
clinician administers the questionnaire to the client ensuring that he stays focused on the
target problem (McConnaughy et al, 1983). Example questions include: “I do not feel
that my problem is interfering with my life” and “Changing my problem would be for the
better.” Participants receive mean scores on each of four subscales: Pre-Contemplation,
Contemplation, Action, and Maintenance (McConnaughy et a1, 1983). The URICA’s
items do not inquire about the Preparation stage, which demonstrated poor statistical
qualities (McConnaughy, et al., 1983). (Correlations among the item subscales can be
found in Tables 8a-8d). When used with a sample of offenders, the original URICA

showed no group differences based on crime type and test-retest reliability ranged from

60

.50 to .74 among the respective subscales (McMurran et al., 1998). In the present study,
the alpha for the subscales ranges from .78 to .88 (see Table 7).

After scoring the URICA, the researcher or clinician can then use this information
to create motivational proﬁles based on the scores from all four “stages”. The original
authors of the URICA advised against assigning participants to one speciﬁc stage, as they
emphasize that this method does not coincide with the theoretical underpinnings of the
measure which, as previously described, seem to be parallel processes rather than stages
(McConnaughy et al, 1983). That is, clients can have similar scores on multiple stages as
they are transitioning from one stage to the next, making assignment to a stage difﬁcult.
An overall Readiness to Change (labeled in the present study as “Motivation to Change”)
score may also be calculated (Blanchard et al., 2003; Project MATCH Research Group,
1997, 1998). This score is the sum of the Contemplation, Action, and Maintenance
subscales (which are positively correlated with one another) minus the score from the
Pre-Contemplation subscale (which is negatively correlated with the other three
subscales). (Subscale correlations can be found in Table 6a). The higher the
participant’s Motivation to Change score, theoretically, the more ready they are to change
their behavior. For the present study, both the individual subscales (i.e., Pre-
Contemplation, Contemplation, Action, Maintenance) and the overall Motivation to
Change score were used in the analyses. The individual subscales were used in the
structural equation modeling to validate Motivation to Change as a latent construct and
predictor of treatment response. After the validation of this construct, only the

Motivation to Change score was used.

61

The URICA and Sex Offenders. The URICA was recently adapted to speciﬁcally
apply to sexually offensive behavior (Tierney & McCabe, 2004). This exact measure
was not available for use in the present study; instead the original URICA was available
in the records and therefore utilized. Tierney and McCabe (2004) revised the scale at a
text level only in an attempt to maintain the underlying theoretical construction of the
measure. Instead of the items stating questions about the individuals’ problem, the items
stated questions regarding their sexual problem speciﬁcally. The revisions of this
measure did not seem substantial enough to warrant disregarding the applicability of the
revised URICA’s reliability and validity statistics to the present sample. This modiﬁed
version of the measure demonstrates an internal reliability ranging from .81 to .90 among
the subscales and a test-retest reliability ranging from .69 to .84 among the subscales
(Tierney & McCabe, 2004). More details on the validation of this speciﬁc measure can
be found in Appendix F. The reliability of these scales were carefully evaluated in the
current study and noted previously.

2. Offense Characteristics

Client Proﬁle/Criminal] History Forms. This questionnaire, which was completed
during the intake process at the treatment facility, provided general demographic
information (i.e., marital status, education level, employment status, and age) as well as
more speciﬁc information about the individual’s criminal history. These forms were
administered during the intake procedure by one of the clinic’s intake workers. The
intake workers read these questionnaires to the clients who then verbally provided their
responses. The criminal history information gathered included type of victim (i.e., adult

or child), the offender’s relationship to the victim (e.g., spouse, stranger, or niece),

62

whether there is a history of prior offenses (i.e., yes/no), whether the offense involved
penetration of the victim (i.e., yes/no), the gender of the victim (i.e., male or female), and
whether the offender was intoxicated (with drugs or alcohol) at the time of the offense
(i.e., yes/no). In the present study, each of these offense characteristics was coded as a
categorical variable and used in chi-square analyses to examine descriptive differences
between clusters. Material obtained from this form was supported by the information
presented in the offenders’ psychological reports, which were also in the offenders’ ﬁles
gathered from the treatment facility. These reports also provided additional information
not found on this questionnaire such as the degree of sexual assault, victim’s age, time
spent in prison or jail, all of which were included in the analyses.

3. Treatment behavior

a. Treatment Attendance. Each time the offender went to the treatment facility for
therapy he signed a sheet in his ﬁle. This sheet was the most accurate piece of
information recording the attendance at the group sessions and, thus, was used to obtain
the number of group therapy sessions each offender attended. The number of sessions
attended was used as a continuous variable and entered into a MANOVA to compare
psychopathology clusters.

b. Treatment Completion. Using a report from the ﬁles at the treatment facility,
information on whether a member completed treatment or not was gathered. This
variable was dichotomized in the analyses as “Yes—treatment was complete” or “No—
treatment was not completed.” Treatment was considered complete if the group
facilitator noted in the report that all of the treatment requirements were met for that

particular offender. If the group facilitator reported that the requirements were not met or

63

if the offender was removed (or voluntarily dropped out) of treatment, treatment was
coded as incomplete in this study. This variable was numerically coded to represent its
dichotomous structure and then used in a chi-square analysis to test the differences
between psychopathology clusters on treatment completion.

c. Pre- and Post-Treatment Variables. Also present in the treatment facilities ﬁles
were monthly reports which the group facilitator completed assessing each sex offenders’
Psychopathology, Denial, Risk of another sexual offense, Impulsivity, and Maturity.
These measures were included in the analyses as an additional test of how motivation to
change may be related to treatment responses. Each of these variables was scored on a
scale of 1 to 10, which 10 indicating greater presence of a particular pathology. (Before
analyses were run, the Maturity variable was recoded so that higher levels indicated less
maturity). Though these reports were supposed to be completed monthly, they were not
consistently available in the ﬁles. Thus, only the ﬁrst month’s scores and the last
month’s scores (i.e., those in the ﬁnal treatment report) were used in the analyses. If an
offender did not complete treatment (20% of total sample), their post-treatment scores
were coded as the same as their pre-treatment scores, indicating no change.

A scale reliability analysis was run on these items to see if pre- and post-treatment
scales could be created. Results indicated that Psychopathology, Maturity, Impulsivity,
and Risk formed a reliable scale but if Denial was included with that scale the reliability
decreased (see Tables 9a and 9b). This pattern was consistent in both the pre-treatrnent
and post-treatment items. Thus, one pre-treatrnent and one post-treatment variable were
created (i.e., summed) which included Psychopathology, Maturity (reverse-coded),

Impulsivity, and Risk. This total score was labeled Pre-Treatment Score or Post-

64

Treatment Score. In addition, the pre- and post-treatment Denial was included in the
analyses as a separate variable. Each of these variables was coded in a manner such that
higher scores indicated more maladaptive pathologies.

Using EM estimations in SPSS 16.0,14 18% of the pre-treatrnent scores were
imputed. One case’s post-treatment score was imputed. This case was the only one
which completed treatment but did not have a treatment response measure. For all other
cases in which treatment was completed an outcome measure was present in the ﬁle. The
same 18% whose pre-treatrnent scores were imputed also did not complete treatment.
Therefore, rather than imputing the post-treatment scores, their post-treatment scores
were coded as equivalent to their pre-treatment scores. For denial, 14% of the pre-denial
scores were imputed using the same methods as the pre-treatment scores. Two cases’
post-denial scores were imputed because these two cases completed treatment but did not
have a treatment response measure in their ﬁle. The remaining cases which did not
complete treatment had their post-denial scores coded as equivalent to their pre-denial
scores.

Additionally, a Clinical Change score and Denial Change score were created by
subtracting the pre-treatment scores from the post-treatment scores. This process resulted
in negative values because the pre-treatrnent scores were higher than the post-treatment
scores; that is, before treatment the sex offenders had higher scores on these maladaptive
pathologies than after treatment. Scores were not reversed to be positive; thus negative

scores are indicative of positive change conceptually (i.e., functioning improved).

Summary

 

’4 SPSS 16.0 was only used to impute the data. SPSS 14.0 was used to run the analyses. The reason for
this is that SPSS 14.0 did not have the imputation methods needed for this study but the primary
investigator did not have access to SPSS 16.0 for all of the analyses.

65

The following chart summarizes the measures which were used and also includes
the person(s) responsible for administering the instruments and approximately when in

the treatment process this information was gathered:

 

 

Measure Administrator When Administered
MMPI Licensed Psychologist Intake
URICA Intake Worker Intake
Client Proﬁle/Criminal History Intake Worker Intake
Treatment Attendance Administrative Assistant Each Treatment Session
Treatment Completion Group Facilitator Last Treatment Session
Pre-Treatment Variables Group Facilitator First Treatment Session
Post-Treatment Variables Group Facilitator Last Treatment Session

 

Note to Chart. The same licensed psychologist administered the MMPI to each of the sex offenders using
the tape recorded version of the MMPI. The intake workers varied depending on when the clients entered
treatment; thus, there is no consistency in who administered the forms. The administrative assistant signed
in each client at each session. The same group facilitator reported treatment completion and clinical ratings
for all of the sex offenders.

Data Analyses

Cluster Analysis. Both agglomerative hierarchical and iterative (k-means) cluster
analyses were used in this study in order to maximize within-group (i.e., cluster)
similarities and minimize between-group similarities. Agglomerative hierarchical
clustering procedures conceptualize each case as a cluster and then group those cases
with each other based on within-group and between-group means. Because the
agglomerative approach can be used with larger sample sizes (Rapkin & Luke, 1993),
this approach was used here. In the present study, Ward’s method was used because the
focus in assigning cases to a cluster is on maximizing between-group heterogeneity and
minimizing within-group heterogeneity rather than on matching a single case to the
overall cluster (Rapkin & Luke, 1993).

The iterative approach involves assigning cases to a particular cluster based on the
researcher’s predetermined speciﬁcation of the number of clusters and the “cluster seed”

(i.e., the mean of all cluster variables for each individual randomly assigned in the ﬁrst

66

clustering attempt) expected in the sample (Hair & Black, 2000; Luke, 2005; Rapkin &
Luke, 1993). The added advantage of this approach is that it allows cases to be
reassigned to clusters (Rapkin & Luke, 1993). The cluster seeds are recalculated when
new cases are added, after which cases may be reassigned to different clusters if their
means better match the new mean of another cluster. The result is a set of more
homogeneous (within-group) clusters (Rapkin & Luke, 1993). Please see Appendix K
for more background about clustering approaches.

Cross-validation. Once clusters were formed using the above mentioned
clustering techniques, they were compared based on a theoretical or empirical
understanding of why the characteristics associated with each cluster emerged (Hair &
Black, 2000). This was an essential feature of the cluster analysis and in applying
meaning to the clusters. If the clusters are not validated then there is no justiﬁcation for
the resulting clusters and less viability for their patterns. Either ANOVAs or chi-squares
are used to examine differences between these clusters. In these analyses, the clusters
served as the between group factor and the variable chosen to validate the clusters serve
as the outcome.

Structural equation modeling. To test the hypothesis that motivation to change is
a valid construct in sex offenders and a valid predictor of treatment responses, a structural

equation model was conducted using Amos 6.0 (Arbuckle, 2005).

67

RESULTS

The total number of records used was 228. Approximately 73% of the sample
was Caucasian, 69% had at least a high school diploma, 57% was single, and 56% was
employed. The average age was 34.6 years. Eighty-one percent of the sample had child
victims. Continuous variables (except offender’s age, victim’s age, number of children,
and the MMPI) were imputed using EM estimations (see Methods section for details).
Detailed descriptions of the means, ranges, missing data, skew, and kurtosis can be found
in Table 2; Table 3 provides a description of the categorical variables.
Hypothesis 1: Sex offenders will form into at least ﬁve homogenous clusters based on
their psychopathology traits

Cluster structure. To evaluate whether clustering was the correct statistical
approach for the data in this study, three steps were taken: Step 1) initial iterative
clustering; Step 2) agglomerative hierarchical clustering; Step 3) ﬁnal iterative clustering.
These steps were employed given the research which indicates that, separately, iterative
and hierarchical clustering approaches have shortcomings which can be minimized when
these two techniques are used in combination (see Appendix K). To compensate for the
shortcoming of the iterative approach, the agglomerative hierarchical approach was used
in conjunction with the iterative approach. The cluster means from the hierarchical
solution were then fed into an iterative cluster where the data was iterated multiple times
until maximum heterogeneity was achieved.

Step I : Initial iterative clustering. The purpose of the initial iterative cluster was
to gain a preview of the likelihood of forming a certain number of clusters in the dataset.

In this analysis, the researcher inputs the number of clusters to search for in the clustering

68

variables based on theory or ﬁndings from previous research. For this study, the initial
iterative cluster analysis was an exploratory step to determine if the hypothesized ﬁve-
cluster solution or another solution would produce meaningful clusters. Thus, the ﬁve-
cluster solution was analyzed, as were three- and four-cluster solutions. These cluster
solutions were chosen so that the cluster sizes would allow for sufﬁcient power in
running the cross-validation analyses. In selecting the ﬁnal cluster solution, the goal was
to maximize the number of clusters which provided clinically signiﬁcant scales but which
were not conceptually repetitive (i.e., interpreting one cluster did not result in the same
understanding of their emotional and behavioral pathology as another cluster). These
criteria were chosen in accordance with researchers’ suggestions (i.e., Rapkin & Luke,
1993)

Multicollinearitv of scale Sc. Per previous research, the MMPI scales included in
the analyses were F, K, D, Pd, Mf, Pa, Pt, Sc and Ma. The inter-correlations between the
MMPI scales showed a high correlation between scale Sc and scales F, Pa, and Pt (i.e.,
correlation above 0.70; see Table 6b).15 This multicollinearity could cause problems in
the cluster solution, such as redundancy in cluster groupings (Hair & Black, 2000).
Having highly correlated variables in clustering also results in those variables being
weighted in the analyses and assumed to be more signiﬁcant in the analyses than they
actually may be (Hair & Black, 2000). The result could be that the software program
treats these highly correlated variables as one variable that is weighted rather than
examining the individual contribution of each variable (e.g., distribution of the variable

throughout the data; Hair & Black, 2000). This statistical assumption could result in the

 

‘5 The correlations reported in Table 6b are prior to the removal of the 46 sex offenders with invalid
proﬁles (i.e., n=274). However, after those 46 offenders were removed, scale Sc continued to be highly
correlated with scales F (r=0.74) and Pt (r=0.75) but not Pa (r=0.61).

69

formation of clusters around the correlated variables which may not be a true
representation of the patterns in the dataset.

Because So was the common variable in each of the high correlations, including it
in the analyses was questioned. To test whether scale So should be included in the
analyses, the initial iterative cluster analysis (Step 1) was run twice using SPSS: once
with Sc included and once excluding Sc. Conducting the initial iterative cluster twice
would allow for an examination of the pattern of means as well as the conceptual
meaning of the clusters if scale Sc is included and if it is removed. In total, this initial
clustering step produced six clusters: a three-, four- and ﬁve-cluster solution including
scale Sc and a three-, four-, and ﬁve-cluster solution excluding Sc (see Tables 10a-12b).
The evaluation of these cluster solutions (discussed next) indicated that Sc needed to be
removed from the analyses.

The results of the initial iterative clustering indicated that the same general pattern
of mean elevations, and thus conceptual interpretation, was evident in the clusters
regardless of if scale Sc was present in the clusters. '6 Two changes were noticeable
when scale So was removed. First, the means in the clusters which included So were
somewhat higher than the cluster means which excluded Sc. Therefore when So was
removed from the cluster analyses, there was a general decrease in the cluster means and
some variables which had reached clinical signiﬁcance when Sc was included no longer
did so. This decrease affected scales Pd and Pa of cluster two in the four-cluster solution
(see Tables 11a and 11b for a comparison). These variables reached clinical signiﬁcance

when So was included but no longer did so when So was removed, with Pa decreasing

 

’6 To avoid confusion, the conceptual interpretation of each of the six initial iterative clusters can be found
in a footnote accompanying each respective table (Tables 10-12b). The conceptualization of the ﬁnal
cluster solution used in the analysis will be detailed in this text.

70

more than Pd (i.e., nearly ﬁve points). While the cluster pattern which included Sc and
signiﬁcant levels of Pd and Pa could be a true reﬂection of the patterns of the data, it
could also be the case that the correlation between Sc and Pa led to those variables being
weighted and more participants with moderately elevated Sc and Pa scores being added
to that cluster. Then, when Sc was removed, Pa was treated without bias and the sample
was distributed only according to the offenders’ scores. Because scale Pd had only just
crossed the threshold for clinical signiﬁcance when So was included (i.e., t-score mean
was 66.70), it is possible that its lack of signiﬁcance after So was removed was due to the
general decrease in means for all of the variables after Sc was removed rather than to a
speciﬁc problem with multicollinearity. These results suggest that having Sc in the
analyses may be negatively inﬂuencing the cluster patterns.

The second observation which may be evidence that Sc is negatively affecting the
cluster solutions regards cluster ﬁve of the ﬁve cluster solution (see Tables 12a and 12b).
In this solution, when scale Sc is included in the analyses, scales Pa and Pt (which are
two of the scales with which So is highly correlated) are below the clinical cut—off.
However, when Sc is removed from the analyses, these two variables meet or surpass the
clinical cut-off. The means for these two variables (as well as scale F with which Sc is
also highly correlated) increased approximately ﬁve points when Sc was removed. This
result again suggests that when So is included in the analyses preference is given to the
highly correlated variables when adding new cases into the clusters. In this particular
case, the cases with high So and Ma scores may be added to cluster ﬁve, ignoring the
elevation of scales Pa and Pt. Then, when Sc is removed from the analyses, the

elevations and contribution of scales Pa and Pt is revealed. Taken together with the

71

previously described problems in the cluster solutions, it was concluded that
multicollinearity with So may be interfering with the precise formation of the clusters.
To maintain the integrity of the data and internal validity of the study, a more cautious
approach was taken and So was removed from subsequent cluster analyses.

Five-cluster solution. Of the three-, four-, and ﬁve-cluster solutions which
excluded scale So, the ﬁve-cluster solution captured the heterogeneity of the data better
than the three- or four-cluster solutions (see footnotes of Tables 10b, 11b, and 12b for
conceptual interpretations of the three-, four-, and ﬁve-cluster solutions, respectively).

The three-cluster solution included one cluster which had no MMPI scale
elevations above 65; one cluster with intermediate elevations (i.e., scores ranging from 60
to 65); and one cluster with several high elevations (i.e., scores above 70). While these
clusters appear to differ in their mean elevations and, thus, their conceptual
interpretations, not much can be interpreted about a set of clusters with low, intermediate,
and high MMPI means other than that they differ at least in degree or severity of
psychopathology.

The fourth cluster solution included a cluster with means below clinical levels,
similar to the three-cluster solution. Also similar to the three-cluster solution was a
cluster with high elevations on the MMPI scales. The last two clusters in the four-cluster
solution both had intermediate elevations. The factor which distinguished these clusters
is the extent of the elevations on scales F, Pd, Pa, and Pt. As will be explained later, both
the scale which is elevated and the extent of the elevation have a signiﬁcant effect on the
conceptual interpretation of these clusters. Thus, though the two intermediate scales have

elevations on the same scales, the extent of the elevations will make a difference in

72

interpretation (see footnote of Table 11b for conceptual interpretation). Based on these
results, the four-cluster solution becomes more conceptually meaningful than the three-
cluster solution because of the type and extent of elevations.

However, the ﬁve-cluster solution was still examined to determine if an additional
meaningful cluster emerged. The ﬁve-cluster solution re-produced some of the previous
clusters and introduced new clusters as well (see Table 12b). The ﬁrst cluster (n=84)
contained MMPI scores below the clinical cut-off (i.e., psychologically healthy
offenders). The second cluster (n=60) had intermediate elevations (i.e., between 60 and
65) on scales F, Pd, Pa, and Pt (i.e., hostile and aggressive offenders). The third cluster
(n=40) had scores below the clinical cut-off except on scale F (i.e., exaggerating or
malingering offenders). The fourth had elevations on scales F, D, Pd Pa, and Pt above 65
and some above 75 indicating severe psychopathology (i.e., severely disturbed
offenders). The ﬁnal cluster had elevations above 65 on F, Pd, Pa, Pt, and Ma indicating
signiﬁcant psychopathology but not as signiﬁcant as the fourth clusters. These offenders
may have personality disorders (Megargee, 2006).

While the three- and four-cluster solutions did produce interpretable results, the
ﬁve-cluster solution had the most potential in not only capturing the heterogeneity of the
sample but also in providing meaningful conceptual interpretations. Based on these
initial analyses, the ﬁve-cluster solution continued to be hypothesized as the best ﬁt for
the data. This solution ﬁt the criteria of producing the most clusters without being
repetitive in conceptual meaning. This solution also produces clusters which are of

appropriate sizes for the cross—validation analysis.

73

Step 2: Hierarchical clustering. Because agglomerative hierarchical clustering is
an exploratory statistical approach that allows the natural patterns of the data emerge, it
was used next to evaluate the viability of the ﬁve-cluster solution in this sample. Unlike
the iterative clustering process, the hierarchical process does not allow the researcher to
specify the number of clusters to search for in the data. Rather, it automatically outputs
the number of cluster solutions that it ﬁnds in the data and which ﬁt the criteria of
maximizing between- group heterogeneity and uses the amount of change in the cluster
means as a indication of whether that particular case ﬁts closely with the other offenders
in that cluster. The hierarchical analysis began with the 228 cases in this study and
clustered them to produce a dendro gram which showed a 17-, nine-, four-, three-, two-,
and one-cluster solution (see Figure 1). Because a ﬁve-cluster solution did not emerge in
the hierarchical analysis, it was concluded that the ﬁve-cluster solution previously
described in Step 1 is not the best ﬁt for the data. The means for the three-, four-, and
nine-cluster solutions were examined to determine if one of these solutions had
conceptual meaning for the study (see Tables 13a—l3c). These solutions were chosen for
further analyses predominantly because they had the greatest potential for yielding
heterogeneity without producing clusters with sizes so small that statistical power would
be lost in the cross-validation analyses. However, before conceptually describing these
cluster solutions, an explanation of how these conceptual descriptions are being derived
is warranted.

MMPI-2 code typing. Each cluster in the present study had its own conceptual
meaning which could be explained via MMPI code typing. Code typing involves

interpreting the MMPI proﬁle based on a single elevated scale (i.e., above the clinical

74

cut-off score of 65) or the two highest scales (above 65) in a proﬁle (Butcher & Williams,
2000; Megargee, 2006).17 The code types which will be used to help explain the clusters
of this study were developed via MMPI validation studies and not via cluster analysis.
The MMPI code typing is only usal here as a guide to help underscore the most
prominent feature of the offender. Given that not all of the clusters of offenders in this
study had MMPI scales above 65, not all scales can interpreted as important to the
manifest behaviors of those offenders. Thus the present study will “layer” the
psychopathology during the conceptual description beginning with the emotional and
behavioral characteristics of the highest score(s), then the next highest, if more than two
scores are elevated above 65. The MMPI validity scales were also described
conceptually for each cluster but they are not traditionally interpreted as part of the code-
typing and, thus, will be another “layer” in the psychopathology description. In this
study, scores below 65 are not interpreted unless otherwise noted and justiﬁed via
previous research.

Clinical scale elevations are associated with different emotional and behavioral
characteristics depending on which other scales are also elevated (Butcher & Williams,
2000). In criminal samples, scale Pd is the single most elevated scale on the MMPI for
both males and females (Megargee, 2006). However, not all criminals act in the same
manner. It is the other scales which may be elevated along with scale Pd as well as the
degree of elevation (i.e., t-scores of 55-64, 65—75, or greater than 75) that deﬁnes the

characteristics which may manifest as a result of the individual’s proﬁle. In the

 

’7 Code typing also involves deciding if each proﬁle is “well-deﬁned,” which typically means that the
proﬁle scale has a t-score elevation at least 5 points above the mean. For the sake of simplicity, discussions
of whether a proﬁle was “well-deﬁned” were excluded from this study. However, the extent of elevations
above 65 was considered in that higher elevations meant different emotional and behavioral outcomes even
on the same scale.

75

description of these clusters, single elevation or two-point code types will serve as the
base for the cluster conceptualization and then the elevations of other clusters (if any are
present) will serve to further explain the possible emotional and behavioral characteristics
of the offenders.

Conceptualizations of clusters. In the three-cluster solution, the ﬁrst cluster was
comprised of 107 sex offenders whose t-scores were below clinical levels (see Table
13a). These offenders may be classiﬁed as having psychological proﬁles within normal
or average limits (Megargee, 2006).18 The second cluster was comprised of one clinical
elevation on scale Pd, which fell below 75. Offenders with this proﬁle may be
characterized as dishonest or arrogant (Megargee, 2006). The third cluster was
comprised of offenders with t-score elevations on scales F, Pd, Pa, and Pt. Scales Pd, Pa,
and Pt had scores between 65 and 75 while scale F was above 75. Offenders with these
elevations may be hostile, aggressive, and suspicious (Megargee, 2006). Similar to the
three-cluster solution of the iterative analyses, this three-cluster solution provided clusters
with some distinct conceptualizations but may not be a ﬁrll representation of the
heterogeneity of the data.

In the four-cluster solution, the ﬁrst and second clusters of the three-cluster
solution were replicated (see Table 13b). The third cluster divided into two separate
clusters. One of these clusters had t-score elevations above 70 on scales F, Pd, and Ma.
These offenders may be impulsive, aggressive, and manipulative (Megargee, 2006). The
last cluster in this group was characterized by t-score elevations on scales F, D, Pd, Pa,

and Pt. Offenders with these elevations tend to be emotionally and behaviorally unstable

 

‘8 Megargee (2006) or Butcher and Williams (2000) were cited throughout the conceptualization of these
clusters because of the code types they describe in their MMPI books, which were relied upon for the
inferences proposed throughout this section about the types.

76

(Megargee, 2006). This solution was similar to the iterative ﬁve-cluster solution
(earlier), except it does not have a cluster with only an elevated F scale. The spread of
heterogeneity in the four factor solution will provide important interpretations for the
purposes of this study.

The nine-cluster solution was unsatisfying. The ﬁrst observation that may be a
disadvantage of this solution is that two of the clusters have less than ten cases (N =7 and
N=9). It may be difﬁcult to detect the effects in the cross-validation analyses with
clusters which are so small, especially when conducting the chi-square analyses and these
clusters have to be divided into smaller cells. However, if this cluster solution were to be
used in this study, these small clusters of offenders may have particular characteristics
which distinguish them from the rest of the sex offenders and those unique characteristics
may be worth exploring.

Another potential disadvantage of this solution is that the ﬁrst four clusters all
have means below the clinical cut-off. They may be unlikely to produce signiﬁcant
results during the cross-validation (i.e., these four clusters may not be signiﬁcantly
different from each other though they may be signiﬁcant ﬁom other clusters) or to lack
clinical value. Having these redundant clusters may also result in less power to detect
effects in other clusters which are distinct from each other. Clusters ﬁve, seven, and nine
of this solution conceptually mimic clusters two, three, and four, respectively, of the four-
cluster solution; however, they have many fewer cases per cluster compared to the four-
cluster solution. The last two clusters of this solution (clusters six [N=7] and eight
[N=31]) are new clusters with new conceptualizations compared to the other cluster

solutions. However, given the size of cluster six (N =7), it may be difﬁcult to detect

77

effects for this cluster. Overall, the weaknesses with this solution were judged to
outweigh the potential beneﬁts of including the additional clusters.

Based on the results of the initial iterative analyses and the agglomerative
hierarchical analyses, it appeared that the four-cluster solution was the best for this
dataset. It maximized the number of clusters with clinical scale elevations while
preserving large cluster sizes and thus power for cross validation.

Cluster solution stabilﬂ Two more hierarchical analyses were run to test the
stability (or reliability) of the four-cluster solution. Using a ﬁrnction in SPSS, the dataset
was randomly split into two groups, each with a sample size of 114. An agglomerative
hierarchical analysis was conducted for each half to determine if a four-cluster solution
was evident in each half of the data. If a four-cluster solution was present in both halves,
then it can be concluded with greater conﬁdence that the four-cluster solution is a reliable
choice for representing the data (Rapkin & Luke, 1993). The hierarchical solution for the
ﬁrst half of the dataset yielded a 14-, eight-, ﬁve-, four-, three-, two-, and one-cluster
solutions (see Figure 2), thus supporting this conclusion. However, the hierarchical
analysis for the other half of the data produced a 15-, seven-, ﬁve-, three-, two-, and one-
cluster solutions (see Figure 3), with no four—cluster solution. While it is not uncommon
to yield inconsistent cluster solutions in smaller sample sizes (Rapkin & Luke, 1993;
Steinley, 2003), further examination of this ﬁve-cluster solution was warranted to
determine if it was similar to the four-cluster solution found in the full dataset.

In examining the means of the four-cluster solution of the ﬁrst half and the ﬁve-
cluster solution of the second half, there were conceptual parallels between them (see

Figures 2 and 3). Each solution had a cluster in which scale Pd was elevated or close to

78

elevated (i.e., antisocial offenders), another cluster featured the elevated F plus elevations
in either Pa or Pd in the low 608 (i.e., exaggerating or malingering offenders); and a third
cluster with elevations on scales F, D, Pd, Pa, and Pt (i.e., severely disturbed). Each
solution also contained a cluster with no scale elevations (i.e., within normal limits
offenders). The difference between the clusters were that the ﬁve-cluster solution
included a cluster with extreme elevations on scales F, D, Pd, Pa, and Pt.

Thus, some evidence supported a four- and some a ﬁve-cluster solution. In all,
because the ﬁve-cluster solution was not present in the hierarchical analysis which
included all 228 cases, it was concluded that the four-cluster was the most appropriate ﬁt.

Step 3: Final iterative clustering. Having conﬁrmed through the hierarchical
analyses that the four-cluster solution would be a meaningful solution for the data,
another iterative analysis was conducted to ﬁnalize the clusters (that is, the assignment of
individual cases). Similar to the process of an exploratory and conﬁrmatory factor
analysis, the purpose of this analysis was to reﬁt the clusters based on the results of the
hierarchical analyses rather than exploring the ﬁt of other solutions as with the ﬁrst
iterative analysis.

Two changes were made for this iterative analysis compared to the ﬁrst iterative
analysis. First, only the four-cluster solution was requested of the statistical program.
Second, in this ﬁnal step of the cluster analyses, the cluster means from the hierarchical
four-cluster solution (Table 13b) were used as the initial clustering seeds (i.e., starting
points) in the ﬁnal iterative clustering solution. Using the means for the hierarchical

analysis and specifying the number of clusters to examine, the ﬁnal iterative analysis has

79

more information to use to reﬁt the clusters and produce the ﬁnal cluster means (see
Table 14 and Figure 4).

The resulting solution produced four clusters which had distinct means and were
conceptually meaningful. The conceptualization of the four-cluster solution (see below)
did not change in any of the prior clustering steps, which indicates stability in the
solution.

This ﬁnal four-cluster solution does not strictly support Hypothesis 1 because less
than ﬁve clusters were formed. However, it did support the existence of meaningful
clusters and therefore was used to address the questions in the subsequent hypotheses.

To help better understand the changes that occurred among the cases during the
last two steps (the four cluster hierarchical and ﬁnal four cluster iterative analyses), Table
15 details the case migration and cluster assignment for those two analyses. Also, Figure
5 depicts the dissolution of the hierarchical clusters and reassignment of the cases in the
ﬁnal iterative analysis. (Because the initial iterative analysis was exploratory and there
was no support yet that the ﬁve-cluster solution was accurate, it was not included in
Table 15). The purpose of Table 15 and Figure 5 is to illustrate the ﬁnalization of the
clusters and the importance of using both types of cluster analysis to create the cluster
solution. The Table shows that approximately 24% of the cases in the four-cluster
solution hierarchical analysis (i.e., Step 2) migrate to new clusters in the ﬁnal four-cluster
iterative analysis (i.e., Step 3). These changes occur because of the inclusion of the
hierarchical cluster means in the ﬁnal iterative cluster solution. It is not surprising that
such changes took place in the ﬁnal analysis because the iterative process involves

reiterating cases until they most closely ﬁt the means of a particular cluster (including the

80

initial seed), whereas the hierarchical cluster does not reassign cases when the means of
the clusters change. Thus, though it may appear at ﬁrst that the case migration in Table
15 and Figure 5 demonstrates instability in the cluster assignment, it actually
demonstrates the importance of using the iterative solution so that the cases can be
reassigned to new clusters as the means change. It was concluded that the four-cluster
solution was reasonably well supported.

Conceptualization of ﬁnal cluster solution. Four distinct clusters formed in this
sample of sex offenders. They vary in severity of psychopathology but have scale Pd as a
core characteristic. The increase in severity may have particular meaning for offense and
clinical characteristics, which is a question of cross-validation and will be analyzed
below. The conceptual meaning presented here is meant to apply interpretation to
clusters based only on their psychopathology proﬁles.

These clusters were conceptually distinct based on an evaluation of the elevations

above 65 on the MMPI scales, which the following chart summarizes:

 

 

Cluster Scale Elevations
Within Normal Limits No elevations
(WNL; N=86)
Hostile (N=69) Pd
Manipulative (N=36) F, Pd

Very Disturbed (N=37) F, D, Pd, Pa, Pt

 

The ﬁrst cluster was labeled Within Normal Limits (WNL). This cluster included
the largest number of cases (N =86), consistent with previous research which has found
that the “within normal limits” proﬁle is the most common (though not itself a majority)
proﬁle evident in an efforts to type samples of offenders (Megargee, 2006). Criminals
with this pattern of scores tend to be psychologically adjusted, have less severe diagnoses

(e.g., adjustment disorders), and good treatment prognoses (Megargee, 2006). Also,

81

though scales D and Pd are not in the clinically signiﬁcant ranges, their scores place them
in the moderately elevated range (i.e., t-score of 55-64), which is interpretable. Because
of scale D (i.e., depression), this cluster of sex offenders may also have a pessimistic
cognitive style. The Pd score indicates that these offenders may be extroverted and
superﬁcially charming (Megargee, 2006). However, as a reiteration, these scales are not
in clinically signiﬁcant ranges. Comparing this cluster to those ﬁndings of previous
research, in two of their clustering studies Kalichman et al. (1989, 1992) had clusters of
offenders with MMPI scores within normal ranges.

The second cluster (N =69), which was labeled as Hostile, is characterized by a
single elevation between 65 and 75 on Scale Pd and a Scale Pa score just below clinical
signiﬁcance. Kalichman et al. (1989) and Duthie and McIvor (1990) both reported
clusters with single clinical elevations on scale Pd. Criminals with this proﬁle tend to be
argumentative, hostile, and suspicious of others’ intentions (Megargee, 2006). They also
can be arrogant and dishonest. They have problems listening to authority and are
difﬁcult to manage in prison and while on probation or parole. Their difﬁcult with others
can negatively affect their interpersonal relationships as well, making them less likely to
have stable marriages (Megargee, 2006). They may be resistant to engaging in treatment,
resentful of the therapist, and quick to scapegoat their problems (Megargee, 2006).

The third cluster (N=36), labeled as Manipulative, has a clinically signiﬁcant
scale Pd (t-score between 65-75) which indicates that they may share some of the
emotional and behavioral problems of the offenders in the Hostile cluster such as
disagreeableness. However this third cluster also has an elevated scale F (t-score above

75). These two scales elevated together are indicative of malingering and symptom

82

exaggeration (Megargee, 2006). This cluster was not found in any previous clustering
studies. These offenders may also engage in manipulation perhaps with the intention of
being classiﬁed as criminally insane or to receive some type of special psychological
service. Engaging in treatment may be difﬁcult for these individuals as they often have
difﬁculty accepting responsibility for their actions or appreciating the consequences of
their behaviors (Megargee, 2006). Because this cluster also has a scale Ma score which
is above 60, it could be the case that these offenders are highly energetic, have a high
sense of self-conﬁdence, and deny the seriousness of their actions (Megargee, 2006).
The ﬁnal cluster (N =3 7), labeled as (Very) Disturbed, is characterized by
elevations on Scales F, D, Pd, Pa, and Pt. These are the most severely disturbed
offenders of the clusters. While severely disturbed offenders were found in the clusters
of previous studies, the same scales were not elevated as here. In this cluster scales Pd
and Pa have the highest elevations on the clinical scales (above 75), thus their two-point
code type was used in the descriptions. However, because these score elevations are
above 75 and they have scales F, Pa, and Pt elevated as well, these offenders can be
interpreted in different ways from the Hostile cluster which also had elevations on scales
Pd and Pa. The Disturbed offenders may be suspicious of others, impulsive, and difﬁcult
to manage (Megargee, 2006). Because their scores are above 75, they may also engage in
violent acting out. They tend to lack empathy and genuine care for other people, making
them less likely to form a relationship and commit to it. These offenders also may
struggle with substance abuse and have a criminal record. Because of the clinically
signiﬁcant scales D and Pt, these offenders may display severe depression and anxiety

symptoms, appear agitated, have feelings of worthlessness, and possibly express suicidal

83

thoughts or behaviors (Megargee, 2006). Because of the number of scales which reach
clinical signiﬁcance in this cluster, the scale F score may be elevated due to acute
psychological distress rather than to malingering (Megargee, 2006). For treatment, these
offenders may want to be in therapy in order to gain some symptom relief but they also
may have difficulty engaging in the treatment process due to the severity of their
symptoms (Megargee, 2006).

Interestingly, scales K and Mf do not reach clinical signiﬁcance in any of these
clusters. In previous clustering research (e.g., Duthie & McIvor, 1990), these scales
appeared to be important factors in the psychopathology of sex offenders. Scale K was
proposed to be associated with treatment attendance (Kalichman et al., 1990) and scale
Mf has been associated with the selection of victims (Erikson et al., 1987). However, in
the present sample, neither of these variables had a prominent presence in this sample.
Because cluster analysis is sensitive to the presence of unimportant or non-theoretically
based variables in the analyses, removing scales K and Mf from the analyses was
considered. However, preliminary hierarchical analyses without these two scales
produced unstable clusters with unevenly distributed cluster sizes. Therefore, it could be
the case that even though these scores are not meeting clinical signiﬁcance in the clusters,
their presence in the solution is providing stability and helping to hold the clusters
together.

Hypothesis 2: The clusters will differ on offense characteristics such as victim status

To ensure that the clusters were signiﬁcantly different from one another in
severity, a MANOVA was run comparing the MMPI clustering variables (i.e., F, K, D,

Pd, Mf, Pa, Pt, Ma) between the clusters. Results indicated that the clusters are

84

signiﬁcantly different from each other on the MMPI scales, meaning that the cluster
analysis did create groups which maximized the between group heterogeneity (see Tables
16a-16b). This result does not indicate that these clusters are distinct types but it does
indicate that they are distinct clusters with signiﬁcant differences between them in terms
of psychopathology. Therefore, the cross-validation analyses were next completed as
proposed.

MANOVA and chi-square analyses (SPSS) were used to cross-validate the
clusters on offense characteristics. A MANOVA19 was run to test the between group
differences on the three continuous offense variables (i.e., age of offender, months in
prison, victim age). (Also, because the prison time variable was one of the imputed
variables, a covariate representing the number of variables each case had imputed for the
study was added as a covariate). The remaining offense variables, which were
categorical, were analyses via chi—square statistics. In both sets of analyses, cluster
membership served as the between-groups variable and each case received a code (i.e., 1,
2, 3, or 4) which coordinated with that case’s assigned cluster. Overall, results indicated
that the offenders’ age, drug use at the time of the offense, the gender of the victim, and
having a juvenile criminal history distinguished the clusters. Results for all of these
analyses are show in Table 17. 20

An analysis of demographic variables indicated that the clusters did not differ

signiﬁcantly on race, education, marital status, or current employment (see Table 17).

 

‘9 A MANOVA was chosen as the statistical approach to minimize the type I error associated with running
several individual f-tests.

20 Table 17 provides a summary of the cross-validation analyses including means and standard deviations
(or percentages if a chi-square analysis was run) for each cluster for each variable in the analyses. Also
provided in this table are chi-square and f-test statistics. The table is organized into demographic
characteristics, offense characteristics, and then clinical variables.

85

The offenders did differ in parenthood, with the Hostile offenders being more likely to
have children than any of the other clusters of offenders.

In terms of the offenders’ demographic characteristics, one crosstab with chi-
square statistics was run, outputting ﬁve separate chi-square tests for the ﬁve categorical
demographic variables (i.e., race, education, marital status, employment, parenthood).
Results indicated that the parenthood variable was statistically signiﬁcant with the
Hostile offenders being more likely to have children (74%) than any of the other clusters
(see Table 17). No other categorical demographic characteristics were statistically
signiﬁcant. It is possible that the Hostile offenders were more likely to have children
because they are older and had more years to engage in romantic relationships. An
ANOVA was run to test whether the clusters differed by the offender’s age. Tukey post-
hoc analyses were conducted to ﬁnd which clusters were statistically different from one
another. Results indicated that the Hostile cluster was signiﬁcantly older (age: 39.83)
than the WNL (age: 31.46) and Manipulative (age: 31.69) clusters but not the Disturbed
cluster (age: 34.88). Interestingly, the average age of the Hostile offenders falls near
what has been called the “burnout” point for offenders (of people with antisocial
personality disorder; APA, 2000). It is possible that the single t-score elevation on scale
Pd which characterizes this group represents the residual effects of an offender who had
reached his burnout point. Further analysis of this cluster’s offense characteristics could
shed more light on this possibility.

A crosstab with chi-square statistics was used to determine if certain categorical
offense characteristics differed by cluster membership: offense type (i.e., CSC 1, 2, 3, or

4), use of penetration, alcohol use around the time of the offense, drug use around the

86

time of the offense, victim status (i.e., adult or child), victim gender, relationship to
victim (i.e., did they know the victim), supervision status (i.e., parole or probation), any
prior offenses in adulthood (i.e., sexual or nonsexual crimes), and having a juvenile
record. Each variable is coded dichotomously except offense type, which has four
categories. Separate chi-square statistics were generated for each of the listed variables
and examined for signiﬁcance at p<.05. Results indicated that none of these variables
were signiﬁcant at p<.05.

Overall, these results suggest no differences between these clusters in terms of
their offense characteristics and only limited differences in terms of demographic
characteristics. Thus this hypothesis is not supported and the clusters are not well
validated on these indictors. The lack of signiﬁcance between these clusters could be
due to selecting a cross-validation measure which does not speciﬁcally address sex
offender pathology. A better measure may have been the MSI (described in the Methods
section) which assesses maladaptive sexual psychopathology (e.g., rape fantasies) rather
than criminal sexual activity as in this study.

Hypothesis 3: Motivation to change will form into one construct which will predict
both treatment attendance and completion for sex offenders

Structural equation modeling (Amos 6.0) was used to test motivation to change as
a latent variable in this sample. The model was run in two parts. The ﬁrst part tested the
structure of the four URICA subscales loading onto one latent factor (i.e., motivation to
change). The URICA subscales (i.e., Pre—Contemplation, Contemplation, Action, and
Maintenance) were placed in the model as the indicators loading onto the latent

motivation to change construct. In order for the model to run properly, each of the four

87

residual weights and one of the indicator paths (i.e., Contemplation) were set to 1.00.
Contemplation was chosen to be set to 1.00 because, when it was left to run free, it had
the largest loading and there was not enough residual variance left to ﬁt the other paths.
The results indicated that the model was a good ﬁt for the data (see Figure 6 which shows
standardized coefﬁcients). The three indicator paths were signiﬁcant at p<.001 and the
ﬁt indices supported the hypotheses that, in a sample of sex offenders, motivation to
change is single latent construct composed.

To test the validity of the motivation construct as a predictor of treatment
responses, the pre- and post-treatment variables (described in the Methods section) were
used. In the SEM model which validated the URICA (Figure 7) as a predictor of
treatment responses, the pre- and post-variables were selected to examine if motivation to
change can predict treatment response (i.e., post-treatment scores) while controlling for
pre-treatment factors. If motivation to change is a valid construct in sex offenders then it
should be able to predict treatment response.

As seen in Figure 7, the ﬁrst part of the model is the previously described latent
structure of the URICA subscales loading onto motivation to change. The second part of
the model includes three paths: one is leading from the latent motivation to change
construct to pre—treatrnent scores, another leading from motivation to change to post-
treatment scores, and the last path was leading from the pre-treatment variable to the
post-treatment variable. The path from motivation to change to the pre-treatment
variable was set to 1.00 in order to control for pre-treatment functioning. The path from
motivation to change to post-treatment scores was allowed to vary and was the path of

central interest in this model.

88

The results presented in Figure 7 (which shows standardized coefﬁcients) indicate
that motivation to change is a signiﬁcant predictor of post-treatment functioning while
controlling for pre-treatrnent functioning. This path was signiﬁcant at p<.001. It is also
negative, which means that as motivation increases post-treatment psychological
variables (i.e., impulsivity, risk of another sexual offense, psychopathology, maturity) as
rated by the therapist decrease (i.e., the offenders have better functioning). The ﬁt
indices for the model were appropriate and indicated an overall good ﬁt for the model
(see Figure 7). Based on these results, motivation to change can be conceptualized as a
latent construct in sex offenders and is a valid predictor of treatment response. The path
coefﬁcient for motivation predicting treatment response is below .30, indicating a low
effect size.

Hypothesis 4: Clinically, these clusters will vary by their level of motivation to change
and treatment variables (e. g., attendance, completion)

MANOVA and chi-square analyses were used to test this hypothesis. The
variables included in the MANOVA were attendance record, pre-treatment scores, post-
treatrnent scores, clinical change, pre-denial scores, post-denial scores, denial change,
and motivation to change.21 Tukey post-hoe analyses were conducted to ﬁnd which
clusters were statistically different from one another. Overall, results indicated that
attendance record and motivation to change score were each statistically signiﬁcant at
p<.05.

For attendance record, MANOVA results indicated that the WNL cluster (mean:

25.43) attended sessions more frequently than the Manipulative cluster (mean: 20.67) but

 

2' The URICA subscales (i.e., Pre-Contemplation, Contemplation, Action, and Maintenance) were excluded
from the analyses given that these factors all load onto one latent construct (i.e., Motivation to Change).

89

no other cluster (see Table 17). Given that the WNL cluster has less severe psychological
problems and a better treatment prognosis (Megargee, 2006), it is not surprising that they
attend treatment more regularly than some of the other more psychologically disturbed
offenders. This ﬁnding contradicts that of a previous study which indicated that the more
severely disturbed sex offenders had higher rates of attendance than the offenders with
average psychological proﬁles (Kalichman et al., 1990). The Manipulative cluster had
the lowest attendance rate. Offenders with this proﬁle tend to manipulate individuals for
their own gain (Megargee, 2006). In treatment they may try to convince the therapist that
they are either more sick than they truly are or are more motivated to change than they
truly are depending on their intentions (Megargee, 2006). Therefore they may only
attend treatment when they see the beneﬁt for their own gain.

In terms of their overall readiness to change, results indicate that the Disturbed
cluster (mean: 8.91) actually had a signiﬁcantly higher motivation to change score than
the WNL cluster (mean: 7.56). These results were unexpected given that the WNL
offenders had fewer psychological problems. It could be the case that the multitude of
problems that the Disturbed cluster had was motivating them to make more changes in
therapy. However, it is also possible that these offenders are manipulating the therapist
into thinking that they are more motivated to change than they actually are.

Overall these results support the hypothesis that these clusters would have clinical
relevance, particularly regarding motivation to change. These clusters differ statistically
on three of the clinical variables, with the greatest differences found between the WNL
and the Disturbed clusters. Contrary to expectations, the Disturbed cluster, not the WNL

cluster, had the overall highest elevations on most of the signiﬁcant clinical variables.

90

Post-h oc Analyses

To tie all of these analyses together, several regressions were run with treatment
responses (i.e., post-treatment score, post-denial scores, clinical change score, denial
change score) as the dependent variables and motivation to change and dummy codes
which served as proxies for psychopathology as the predictors. The purpose of these
analyses was to determine if motivation to change adds value to the prediction of
treatment response. No hypotheses were generated for these analyses.

Because the resulting four clusters could not be completely conceptualized as
ordinal, three dummy codes were used in the analyses to represent psychopathology.
Each dummy code was dichotomized to represent the presence or absence of a particular
cluster (i.e., Hostile, Manipulative, and Disturbed). No covariates were used in these
analyses. The regressions were run in one step with motivation to change and the three
dummy codes as the independent variables and each of the treatment response variables
as an outcome variable.

The ﬁrst regression used cluster membership and motivation to change to predict
post-treatment scores (see Table 18a). Results indicated that motivation to change was a
signiﬁcant predictor of treatment responses but the dummy codes were not. The beta for
motivation to change predicting treatment responses was negative, indicating that as
motivation to change increased, negative treatment response decreased (i.e., the
offender’s functioning improves). However the model only accounted for 5.3% of the
variance in treatment response.

The second regression used motivation to change and the dummy codes to predict

post-denial scores (see Table 18b). Results indicated again that motivation to change was

91

a signiﬁcant predictor of post-treatment denial scores while the dummy codes were not
signiﬁcant predictors. The beta was negative, indicating that as motivation to change
increased, level of denial decreased. The model accounted for 16.6% of the variance in
the denial score.

The third and fourth regressions tested whether motivation to change or
psychopathology types (as represented by the dummy codes) could predict the actual
degree of change at the end of treatment. Thus the clinical change variable and the denial
change variable were used as the outcome measures in two separate regressions. Results
indicated that the total regression model as insigniﬁcant and neither motivation to change
nor psychopathology could predict the degree of change (see Tables 18c and 18d).

Overall, these analyses indicate that motivation to change is a valid predictor of
treatment response (i.e., clinician ratings of clinical functioning at the end of treatment) in
sex offenders but it did not predict the degree of change (i.e., the difference between the
clinician ratings of clinical functioning at the beginning of treatment and at the end of
treatment) that can be expected during the course of treatment. Further, psychopathology
types did not add to the prediction of treatment response as it did not predict any of the
treatment response variables. In fact, when motivation to change is removed from the
analyses, none of the psychopathology types (i.e., dummy codes) could predict any of the
treatment response variables. Thus motivation to change alone may have greater clinical

utility than psychopathology.

92

CONCLUSIONS AND DISCUSSION

The present study used archival data to test whether the heterogeneity in sex
offenders’ psychopathology could be lawfully described using a clustering technique, and
whether along with motivation it helped explain variation in their treatment response.
This study had four main ﬁndings. First, four sex offender clusters which varied in type
and severity of psychopathology emerged. Second, these clusters differed signiﬁcantly
on measures of treatment behavior and responses but not on offense characteristics.
Third, motivation to change was a valid single latent construct in sex offenders with each
of the URICA subscales serving as indicators. Fourth, multiple regression analyses
indicated that motivation to change was a better predictor of treatment response than was
psychopathology. Each of these main ﬁndings will now be discussed in more detail
beginning with a comparison of the clusters in the current study to those found in
previous clustering studies.
Comparison of Current Clusters to Previous Clustering Attempts

The cluster analyses performed in this study improved upon the cluster studies of
the past through the use of a larger sample size and a mixed sample of offenders (i.e., sex
offenders against adults and sex offenders against children). Because of the mixed
sample of offenders, the clusters found here may represent hi gher-order sex offender
types which are not speciﬁc to either sex offenders against adults or sex offenders against
children. While there is some overlap between the clusters found in the current study and
those of previous research, this study also produces new clusters (both in regards to
scales means and conceptualizations) which are potential unique contributions to the

literature.

93

The WNL cluster is similar to one of the clusters in Kalichman et al. (1989),
which analyzed adult sex offenders, and two clusters in Kalichman et al. (1992), which
analyzed child sex offenders. This comparison suggests that, regardless of victim status,
there are sex offenders who do not have any manifest psychopathology and these
offenders comprise the majority of sex offenders in this study and each of the Kalichman
studies. The Hostile cluster in the present study had a single clinical elevation on scale
Pd similar to one of the clusters in each of the studies by Kalichman et al. (1989), which
analyzed sex offenders against adults, and Duthie and McIvor (1990), which analyzed sex
offenders against children. Similar to the WNL cluster, these results suggest that
antisocial behaviors and hostility may be a central factor in sex offenders regardless of
their victim status.

The WNL and Hostile clusters were the only two which were replicated from
previous studies. The Manipulative and Disturbed clusters were new clusters to emerge
in the current analyses. The Manipulative cluster, which displayed some of the more
severe offensive behaviors in comparison to the other clusters, may represent a
personality disordered sex offender but, if not, at least represents a manipulative and
cunning sex offender. This cluster may have emerged in the present study because of the
larger sample size and thus, the greater number offenders who would potentially have this
proﬁle. It is particularly interesting that this cluster emerged in the data even after those
offenders with elevated validity scales (i.e., those who deﬁnitely were manipulative and
malingering) were removed from the analyses. This result suggests that manipulation
may be a genuine feature of sex offenders’ psychopathology. Indeed each of the other

sex offender clustering studies found a cluster which had manipulation as a feature but it

94

was never paired with only antisocial characteristics. Instead, it was paired with other
scales which may be speciﬁc to the type of offenders being analyzed (e.g., severe anxiety
and thought disorders or anxiety; Kalichman et al., 1989; Shealy et al., 1990). Thus, the
Manipulative cluster found in this study could represent a core feature of a particular
subtype of the sex offender population that is further developed depending on the type of
sex offender being analyzed (i.e., sex offender against adult or child).

Finally, the Disturbed cluster was also a new cluster in this study, though it does
conceptually mirror the presence of psychologically disturbed offenders in previous
clustering studies. The interesting ﬁnding about this cluster of offenders is that, when
examining the previous clustering studies, the scales for the most severely disturbed sex
offenders are not consistent. That is, each cluster study produces a new set of MMPI
scale elevations as representative of the severely disturbed offender regardless of the
victim status associated with the cluster. Therefore, it cannot be concluded in this
particular study that this pattern of scale elevations is any more representative of a
higher-order psychopathology type in sex offenders than the means found in other
studies. What can be deduced is that in sex offender samples there will be a cluster of
offenders with severe psychopathology. This ﬁnding actually supports the previously
described treatment plan for the Disturbed cluster of offenders. Because of the lack of
consistency in the psychological proﬁles of the severely disturbed sex offender, an
individualized treatment plan which addresses the speciﬁc needs and concerns of each
offender may result in a successﬁrl rehabilitation.

The cluster analyses performed in this study served useful in identifying

psychopathology proﬁles which may be key features of sex offenders’ psychological

95

presentations regardless of their victims’ status as an adult or child. The use of a large,
mixed sample replicated two clusters from previous studies and yielded two new clusters
not found in previous research. As described in the Results section, these clusters may
not have many differences between them in terms of their offense characteristics but
interpretations could still be made about the variation in psychopathology and its
relationship to motivation to change and treatment. Future studies should use a mixed
sample of sex offenders to test the replicability of the clusters which emerged in this
study, particularly the Manipulative cluster which may provide a new understanding of
the psychopathology of sex offenders.

Based on the integration of these ﬁndings, the clusters found in this study have a
limited degree of external validity. The WNL and Hostile clusters may represent
consistent types of psychopathology in the sex offender population. However, they could
also be reﬂections of a general pattern in all types of criminals. Future research should
compare the ﬁndings of this study to the ﬁndings of clustering general populations of
criminal offenders to determine if the WNL and Hostile clusters are speciﬁc to sex
offender populations. The lack of replicability of the Manipulative and Disturbed
clusters between the present study and previous studies suggests that more research is
needed to determine the applicability of these clusters to the sex offender population in
general. Also, because the sample of the present study was mostly composed of sex
offenders against children, future studies should test the validity of these clusters in a

balanced sample of sex offender types.

96

Integration of Cluster Analyses and Treatment Implications

The next major ﬁnding of this study revealed that these clusters differed in
treatment response but not offense characteristics or demographics. Further, when
considering the signiﬁcant treatment response variables (i.e., treatment attendance and
motivation to change), only some clusters signiﬁcantly differed from each other.
Because of both of these ﬁndings, the clusters are only partially validated. However,
even this limited validity is important to consider further from the viewpoint of
generating hypotheses for future studies. The following discussion of each cluster
provides a speculative interpretation of the treatment response results and the potential
treatment implications associated with those results. Because of the limited signiﬁcant
ﬁndings, the interpretation of these ﬁndings should not be overstated.

Within Normal Limits. Based on the information collected in the present study,
approximately half of the WNL sex offenders were fathers and they were the youngest of
the clusters (along with the Manipulative cluster). Criminal offenders with psychological
proﬁles within normal ranges are also likely to have a good treatment prognosis and
better treatment responses than offenders with clinically elevated scales (Megargee,
2006). Consistent with these proﬁle characteristics, the WNL cluster had the highest
attendance rate of the clusters. Their motivation to change score indicates a moderate
level of motivation which seems appropriate given that they are just beginning treatment.
This motivation to change score also seems to coincide with the psychopathology of the
cluster. These individuals are mostly well adjusted but do have some problems in their
lives, such as their current sexual offense, which may be a manifestation of underlying

psychological problems.

97

The motivation scores of the WNL offenders indicate that they were not quite
committed to the idea of changing their behavior. This ambivalence may be targeted
through motivational interviewing prior to participation in a group therapy program. The
motivational interviewing could focus on expressing empathy to build rapport with the
client and attempting to understand the difﬁculty that he may be facing in changing a
pattern of behavior in which he has been engaging since adolescence. This motivational
interviewing could also attempt to engage him in explicit discussions about the
advantages and disadvantages of changing. Successful resolution of these offenders’
ambivalence could result in more successful treatment response.

Hostile. Based on the information collected during this study, this cluster is most
likely to be parents, perhaps because they are also the oldest of the four clusters. Because
of their MMPI proﬁle, these offenders may have problems with authority ﬁgures and thus
may be resistant to the therapist telling them what is best for them to do in their lives.
They also may have trouble accepting responsibility for their actions and may blame
others for their behaviors (Megargee, 2006). Contrary to these conjectures, the Hostile
offenders of this study had clinical ratings similar to the WNL cluster. The Hostile
cluster had the second highest attendance rate but the second lowest motivation to change
score. Because of this, it could be that these offenders were only attending treatment
because it was a mandatory part of their probation or parole requirements.

Motivational interviewing could also work to help this cluster of offenders

through their ambivalence, though it may be a more difﬁcult process for the therapist. It

98

has been proposed that using reﬂective listening22 with hostile client during the
motivational interviewing process can help him to engage in therapy but the process may
need to be repeated several times before the client becomes receptive to it (F arbring &
Johnson, 2008). For hostile offenders, it is speculated that their hostility is reﬂective of
an underlying value system that is important to them but by which they are not living
(Farbring & Johnson, 2008). Thus, reﬂective listening can help the therapist build
rapport with the client through communicating to the client empathically that the therapist
understands the ﬁ'ustration and anger that the client is feeling and describing the
discrepancy between their values and behaviors which may be one possible source of
their hostility. Consistency in this approach could result in the client being more willing
to engage in conversations about speciﬁc plans to change, thereby aligning their
behaviors with their value systems (Farbring & Johnson, 2008).

Manipulative. The Manipulative cluster may have behaviors involving the
manipulation of situations or other individuals in order to meet a particular agenda
(Megargee, 2006). Similar to the Hostile cluster, offenders in the Manipulative cluster
may be dishonest, narcissistic, and superﬁcially charming. They may also be impulsive,
have a high self-esteem, and deny the seriousness of their actions (Megargee, 2006).
Based on the patterns in their MMPI proﬁles, the Manipulative cluster may be a more
severe offender than the WNL and Hostile offenders. Approximately half of these
offenders are parents and they are the youngest of the clusters (along with the WNL
cluster). It could be hypothesized the sex offenders in this cluster currently meet criteria

for Antisocial Personality Disorder (ASPD). Their pathology indicates that they are

 

22 Reﬂective listening involves communicating back to the client the information that he has conveyed to
the therapist but in an empathic way and adding an interpretation to the client’s statements (Arkowitz &
Miller, 2008; Farbring & Johnson, 2008).

99

cunning, charming, defensive, and impulsive. Though giving a formal diagnosis based
on this information alone is not possible, the types of factors which are associated with
ASPD coincide with the characteristics of these offenders but not the other clusters
(APA, 2000).

Clinically, engaging in treatment may be difﬁcult for these individuals
(Megargee, 2006). Sex offenders in this cluster tend to have difﬁculty accepting
responsibility for their actions or appreciating the consequences of their behaviors. They
sometimes are resistant to engaging in treatment or wanting to change because of their
tendencies to scapegoat (Megargee, 2006). Consistent with these notions, the offenders
in the Manipulative cluster had the lowest attendance rate but the second highest
motivation to change score. Similar to the Hostile cluster, these behaviors are not
congruent with their level of motivation. Because of their psychological proﬁles, it could
be that these offenders were malingering and tried to impress the therapist with their level
of motivation but were not genuine.

The use of motivational interviewing would be limited with this cluster of
offenders. Because of the complexity of their psychopathology and use of manipulation,
during motivational interviewing these offenders may appear to be more motivated than
they actually are. They also may present as resistant rather than ambivalent, which is
harder to overcome (Arkowitz & Miller, 208). When working with these types of clients,
it has been suggested that tolerance of the resistance and consistent empathy may be the
best method for intervention (Farbring & Johnson, 2008). Also, frequent assessments of
the client’s motivation and willingness to develop a speciﬁc plan for change may assist

the therapist in understanding if the client is genuinely motivated or not (Farbring &

100

Johnson, 2008). However, if these offenders do not show improvement in motivation,
then therapy may be less effective and it may be better for them to have a longer prison
sentence. 1

Disturbed. The Disturbed cluster of offenders is the most severely pathological of
the clusters and may be suspicious of others, impulsive, and difficult to manage
(Megargee, 2006). They may also engage in Violent acting out, lack empathy, and have
troubles forming and maintaining emotional relationships. Of the four clusters, this is the
only one to have clinically signiﬁcant levels of depression and anxiety symptoms as well
as agitation, feelings of worthlessness, and possibly suicidal thoughts or behaviors
(Megargee, 2006).

Based on the information gathered in this study, the offenders in this cluster have
the second highest rate of parenthood and are younger than the Manipulative offenders
but older than the WNL and Hostile offenders. Because of the emotional component in
their psychopathology proﬁles, it could be that this cluster of sex offenders represents
those that are emotionally-driven. Hudson et al. (1999) described an emotional sex
offender as part of their offense pathways model. These offenders had alternating cycles
of excitement and guilt during the course of the offense process. Another possible
explanation for these offenders is that they have a more serious psychological disorder,
such as a thought disorder, which impairs their judgment and results in a number of
emotional and behavioral problems

Clinically, the offenders in the Disturbed cluster may have difﬁculty engaging in
treatment because of the severity of their symptoms (Megargee, 2006). In the present

study, these offenders had an attendance rate higher than the Manipulative cluster but

101

lower than the other two clusters. These slightly lower rates of treatment attendance
suggest that these offenders may have difﬁculty committing to the therapy process,
perhaps because of severe psychopathology or because of a disingenuous interest in
therapy. In terms of their motivation to change, these offenders had the highest
motivation to change of all of the clusters. This high motivation could be due to the
severity of psychological disturbances and their seeking symptom relief. If these
offenders represented the emotionally-driven offenders which Hudson et al. (1999)
described, then elevated motivation would be consistent with that presentation. However,
because of their elevated F scale, it could also be the case that these offenders were
manipulating the therapist in the same manner that the Manipulative cluster may have
been.

Ambivalence and motivation may not be the problem for these offenders in terms
of their clinical presentations and treatment responses; thus, motivational interviewing
may not be the best technique to use with them (F arbring & Johnson, 2008). These
offenders may need long-term, individual treatment which can target each of the
behavioral and emotional problems that they are having. If the hypothesis that these
offenders are committing sexual offenses because of their emotional disturbances, then
targeting those problems in additional to the maladaptive sexual behaviors may be one
possible way to reduce their recidivism rates. However, regardless of the content of
treatment and the structure of the treatments, individualized, long-term therapy may be
the key to helping these offenders stop offending. This therapy may need to take place in

prison if the offenders cannot stop committing sexual offenses while in the community.

102

Summary. Though only partially validated, these clusters represent some of the
possible ways in which psychopathology may vary in a sample of sex offenders and the
effect that this variation can have on clinical variables. Further, the speculation about the
types of treatments which may work with these offenders given their psychopathology
and motivation to change underscores the variation in treatment planning which may
need to take place if the goal of reducing recidivism rates even more is to be obtained.
Future studies should investigate these possibilities.

Because these clusters are only partially validated, their application in clinical
practice at this time would be premature. It may be that with further work, these clusters
may be used as a general screen for sex offender psychopathology types when treatment
begins. These psychopathology types may provide the clinician with an idea of the types
of problems that the offenders may be struggling with and those which may affect their
treatment prognosis. However, caution should be used when interpreting these cluster
proﬁles because of their limited validation. Because motivation to change was different
between two of these clusters, these psychological proﬁles may be used to help
understand why an offender may have a certain level of motivation to change but should
not be used as a substitute for directly measuring motivation to change. F urtherrnore,
because motivation to change was a more robust predictor of treatment response than
psychopathology (discussed below), these clusters cannot be used to predict treatment
response or outcomes until further research has been conducted.

Motivation to Change as a Latent Construct
The third major ﬁnding for this study was that motivation to change was a latent

construct in sex offenders. The present study was one of the ﬁrst to empirically study the

103

construct of motivation to change in a sample of sex offenders. It has been postulated
that there are four processes associated with motivation to change (i.e., Pre-
Contemplation, Contemplation, Action, and Maintenance) and that each of these
processes is associated with a different aspect in the change process (Prochaska et al.,
1992). It has further been postulated that treatment responses may differ because of the
level of motivation to change a client may have (Prochaska et al., 1992; Tierney &
McCabe, 2005). However, in sex offenders, motivation to change has not been
empirically tested as a latent construct nor has it been validated as a predictor of
treatment responses. The present study empirically tested these two constructs.

Based on the results in this study, motivation to change was a latent construct
with Pre-Contemplation, Contemplation, Action and Maintenance as its indicators.
These results provided some insight into the processes that may be involved when a sex
offender is trying to stop their offending behaviors. Sex offenders do not seem to move
through distinct stages, based on these results. Rather there are parallel processes
working within the sex offender and the extent to which an offender endorses the
characteristics associated with a particular change process determines the extent of his
motivation. This ﬁnding supports the theory that motivation is a continuous, oscillating
process that involves many changes and relapses as one tries to end patterns of behavior.
In order to know the utility of this construct, it was used to predict treatment responses.
Results indicated that motivation was a signiﬁcant predictor of treatment response when
controlling for psychological characteristics at intake. Because of this ﬁnding, one may
conclude that assessing motivation to change at the beginning of treatment may be able to

provide useful information about how the offender will progress in treatment.

104

Results ﬁom this study also indicated that variations in the severity of
psychOpathology are related to variations in motivation to change, as evidenced by the
signiﬁcant differences between clusters on motivation to change and its components.
Therefore, psychopathology may be one possible source of variation in sex offenders’
motivation to change (though it was less so for the offense characteristics). These
ﬁndings could mean that clinically addressing sex offenders’ psychopathology in
treatment in addition to their motivation to change could have signiﬁcant (and potentially
positive) effects on their motivation to change and, possibly, treatment responses. Future
studies should consider testing the validity of this interpretation through examining a
treatment response study in which motivation to change and psychopathology are both
treated during the course of treatment.

Motivation to Change as a Predictor of Treatment Responses

The ﬁnal main ﬁnding for this study regards the predictive capabilities of
motivation to change and psychopathology. Results indicated that motivation to change
was a more robust predictor of treatment response than was psychopathology. This result
suggests that, even though some of the psychopathology clusters of this study
signiﬁcantly differed by motivation to change thereby suggesting a relationship between
these two variables, motivation to change is the more important of the two when
predicting treatment response. The treatment implications of this ﬁnding include using
motivation interviewing or other enhancement techniques in order to increase the
likelihood of a better treatment response. However, this ﬁnding does not indicate that
psychopathology should be ignored when considering treatment responses. The cross-

validation results which indicated that motivation differed by psychopathology type

105

implies psychopathology may have an inﬂuence on level of motivation to change. It
could be the case that motivation is a moderator for psychopathology and, thus, utilizing
the interaction between the two constructs may provide even more clinically meaningﬁrl
information than examining them alone. The present study did not analyze this
interaction and future research should consider exploring this element. Further, caution
should still be taken in assessing the manner in which psychopathology may be affecting
motivation to change and in ensuring that the motivation that the cluster types are
endorsing is genuine. In doing so, the utility of using motivation to change to predict
treatment response would be enhanced.

One ﬁnding which was unexpected was that motivation to change was not a
predictor of the extent of change. That is, the degree to which an offender was ready to
change did not determine the degree of change. This ﬁnding may suggest that other
factors besides motivation to change are inﬂuencing the extent to which a sex offender
changes in therapy. One possible factor could be characteristics of the therapist (e. g.,
empathy) which have been hypothesized as important to enhancing motivation and
improving treatment prognosis in criminal populations (Farbring & Johnson, 2008). The
ﬁnding that motivation to change does not predict the degree of change in therapy also
highlights one of the limitations of the clinical application of motivation to change or the
URICA measure. Future research should work toward examining other clinical variables
which could be used in conjunction with motivation to change to predict the degree of

therapeutic change.

106

Treatment Implications

Speciﬁc treatment implications have already been discussed per cluster type in the
preceding section; thus, that information will not be repeated here. Rather, this section
takes a step back. Broadly speaking, these results suggest that motivation to change is an
important factor to consider during a sex offenders’ rehabilitation process. Motivation to
change has speciﬁc meaning for treatment responses and varies depending on the cluster
of offenders being considered. Because of this variation in motivation to change across
clusters, careful and repeated assessments of motivation to change may be necessary to
ensure that the therapist is gathering accurate information about the client’s current
motivational proﬁle. It may also be worth developing a clinician-rated motivation to
change measure which can be used in addition to the self-report measure. Especially
when considering the more manipulative sex offenders who may not be genuinely
interested in change, taking multiple measures from multiple reporters may increase the
viability of the results.

As discussed earlier, motivational interviewing may be worth considering as a
pre-treatment for the traditional cognitive-behavioral treatment approach with sex
offenders or as a technique to be incorporated into current cognitive-behavioral
frameworks. It has been suggested that of all criminals, sex offenders are the ones who
need and may best respond to motivational interviewing (Mann, 1996 as cited in
Ginsburg et al., 2002). The results of this study suggest that motivation to change should
be assessed and monitored during the treatment, thus motivational interviewing may be
one way to accomplish that goal. However, if the training necessary to conduct

motivational interviewing is not available, then objectively assessing motivation to

107

change through measures such as the URICA (or the modiﬁed URICA speciﬁc for sex
offenders) may serve as an alternative solution.

Concerning psychopathology, based on these results, psychopathology types only
add limited value to the understanding of treatment response and outcome in sex
offenders. These types differ in treatment behaviors such as attendance and completion
but these results are confounded by extrinsic motivating factors such as the consequences
of not completing treatment (e. g., returning to prison). Also, psychopathology types were
not predictive of treatment responses or the degree of clinical change. These results
suggest that psychopathology types have limited direct clinical applicability regarding
treatment responses. However, because these types differed signiﬁcantly in their
motivation to change, using psychopathology types in conjunction with motivation to
change may be useful in treatment planning. That is, the psychopathology types may
lend insight to the motivation of the offender which may then help in understanding that
offenders’ treatment prognosis. Thus, psychOpathology types do not have universal or
direct utility in understanding treatment behavior and response but the combination of
psychopathology types and motivation to change may have more utility.

Limitations

Though the present study is a good start to the understanding of motivation to
change in sex offenders and provides useful information as a starting point for other
empirical studies in this area, there are some limitations. One of the most signiﬁcant
limitations in this study is the use of the substance abuse literature to conceptualize
motivation to change. Because motivation to change has not been fully conceptualized in

sex offenders, it was necessary to draw from other literature to gather an understanding of

108

this construct and how it may operate in this population. However, the actual parallels
between substance abuse and sex offending are severely limited. Substance abuse often
involves physiological dependence on a substance that mostly causes suffering to the
substance abusers. However sex offending is not a physiologically-driven behavior and
innocent victims, including children, are the ones who signiﬁcantly suffer from the
choices that sex offenders make. Because of the differences between these behaviors, it
is also likely that the determinants of motivation to change may be different in these two
classes of psychopathology. For example, substance abusers may be highly motivated to
change but the physiological dependence on the substance makes that change difﬁcult.
On the other hand, some sex offenders’ lack of motivation to change may be due to
intrinsic motivations or belief systems related to the behavior (e.g., power seeking,
rationalization that the behavior is harmless, etc). Thus, how motivation is
conceptualized in each population has very distinct implications for enhancing that
motivation and the behavioral or treatment responses which may be associated with that
motivation. A key next step is to develop better measures of motivational indicators in
sex offenders preparatory to treatment.

The use of archival data created a number of limitations in the study. First, it
restricted which measures could be used in the analyses. If this study were to be
replicated, it would have the strength of being able to choose measures that were
speciﬁcally designed for sex offenders and, perhaps, will yield more signiﬁcant results
during the cross-validation process. Similarly, though the presence of the URICA in the
intake battery was fortuitous, the URICA may not be the best measure of motivation to

change, given its reliance on stages and processes of change which are difﬁcult to

109

interpret. Had the data been directly collected from the offenders, then perhaps a better
method could have been used, such as the inclusion of qualitative data to supplement the
objective measure. Though sex offenders have a wide variety of motives and reasons to
be honest or dishonest, a study assessing such an idiographic variable as motivation to
change may be enriched with the inclusion of qualitative data which can supplement the
quantitative measures. Lastly, because this data was taken from a treatment facility
which was not involved in research, there was no rigorous control of the data collection
procedures. Thus, the level of subjectivity and bias introduced in the data during the
collection procedures may have affected the amount of error present in the results.

In relation to the measurement problems, another limitation of this study is the
lack of a true measure of treatment outcome in terms of real world behavior over time
(i.e., recidivism). One of the implications of this study is that using the clusters may
provide information about treatment prognosis but without a measure of recidivism, the
full utility of these clusters is unknown. The use of treatment response in the present
study provided a starting point for understanding the clinical utility of clustering
psychopathology but further research is necessary to continue moving the ﬁeld forward.
Future research should consider using a longitudinal method which would involve using
these clusters of psychopathology to predict recidivism rates.

Another limitation of this study was the sensitivity of the cluster analyses to the
validity of the MMPI. Having to remove several of the offenders because of potential
invalidity decreased the sample size and power to detect effects, though it did not seem to

alter the results. Because those offenders represented even greater heterogeneity in this

110

sample, keeping them in the analyses may have yielded even more interesting results
about the psychopathology proﬁles in sex offenders.

Statistically, the effect sizes for some of the results were small based on Cohen’s
deﬁnition (i.e., beta less than .30). However, small effect sizes are not necessarily
clinically unimportant (Kendall, Flannery-Schroeder, & Ford, 1999). Though the effect
sizes of this study were small in terms of group means, they could have clinical
signiﬁcance in terms of individual outcomes, as can often happen in health-related
research (Kendall et al., 1999). For example, the SEM model testing motivation to
change as a latent construct had an r of .21 for motivation predicting treatment response.
According to the binomial effect size for display (BESD; Rosenthal & Rubin, 1982),
which translates the correlation into a proportion of cases leading to change, the r of .21
would be associated with an increase in treatment success from approximately 25% to
75% (Rosenthal & Rubin, 1982). Thus, the potential prediction of treatment success may
be viewed as clinically signiﬁcant.

A ﬁnal limitation regards the number of variables included in the study. Cross-
validating such a large number of offense characteristics and treatment variables
produced the possibility for an increase in Type I errors. However, one beneﬁt of having
included a variety of variables is that this study has potentially eliminated variables
which are less likely to differ between clusters and narrowed the scope on which
variables are important to attend to in research and treatment.

Future Research Directions
Future research can focus on understanding the role that motivation to change

may have in sex offenders. One ﬁnding which emerged ﬁom this study is that

111

psychopathology (as measured by the MMPI) and motivation to change are not highly
correlated (see post-hoc analyses in Appendix L). Therefore, future research may want to
focus on a wide variety of factors which may be related to motivation to change (e. g.,
denial, cognitive distortions) with the goal of understanding the offender characteristics
which may be driving motivation. If these factors can be identiﬁed then perhaps
treatment can be targeted toward these speciﬁc factors and become even more specialized
for sex offenders.

Similarly, future research can also consider continuing to explore the connection
these ﬁndings may have in other bodies of literature, such as the personality literature.
For example, characteristics of the Hostile cluster parallel ﬁndings in the temperament
and personality literature which suggest that afﬁliation/hostility are basic dimensions of
human personality. Perhaps these clusters represent higher order personality traits that
were not captured in the MMPI but could also have signiﬁcant clinical implications.

A ﬁnal future research direction regards the measure for motivation to change.
The URICA was used in the present study and was a valid measure. However, if a larger
sample size were used, a second-order factor analyses including the scale items may be
worth analyzing to ensure that the URICA subscales and motivation to change are latent
constructs. Also, the modiﬁed version of the URICA which as specially designed for sex
offenders needs empirical validation. Having this measure available in professional
settings would ensure that sex offenders are considering their sexual offending when they

answer questions about changing their behaviors.

112

Conclusions

Sex offending is a problem in society that many researchers have been trying to
address and reduce recidivism. However, one of the problems with studying sexual
offenders and treatment responses is that the varied outcomes which are apparent in the
results and inability to take those varied outcomes and develop one solution to the
problem. The present study therefore sought to study the variation in sex offenders,
focusing on psychopathology, to examine if this variation has implications for treatment
attitudes and behaviors. The results indicated that psychopathology may be one source of
variation in treatment attitudes such as motivation to change as well as treatment
responses. The current study has attempted to empirically test a new perspective in
understanding sex offenders and their treatment responses. Taking the suggestions for
ﬁrture research, if other researchers continue in this train of understanding motivation to

change, then perhaps the sex offender treatment literature can continue to develop.

113

APPENDIX A

Tables for Statistical Results

Table 1. ReliabilityStatistics for All Variables in the Dataset.

 

 

Variable No. of N Reliability Reliability Percent Percent
Raters Statistic Statistic Correct Correct
(ICC)l (ICC)2 (Rater A)3 (Rater B)3
Racea 2 19 0.89 1.00 100 100
Educationa 2 20 0.73 0.93 100 100
Marital Statusa 2 20 -- 1.00 100 100
No. of Children 2 19 0.99 0.99 100 80
Employeda 2 l9 1 .00 l .00 100 80
Offense Typea 2 19 0.92 1.00 80 80
Alcohol at Offensea 2 20 0.88 1.00 100 80
Victim Statusa 2 20 1.00 1.00 100 100
Victim Gendera 2 20 (all meBIC) (all female) 100 100
Victim Age 2 17 0.97 1.00 80 80
Relationship to 2 20 -- 0.93 100 80
Victima
Supervision Statusa 2 I 5 1.00 1.00 100 80
Months in Prison 2 16 1.00 1.00 80 100
Any Prior Oﬂensesa 2 14 1.00 1.00 100 100
Juvenile RCCOI’da 2 1 2 1.00 1.00 100 100
Attendance Record 2 20 0.99 1.00 100 100
Requirements 2 20 1 .00 l .00 100 100
Completea
Pre-Psychopathology 2 15 l .00 l .00 100 80
Pre-Maturity 2 l 8 0.99 l .00 100 100
Pre-Risk 2 16 0.98 1.00 100 100
Pre-Denial 2 l 8 0.99 l .00 100 100
Pre-Impulsivity 2 17 l .00 1.00 100 100
Post- 2 l7 1 .00 l .00 100 100
Psychopathology
Post-Maturity 2 17 l .00 1 .00 100 100
Post-Risk 2 1 7 l .00 l .00 100 100
Post-Denial 2 17 l .00 1.00 100 l 100
Post-Impulsivity 2 17 l .00 1 .00 100 100
URICA l 2 18 1.00 1.00 100 100
URICA 2 2 18 1.00 1.00 100 100
URICA 3 2 18 0.99 1.00 100 100
URICA 4 2 18 1.00 1.00 100 100
URICA 5 2 17 1.00 1.00 100 100

 

a = Kappa statistic

I = Before retraining of raters

2 = After retraining of raters

3 = Percent correct of raters after the retraining.

ll4

Table 1 (cont’d).

 

 

Variable No. of N Reliability Reliability Percent Percent
Raters Statistic Statistic Correct Correct
(ICC)l (ICC)2 (Rater A)3 (Rater 13)3
URICA 6 2 18 1.00 1.00 100 100
URICA 7 2 18 1.00 1.00 100 100
URICA 8 2 18 0.99 1.00 100 100
URICA 9 2 18 1.00 1.00 100 100
URICA 10 2 18 1.00 1.00 100 100
URICA ll 2 17 1.00 1.00 100 100
URICA 12 2 18 1.00 1.00 100 100
URICA l3 2 17 1.00 1.00 100 100
URICA 14 2 18 1.00 1.00 100 100
URICA 15 2 18 0.99 1.00 100 100
URICA 16 2 18 1.00 1.00 100 100
URICA 17 2 18 1.00 1.00 100 100
URICA l8 2 18 1.00 1.00 100 100
URICA l9 2 18 1.00 1.00 100 100
URICA 20 2 16 1.00 1.00 100 100
URICA 21 2 18 1.00 1.00 100 100
URICA 22 2 18 1.00 1.00 100 100
URICA 23 2 17 1.00 1.00 100 100
URICA 24 2 18 1.00 1.00 100 100
URICA 25 2 18 1.00 1.00 100 100
URICA 26 2 18 0.99 1.00 100 100
URICA 27 2 18 0.99 1.00 100 100
URICA 28 2 18 1.00 1.00 100 100
URICA 29 2 18 1.00 1.00 100 100
URICA 30 2 18 1.00 1.00 100 100
URICA 31 2 17 1.00 1.00 100 100
URICA 32 2 18 1.00 1.00 100 100
MMPI F 2 20 1.00 1.00 1.00 1.00
MMPI L 2 20 1.00 1.00 1.00 1.00
MMPI K 2 20 1.00 1.00 1.00 1.00
MMPI Hs 2 20 1.00 1.00 1.00 1.00
MMPI D 2 20 1.00 1.00 1.00 1.00
MMPI Hy 2 20 1.00 1.00 1.00 1.00
MMPI Pd 2 20 1.00 1.00 1.00 1.00
MMPI Mf 2 20 1.00 1.00 1.00 1.00
MMPI Pa 2 20 1.00 1.00 1.00 1.00
MMPI Pt 2 20 1.00 1.00 1.00 1.00
MMPI Sc 2 20 1.00 1.00 1.00 1.00
MMPI Ma 2 20 1.00 1.00 1.00 1.00
MMPI Si 2 20 1.00 1,00 1.00 1.00

 

T . .
= Before retrammg of raters
2 . .
= After retraining of raters
3 = Percent correct of raters after the retraining.

115

Table 2. Descriptive Statistics: All Variables (Before Immtation).

 

 

Variable Sample N Percent Mm. Max. Mean Skew Kurtosis
(max=274) Missing (Std Dev) (Std Err) (Std ErrL

Age 261 4.74 17.81 76.76 34.31 0.81 0.60
(12.07) (0.15) (0.30)

Race 266 2.92 1.00 6.00 1.42 3.01 10.62
(0.92) (0.15) (0.29)

Education 265 3 .28 1.00 9.00 4.23 0.65 0.86
(1.31) (0.15) (0.29)

Marital Status 257 6.20 1.00 4.00 1.72 0.60 -1.31
(0.87) (0.16) (0.30)

No. of Children 271 1.09 0.00 8.00 1.46 1.19 1.36
(1.62) (0.15) (0.29)

Employed 274 0.00 0.00 1.00 0.56 -0.25 -1.95
(0.49) (0.15) (0.29)

Offense Type 272 0.73 1.00 4.00 3.06 -0.53 —l .02
(0.98) (0.15) (0.29)

Penetration 273 0.36 0.00 1.00 0.60 -0.43 -1.83
(0.49) (0.15) (0.29)

Alcohol at 272 0.73 0.00 1.00 0.36 0.59 —1.67
Offense (0.48) (0.15) (0.29)
Drugs at Offense 270 1.46 0.00 1.00 0.21 1.39 -0.05
(0.41) (0.15) (0.29)

Victim Status 259 5.47 1.00 2.00 1.19 1.57 0.55
(0.39) (0.15) (0.30)

Victim Gender 269 1.82 1.00 2.00 1.94 -3.61 11.12
(0.24) (0.15) (0.29)

Victim Age 236 13.87 3.00 65.00 13.81 3.01 16.82
(6.79) (0.16) (0.32)

Relation to Victim 270 1.46 1.00 6.00 3.66 -0.06 -1.38
(1.74) (0.15) (0.29)

Supervision 265 3.28 1.00 2.00 1.37 0.53 -1.72
Status (0.48) (0.15) (0.30)
Months in Prison 259 5.47 0.00 252.00 29.57 2.10 4.98
(42.65) (0.15) (0.30)

Prior Offenses 250 8.76 0.00 1.00 0.64 -0.58 -1.67
(0.48) (0.16) (0.31)

Juvenile Record 245 10.58 0.00 1.00 0.23 1.27 -0.38
(0.42) (0.16) (0.31)

Attendance 271 1.09 0.00 47.00 23.93 -1.43 1.49
Record (8.53) (0.15) (0.30)
Requirements 273 0.36 0.00 1.00 0.76 -1.24 -0.47
Complete (0.43) (0.15) (0.29)
Pre- 240 12.41 1.00 7.00 3.09 0.85 0.57
Psychopathology (1.1 1) (0.16) (0.32)
Pre-Maturity 244 10.95 4.00 9.00 6.51 0.12 0.10
(0.91) (0.16) (0.31)

Pre-Risk 238 13.14 1.00 6.00 2.61 0.44 0.23
(0.93) (0.16) (0.31)

Pro-Denial 242 1 1.67 1.00 10.00 3.29 1.22 2.29
(1.60) (0.16) (0.31)

Pre-Impulsivity 241 12.04 2.00 8.00 3.91 0.78 0.72
(1.18) (0.16) (0.31)

Post— 205 25.18 1.00 7.00 2.83 0.91 1.48
Psychopathology (1 .16) (0.17) (0.34)

 

116

Table 2 (cont’d).

 

 

Variable Sample N Percent Min. Max Mean Skew Kurtosis
(max=274) Missing (Std Dev) (Std Err) (Std Err)

Post-Maturity 206 24.82 4.00 8.00 5.65 0.18 -0.28
(0.96) (0.17) (0.34)

Post-Risk 206 24.82 1.00 4.00 1.96 0.59 -0.21
(0.84) (0.17) (0.34)

Post-Denial 206 24.82 1.00 9.00 2.69 1.65 2.57
(1.75) (0.17) (0.34)

' Post-Impulsivity 206 24.82 1.00 7.00 3.42 0.65 0.47
(1.03) (0.17) (0.34)

URICA 1 262 4.38 1.00 5.00 2.93 0.11 -1.24
(1.33) (0.15) (0.31)

URICA 2 265 3.28 1.00 5.00 3.96 -1.24 1.28
(1.03) (0.15) (0.30)

URICA 3 264 3.65 1.00 5.00 4.00 -1.43 2.16
(0.99) (0.15) (0.30)

URICA 4 261 4.74 1.00 5.00 3.89 -1.06 0.61
(1.09) (0.15) (0.31)

URICA 5 260 5.11 1.00 5.00 2.49 0.57 -0.61
(1.24) (0.15) (0.31)

URICA 6 265 3.28 1.00 5.00 2.60 0.38 -1.28
(1.45) (0.15) (0.30)

URICA 7 262 4.38 1.00 5.00 3.70 -0.96 0.51
(1.06) (0.15) (0.31)

URICA 8 264 3.65 1.00 5.00 3.79 -1.09 0.91
(1.03) (0.15) (0.30)

URICA 9 261 4.74 1.00 5.00 2.66 0.41 -0.83
(1.23) (0.15) (0.31)

URICA 10 261 4.74 1.00 5.00 2.95 -0.09 -1.26
(1.29) (0.15) (0.31)

URICA 11 260 5.11 1.00 5.00 2.33 0.76 -0.16
(1.16) (0.15) (0.31)

URICA 12 260 5.11 1.00 5.00 3.86 -0.88 0.65
(0.98) (0.15) (0.31)

URICA 13 258 5.83 1.00 5.00 2.46 0.72 -0.27
(1.16) (0.16) (0.31)

URICA 14 264 3.65 1.00 5.00 4.06 -l.24 1.50
(0.95) (0.15) (0.30)

URICA 15 262 4.38 1.00 5.00 3.43 -0.52 -0.52
(1.16) (0.15) (0.31)

URICA 16 260 5.11 1.00 5.00 2.55 0.50 -0.76
(1.22) (0.15) (0.31)

URICA 17 262 4.38 1.00 5.00 3.56 -0.81 -0.24
(1.16) (0.15) (0.31)

URICA 18 260 5.11 1.00 5.00 2.52 0.39 -0.86
(1.23) (0.15) (0.31)

URICA 19 258 5.83 1.00 5.00 3.19 -0.29 -0.89
(1.22) (0.16) (0.31)

URICA 20 257 6.20 1.00 5.00 3.39 -0.53 -0.55
(1.16) (0.16) (0.31)

URICA 21 260 5.11 1.00 5.00 3.57 -0.79 0.39
(1.01) (0.15) (0.31)

URICA 22 257 6.20 1.00 5.00 3.25 -0.30 -0.88
(1.19) (0.16) (0.31)

 

117

Table 2 (cont’d).

 

 

Variable Sample N Percent Min. Max. Mean Skew Kurtosis
(max=274) Missing (Std Dev) (Std Err) (Std Err)

URICA 259 5.47 1.00 5.00 2.43 (1.17) 0.55 (0.15) -0.60 (0.31)

23

URICA 260 5.11 1.00 5.00 4.04 (0.83) -1.26 (0.15) 2.62 (0.31)

24

URICA 258 5.83 1.00 5.00 4.13 (0.89) -1.41 (0.16) 2.75 (0.31)

25

URICA 257 6.20 1.00 5.00 2.02 (1.03) 1.08 (0.16) 0.89 (0.31)

26

URICA 260 5.11 1.00 5.00 3.52 (1.31) -0.58 (0.15) —0.83 (0.83)

27

URICA 258 5.83 1.00 5.00 2.03 (1.01) 1.05 (0.15) 0.79 (0.31)

28

URICA 257 6.20 1.00 5.00 2.28 (1.07) 0.73 (0.16) —0.16 (0.31)

29

URICA 258 5.83 1.00 5.00 3.98 (0.89) -1.38 (0.16) 2.61 (0.31)

30

URICA 257 6.20 1.00 5.00 2.02 (1.01) 1.04 (0.16) 0.71 (0.31)

31

URICA 258 5.83 1.00 5.00 2.71 (1.23) 0.17 (0.15) -1.01 (0.31)

32

Pre- 266 2.92 1.00 4.43 2.38 (0.73) 0.29 (0.15) -0.41 (0.29)

Contem-

plation

Contem- 266 2.92 1.25 5.00 3.72 (0.77) -0.59 (0.15) 0.37 (0.29)

plation

Action 266 2.92 1.13 5.00 3.72 (0.69) -0.95 (0.15) 1.57 (0.29)

Mainte- 266 2.92 1.00 5.00 2.73 (0.89) 0.19 (0.15) -0.43 (0.29)

nance

MMPI F 274 0.00 36.00 120.00 65.29 0.92 (0.15) 0.06 (0.29)
(20.10)

MMPI L 274 0.00 34.00 91.00 62.23 0.23 (0.15) -0.28 (0.29)
(11.69)

MMPI K 274 0.00 30.00 87.00 55.22 -0.20 (0.15) -0.41 (0.29)
(10.20)

MMPI Hs 274 0.00 30.00 110.00 59.30 0.41 (0.15) 0.60 (0.29)
(12.14)

MMPI D 274 0.00 34.00 104.00 57.38 0.46 (0.15) 0.76 (0.29)
(11.07)

MMPI Hy 274 0.00 31.00 116.00 57.91 0.64 (0.15) 1.50 (0.29)
(11.76)

MMPI Pd 274 0.00 39.00 105.00 65.05 0.35 (0.15) 0.00 (0.29)
(11.22)

MMPI Mf 274 0.00 30.00 84.00 47.94 (8.86) 0.45 (0.15) 0.53 (0.29)

MMPI Pa 274 0.00 31.00 119.00 63.23 0.60 (0.15) 0.44 (0.29)
(14.12)

MMPI Pt 274 0.00 34.00 104.00 57.90 0.69 (0.15) 0.28 (0.29)
(12.53)

MMPI Sc 274 0.00 34.00 120.00 63.97 0.81 (0.15) 0.33 (0.29)
(16.10)

MMPI Ma 274 0.00 35.00 101.00 54.20 1.09 (0.15) 1.64 (0.29)
(10.88)

MMPI Si 274 0.00 32.00 81.00 52.64 (9.25) 0.26 (0.15) -0.05 (0.29L

 

118

Table 3. Distribution of Categorical Variables.

 

 

Variable Sample N Percent of Variable Sample N Percent of
(max=274) Total (max=274) Total
Sample Sample
Race Alcohol at
Offense
Caucasian 200 72.99 No 174 63.50
African-American 38 13.87 Yes 98 35.77
Hispanic 21 7.66 Drugs at
Offense
Other 7 2.55 No 212 77.37
Education Yes 58 21 . 17
Elementary 3 1.09 Victim Status
Junior High 9 3.28 Child 210 76.64
Some High School 67 24.45 Adult 49 17.88
H.S. Degree 94 34.31 Victim
Gender
GED 46 16.79 Male 17 6.20
Some college 36 13.14 Female 252 91.97
Associate’s degree 3 1.09 Relationship
to Victim
Bachelor’s degree 6 2.19 Romantic 34 12.41
Partner
More than BA/BS 1 0.36 Non-romantic 58 21.17
Partner
Marital Status Stepchild 34 12.41
Single 142 51.82 Stranger 40 14.59
Married 46 16.79 Biological 47 17.15
Relative
Div/Sep/W id 69 25.18 Other 57 20.80
Employed Supervision
Status
No 120 43 .79 Probation 166 60.58
Yes 154 56.20 Parole 99 36.13
Offense Type Any Prior
Offenses
CSC 1 17 6.20 No 90 32.85
CSC 2 71 25.91 Yes 160 58.39
CSC 3 62 22.63 Juvenile
Record
CSC 4 122 44.53 No 188 68.61
Penetration Yes 57 20.80
No 108 39.42 Requirement
Complete
Yes 165 60.22 No 65 23.72
Yes 208 75.91

 

119

.35 So a gauge? £8288 ...a.
.63: mod 3 EonGw_m mucus—2.80 .*

 

 

 

€0qu
-- Ivmd wdd bod wdd- vod- dﬁd odd- 32d 8d cod odd Ivmd- 0:52;.
owcoto
dorm
-- 35d 33d cod- Nod- ...Sd Nod- immd aim—d Ed 8d- Ed. has
semen
-- 13d nod «dd- Nod- Nod 2d 2d ...2d *emmd- Immd 5.82
38m
-- ddd- add- Sd odd- ...w _ d *3 .d :vmd :de- :d oﬂtoasm
8:05 9
-- 13d- ...Ed- 32.? mod wdd **dmd- ...Sd- 12d couﬁom
ow<
: odd :cod mod- odd ...c _ .d ﬁend 3d~d 8:05
5980
-- 5d mod \rdd nod mod 9 d- 8:0;
33m
-- mdd- «...Ed 8d aim—d add 8:05
canto
-- :Rd .25 mod- 2.? 33:5
omqowo
“a
-- Vdd 3d mod- 6:82
no:
-- mod- ...m _ d- -goaom
09$.
-- odd- canto
-- ow<
omcobo acmtm Etc; 038.5 omaoto
908% SE E 33% 8 ow< .8280 mBSm “a “a new 093.
2833. be .82 ommtoasm noun—om 8:05 830; 52> amen 3:83. tabocom 33me ow<
.mormtodoﬁwnu omnobO mo mcouﬁobouruouﬁ .mv 033.

120

Table 4b. Correlations between URICA Subscales and Offense Characteristics.

 

 

Pre-Contemplation Contemplation Action Maintenance
Age 008 0.05 0.07 0.16*
Offense Type 0.25** -0.18** -0.18** -0.16*
Penetration -0.10 0.05 0.05 -0.02
Alcohol at -0.17** 0.12* 0.17** 0.15*
Offense
Drugs at -0.12 0.12 0.14* 0.13*
Offense
Victim Status 0.09 -0.13* -0.11 -0.10
Victim 0.03 -0.07 -0.02 -0.08
Gender
Victim Age 017* -0.24** -0.19** -0.14*
Relationship -0. 13* 0.18** 0.15* 0.21**
to Victim
Supervision -0.50** 0.36** 0.39** 0.34**
Status
Months in -0.38** 0.24** 0.30** 0.27**
Prison
Any Prior -0.02 0.12 0.17** 0.19**
Offenses
Juvenile 0.08 -0.09 -0.08 -0.04
Record

 

*. Correlations signiﬁcant at 0.05 level.
**. Correlations significant at 0.01 level.

121

._o>o_ 8d 8 :8chme 30:20:00 .1.

.38“ dd 8 EmoGEwB 8052280 .*

 

 

30qu

8d 1.: .d mod- 5d mod- 2 d- *3 .d odd- Nod- wdd- *2 d- ddd- wed 0:82:

880me

mod ...m _ .d 3d mod Vdd mod- 3..de mod- 5d- vod- uammdt odd- mod 8th 52.

nomtm

mdd nod- wdd- add- d_d- mod wdd vod- Nod odd- mdd- mod- *2 d- 5 3802

*w _ .d 385

5d- odd- mod- mod- - dd *3 .d Ndd Vdd mod- mod odd add- 883.89%

885 8

mod d_d.. mod- mdd mdd- :d mod- Nod ...Sd mod mod- mdd- Nod- 9:80:23“

ow<

odd- 5d- 5d 5d- mdd mod- ddd- vod vod- mod bod odd 8d- 53>

88:00

*2 d- mod mod- *N d d- odd- _ _d- vod add. .18 _ d- ddd- mod wdd mod 885

38w

2d. 3d- mod Nod- vod Nod- 5d mod nod- odd wdd Ndd 5d 885

omaoho

add .18— .d *Ed 2 .d nod Vdd- ...MZd aid- Ndd ...Ed *Lmd- *mﬁd- :d 8 meQ

omaobo

Nod- wdd mod- mdd Sd odd am d .d ddd mod- cod- _ _d- vod- mod. 8 8:82

Nod- vod ddd mod- mod. 8.? Ndd Sd- ddd- odd mod bod 3d- notabowmm

o P

mod- odd vdd 5d Ndd odd- 3d- Vdd nod- mod 2d 5d- ddd nacho

Vdd :dmd- mod- 3d odd aid—d mod- Immd «Ed 35d odd Nod. ddd- ow<
_m «2 om E «m :2 dm mm Q m: M A m

 

8588830 omcoto d5 860m 7:22 8838: 80:29:00 .3 2an

122

.8288 .388 2 ..82880 3888882.. 8:83 88888 on 88 2:8 822880 .e

.822 5d 8 88286 2832880 .2.

.732 mod 8 “80:88 8228.80 .*

 

 

123

3809
I. .1.de .8..de *amm.o 1.de “.2..de sag—Nd :6 3.26 «Ed at. N26- tamom
02:95
-- I _ md aimed Idmd wdd aided aw _ .d *ammd .2..de a -- vod- -Hmom
-- 33d aimed vod 39d 2 .d *ahmd ...admd e -- mod- 222-892
8:82
.. aimed 1.de Immd 33d .1de Ed at odd- -Hmom
$2058
-oaozwm
.. ...Ed Ivmd INNd ...Cd .3..de a: mod -Hmom
28G
-- Idmd 3..de *admd 3..de :d- 33d- -8m
02:95
1 33.0 ..thd :mmd 1&th :det -95
-- *anmd 3mmd Qd- 12d- 2288.2
8:82
-- 33d Immd- Inmd- -8m
32058
-onohmm
-- :d- 1.3 d- -8m
829.80
-- 1m; 8:
808M
-- 2883‘
awe—0522 $20582
28G 02:95 22¢ 8:82 68.2mm 285 8.8:: 8:82 .0822 82980 8082
.28.; -umom 28m -umom -.mom -8m -8m 22805 -8m -8m 83. 2803.

 

.85mta> 308:0 mo 8032880885 .3 28,—.

.2262 Ed 8 2802-2228.;- 822880 .2.
.2082 mod 8 8802-22288 08222880 .*

 

 

c8032
2 d *m 2 d odd 3 d Nod- mod- 2 d vod 22d mod 1.3.? Ide- 028082.
0080200
cod- m2d 2 d ddd- *m 2 d nod- 2 d mdd 2dd mod 02d- 2 d- 82.252 be
iomd- 22 d- bod- ibmd- vod :wmd- 22d- vod 382.c- mod- odd add 082-2- 82.5
33m
3.. :3. *m 2 d- odd- Idmd- 2dd- :cmd- odd- mod N 2 d- mod- 3 d- mod- 82280.2:m
8285 9
odd- udd- odd- m 2 d- mod 5.? ood- Nod 202 d- odd- Nod Nod- n22228028232
vod m 2 d- ~nod. 2 2d- *c2d- Nod mod- 02 d- *3 d- *2: d- mdd- mod- 0w< 8285
802800
wod- 25d ddd Nod vod- mod- wed- 2.? mod mod- Nod- ood E285
masm
2dd- wdd- odd- cod- 2 .0. mod :3 vod- odd- nod- mdd mod- 9:85
0802.20
“1.2.0- 2.0 mod mod .8 2 d odd- 1.2d 2 2d vod cod. 2 2d- vod- 8 meQ
0802,20
282d- wod mod Bd- 2dd add- 2od- vod- Nod- 5.? odd- ~wdd 8 222002
0222 d- odd- ddd mod- Bd- mod- .voo vod odd vod- Ed Sd 282888802
092.2.
mod dd *2 d mod 2 .0. mod *3 d nod- *m 2 d 22d 3 d- *m 2 d- 0802,20
2302022222 30222822
22:09 822282 0222 8:82 82082 28809 82:22:22 022% 8:82 -022920m 082980 208032
-282 -892 28m -2852 -282 .95 -8n2 -08n2 -8m .082 .832 2805‘

 

802828888220 0mc0b0 2x8 m0288> 808220 80382 80282880 .nm 028-2-

124

Table 5c. Correlations between URICA Subscales and Clinical Variables.

 

 

Pre- Contemplation Action Maintenance

Contemplation
Attendance 0.02 -0.07 -0.06 -0.06
Record
Requirement 0.02 -0.07 -0.05 -0.08
Complete
Pre- -0.07 0.05 0.08 0.10
Psychopathology
Pre—Maturity 0.13 -0.06 -0.05 0.01
Pre-Risk -0.05 0.03 0.01 0.08
Pre-Impulsivity 0.04 0.03 0.08 0.02
Pre-Denial 0.36“ -0.27** -0.31** -0.17*
Post- -0.07 0.22" 0.24* * 0.18*
Psychopathology
Post-Maturity 0.23 ** -0. 12 -0.06 -0.13
Post-Risk -0.02 0.07 0.10 0.07
Post-Impulsivity 0.06 0.1 1 0.1 1 0.03
Post-Denial 0.54** -0.30** -0.33** -0.26**

 

‘2 Correlations signiﬁcant at 0.05 level.
"2 Correlations signiﬁcant at 0.01 level.

125

.2202 Ed 8 22802-222820 08282880 .3."
.2902 mod 8 280m22cm2m 28282880 .*

 

2od- wdd- ood- mdd- vod- Ed- 2d- N2d odd *o2d cod 182d vod- 22800-282
53222282

wdd ~2d d2d vdd 222 d Nod- *o2d odd- wdd ddd :NNd- cod Immd -282
mod 02 d add mod add 20d- mod odd odd odd ...m 2 d- odd 2 d 0222-892
odd 182d 2: d 2dd vdd .222d- 2 2d Nod- 2od mod 2 d- mod 182d ~222-582-282
.30222822

..immd 2 2d Iomd .3..de 182d m2d immd ...w2d Idmd *3 d :2Nd- odd 33d 622082-282
add 202 d ...2d 2 2d 2.22 d d2d mod- 2 2d mod *o2d odd- 2od- 2 d 22:32-82
5232:2282

22d d2d :w2d *02d .2 2 md Bd- 2 2d add *m2d 182d nod- Vdd- 2 2d -82
vod 2od ddd vod d2d mod 222.22 222.22 mod 2 2d Nod- mod- vod- 0222-82
«22 d 182 d Immd ...m2d 2 2d odd- 0% 2 d nod ...Ed ...m. 2 d Ed- vod- .1. 2 md N02882-82
302022822

".02 d rod 2 2d mod odd vod nod *m2d m 2 d *3 d 5.? Nod 02 d 822082-82
0202980

0 2 d- ...m 2 d- *02 d- 2 2d- vod- odd- *m 2 d- odd- 2 2d- wdd- d2d 5.0. gm 2 d- 32888232032

208032
1.3 d- mod- *2 d- nod- 2dd- 222 d- *3 .0. mod- 2 d- nod- ndd vod- 2 2 d- 002828052.

 

a a: 8 a «.2 E E 2: a a: 02 2 "2

 

.8285» 2888 Ba 82an E22: 80308 2822.50 .8 02%;

126

Table 6a. Inter-Correlations of URICA Subscales.

 

 

Pre- Contemplation Action Maintenance
Contemplation
Pre- --
Contemplation
Contemplation -0.62** --
Action -0.54** 0.76** --
Maintenance -0.46** 0.72" 0.57** --

 

*. Correlations signiﬁcant at 0.05 level.
**. Correlations signiﬁcant at 0.01 level.

127

.2252 :8 a 88282.. 3082250 ...-
.20>02 mod 8 2802-2282... 08282880 .1.

 

 

- nod 1.2md 1.3.0 Imvd ...m2d fummd N2d idmd 1.de 1:..ovd- bdd- .1.de 2m
- .1.dvd immd .1.mmd vod- 1:..de mod- mod- Nod «*dmd- .1.mmd- 1.2Vd 82
- 1.05.20 165.0 *m2d frond **om.o gimmd .1.mmd ...zd- 22.0- .1.de om
- 1.50.20 “1.2.2.0 1.080 1:..mvd ﬁrmed 1.de *N2d- 1.02.0- _1.m©d #2
- .1.de 1.310 1.2md 1.0212 idvd 1....2de- 132.c- .1.mod am
- odd 1.de 1.02.22 2 2d dod- hdd- ...v2d .232
- ......mmd 1&vd 1....dmd mod- .1.:d- tovd 25
-- :02 :86 :85 m; 28.20 2:
- 1. 20d 2 2d- mod .1. 3.0 Q
- 1.02.22 N2d 1....Nvd m:
- 1:..2vd .1.dmd- v2
- 22d- ..—
- m
2m «2 8 0.2 8 E 20.2 2: a .22 v2 .2 ..2

 

.8280. :22 .6 32.28858 .8 2.3

128

Table 60. Correlations between URICA Subscales and MMPI Scales.

 

 

 

 

 

Pre-Contemplation Contemplation Action Maintenance

MMPI F 0.15* 0.02 0.01 0.12*
MMPI L 0.08 -0.08 0.04 -0.11
MMPI K -0.02 -0.13* -0.09 0.14*
MMPI Hs 0.09 0.03 0.02 0.05
MMPI D 0.02 0.13* 0.12 0.15*
MMPI Hy -0.01 0.01 0.04 -0.02
MMPI Pd -0.09 0.12* 0.09 0.12*
MMPI Mf -0.17** 0.15* 0.18”” 0.12*
MMPI Pa 013* 0.04 0.03 0.07
MMPI Pt 0.07 0.10 0.02 -0.14*
MMPI Sc 0.11 0.07 0.01 0.11
MMPI Ma 0.10 0.00 -0.04 -0.04
MMPI Si -0.04 0.19** 0.12* O.25**

*. Correlations signiﬁcant at 0.05 level.

**. Correlations signiﬁcant at 0.01 level.

Table 7. Scale Reliability Statistics for URICA.

N Scale Items Alpha

Pre-Contemplation 234 8 0.78
Contemplation 243 8 0.88
Action 242 8 0.83
Maintenance 242 8 0.87

 

129

Table 8a. Inter-Item Correlations for URICA Pre-Contemplation Subscale.

 

 

U1 U5 U11 U13 U23 U26 U29 U31
U1 --
US 0.53 --
U1 1 0.48 0.67 --
U13 0.39 0.49 0.49 --
U23 0.23 0.36 0.34 0.36 --
U26 0.21 0.32 0.35 0.27 0.23 --
U29 0.07 0.25 0.27 0.12 0.12 0.39 --
U31 0.15 0.13 0.17 0.09 0.02 0.38 0.40 --

 

Table 8b. Inter-Item Correlations for URICA Contemplation Subscale.

 

 

U2 U4 U8 U12 U15 U19 U21 U24
U2 --
U4 0.53 --
U8 0.54 0.43 --
U12 0.42 0.47 0.41 --
U15 0.48 0.61 0.46 0.50 --
U19 0.39 0.51 0.41 0.52 0.59 --
U21 0.49 0.58 0.39 0.69 0.57 0.61 --
U24 0.29 0.46 0.28 0.53 0.39 0.45 0.59 --

 

Table 80. Inter-Item Correlations for URICA Action Subscale.

 

 

U3 U7 U10 U14 U17 U20 U25 U30
U3 --
U7 0.49 --
U10 0.19 0.38 --
U14 0.44 0.53 0.25 --
U17 0.26 0.37 0.28 0.28 --
U20 0.35 0.46 0.45 0.37 0.41 --
U25 0.32 0.46 0.13 0.51 0.36 0.40 --
U30 0.43 0.54 0.20 0.56 0.38 0.47 0.62 --

 

Table 8d. Inter-Item Correlations for URICA Maintenance Subscale.

 

 

U6 U9 U16 U18 U22 U27 U28 U32
U6 --
U9 0.60 --
U16 0.51 0.45 --
U18 0.56 0.53 0.46 --
U22 0.49 0.43 0.39 0.52 --
U27 0.46 0.40 0.37 0.40 0.57 --
U28 0.44 0.43 0.41 0.63 0.36 0.31 --
U32 0.45 0.34 0.37 0.56 0.37 0.34 0.41 --

 

130

Table 9a. Scale Reliability Statistics for Pre-Treatment Variables.

 

 

N Mean Alpha If Item Deleted
Pre—Psychopathology 223 3.09 0.69
Pre-Maturity 223 6.50 0.69
Pre-Risk 223 2.61 0.71
Pre—Denial 223 3.25 0.77
Pre-Impulsivity 223 3 .91 0.64

 

Table 9b. Scale Reliability Statistics for Post-Treatment Variables.

 

 

N Mean Alpha If Item Deleted
Post-Psychopathology 198 2.81 0.66
Post-Maturity 198 5.63 0.64
Post-Risk 198 1.94 0.67
Post-Denial 198 2.68 0.79
Post-Impulsivity 198 3 .40 0.64

 

131

Table 10a. Initial Iterative Cluster Analysis with Scale Sc (Three-Cluster Solution).

 

 

 

Cluster
1 2 3
(N=101) (N=90) (N=37)
MMPI F 50.58 64.18 84.46
MMPI K 55.64 56.34 47.19
MMPI D 49.78 58.90 67.43
MMPI Pd 59.03 66.12 75.57
MMPI Mf 46.39 49.28 50.16
MMPI Pa 52.59 65.37 76.49
MMPI Pt 48.24 58.88 73.08
MMPI Sc 50.36 64.57 82.27
MMPI Ma 52.09 51.99 61.51

 

Footnote. In examining the means, Cluster 1 is comprised of 101 sex offenders who all have mean MMPI
scale t-scores below the cut-off for clinical signiﬁcance (i.e., 65). One may characterize this group as
having proﬁles within normal limits (Megargee, 2006). Cluster 2 is comprised of sex offenders with
clinically signiﬁcant t-scores on Scales Pd and Pa, with Scales F and Sc close to clinical elevations. This
proﬁle is common in sex offenders and is indicative of hostility, distrust, and violent behaviors (Megargee,
2006). However, because these elevations are below 75, the extent of these behaviors may be mild
(Megargee, 2006). Cluster 3 is comprised of sex offenders with clinically signiﬁcant elevations on Scales
F, D, Pd, Pa, Pt, and Se. Criminals with scores above 75 on these scales tend to be emotionally and
behaviorally unstable (Megargee).

Table 10b. Initial Iterative Cluster Analysis without Scale Sc (Three-Cluster Solution).

 

 

 

Cluster
1 2 3
(N=104) (N=83L =41)
MMPI F 50.30 64.70 83.15
MMPI K 57.03 54.92 47.51
MMPI D 49.79 58.54 67.98
MMPI Pd 58.89 66.25 75.24
MMPI Mf 45.05 50.83 50.54
MMPI Pa 52.94 64.83 76.54
MMPI Pt 48.83 58.02 72.71
MMPI Ma 52.20 51.94 60.39

 

Footnote. The pattern of means in this set of clusters which excludes Scale Sc is similar to the previous

described pattern of clusters which included Scale Sc.

132

Table 11a. Initial Iterative Cluster Analysis with Scale Sc (F our-Cluster Solution).

 

 

 

Cluster
1 2 3 4
(N=103) (N=66L (N=41) (N=18)
MMPI F 51.78 58.29 80.56 84.83
MMPI K 54.59 59.65 50.41 45.00
MMPI D 49.20 61.02 57.66 75.83
MMPI Pd 58.74 66.70 69.24 78.78
MMPI Mf 46.03 50.14 47.02 55.44
MMPI Pa 53.08 65.18 68.05 81.44
MMPI Pt 47.98 59.36 63.80 77.72
MMPI Sc 50.70 63.53 72.39 86.56
MMPI Ma 52.83 48.83 61.80 56.50

 

Footnote. Clusters l and 4 are conceptually similar to Clusters l and 3 in Tables 10a and 10b. Clusters 2
and 3 of this four-cluster solution are similar to Cluster 2 in the three-cluster solution. One difference is
that the Scale F score of Cluster 2 in this solution is lower than the Scale F score of the previously
described Cluster 2. This could mean that the offenders in this Cluster 2 are exaggerating less or
malingering less than the offenders in the previously described Cluster 2. Another difference is that Cluster
3 of this four-cluster solution has higher scores on Scale F, Sc, and Ma, though the Ma t-scores do not reach
clinical signiﬁcance. These offenders may be impulsive and have signiﬁcant problems with thought
disorders (Megargee, 2006).

Table 11b. Initial Iterative Cluster Analysis without Scale Sc (Four-Cluster Solution).

 

 

 

Cluster
1 2 3 4
(N=70) (N=80) (N=52) (N=26)
MMPI F 51.34 54.10 77.35 79.46
MMPI K 53.17 61.60 49.62 46.42
MMPI D 46.91 58.15 57.29 73.42
MMPI Pd 56.80 64.86 67.19 78.85
MMPI Mf 45.13 49.16 47.69 54.00
MMPI Pa 51.77 60.85 66.35 80.12
MMPI Pt 46.46 56.36 60.37 75.96
MMPI Ma 54.79 47.49 59.33 57.58

 

Footnote. The pattern of means in this set of clusters which excludes Scale Sc is similar to the pattern of
means which included Scale Sc. One noticeable difference is that the means of these clusters are generally
lowering, making some of the scales which were clinically signiﬁcant in the solution which included Scale
Sc no longer so (e.g., Scales Pd and Pa of Cluster 2). This could be due to the high correlation between
Scale Sc and Pa. Please see main text for further explanations.

133

Table 12a. Initial Iterative Cluster Analysis with Scale Sc (Five-Cluster Solution).

 

 

 

Cluster
1 2 3 4 5
(N=74) (N=60) (N=49) (N=35) (N=10)
MMPI F 48.53 54.48 71.02 80.31 85.90
MMPI K 54.57 62.28 50.49 49.63 45.10
MMPI D 48.78 58.90 54.45 70.14 55.70
MMPI Pd 56.64 67.15 63.49 74.94 75.50
MMPI Mf 45.55 49.45 48.24 52.86 42.40
MMPI Pa 51.99 61.83 62.86 78.77 63.10
MMPI Pt 46.61 57.22 56.47 74.86 60.60
MMPI Sc 48.59 61.30 62.61 80.31 78.80
MMPI Ma 52.58 48.37 54.82 55.57 79.20

 

Footnote. Cluster 1 is comprised of sex offenders with t-scores within normal ranges. Cluster 2 is
comprised of a single elevation on Scale Pd. Offenders with this proﬁle tend to be dishonest, manipulative,
and have unstable relationships (Megargee, 2006). Cluster 3 is comprised of a single elevation on Scale F.
These offenders may be intentionally exaggerating their symptoms or randomly responding to the questions
(Megargee, 2006). Cluster 4 is characterized by elevations on Scales F, D, Pd, Pa, Pt, and Sc similar to a
previously described cluster which was conceptualized as emotionally and behaviorally unstable. Cluster 5
is characterized by elevations on Scales F, Pd, Sc, and Ma. Offenders with this proﬁle are oﬁen irritable
and difficult to manage, even in correctional settings (Megargee, 2006). They may need further evaluation
for the presence of thought disorders or suicidal ideation (Megargee, 2006).

Table 12b. Initial Iterative Cluster Analysis without Scale Sc (Five-Cluster Solution).

 

 

 

Cluster
1 2 3 4 5
(N=84) (N=60) (N=40) (N=29) (N=15)
MMPI F 48.07 62.48 65.18 82.97 80.67
MMPI K 58.44 58.28 48.63 48.45 45.40
MMPI D 50.77 61.73 47.58 71.66 58.27
MMPI Pd 59.92 66.70 59.15 74.59 76.33
MMPI Mf 45.04 50.22 49.15 53.31 44.53
MMPI Pa 52.45 66.33 58.35 78.45 68.67
MMPI Pt 48.95 60.50 49.33 75.07 65.53
MMPI Ma 50.55 49.63 57.05 53.59 77.07

 

Footnote. The pattern of means in the set of clusters excluding Scale Se is similar to the pattern of means
including Scale Sc. One difference is, when Scale So is removed, Scales Pa and Pt in Cluster 5 reach
clinical signiﬁcance which suggests that, because these variables highly correlate with Sc, they may be
treated as one variable when Se is included and their individual variance is being overlooked. After
examining these means, in general, there is no conceptual difference between the cluster solutions with and
without Scale Sc. Because of this and the high correlations between Scale Sc and Scales F, Pa, and Pt (see
Table 6b), a more conservative approach was taken and Scale Sc was excluded from the subsequent cluster
analyses.

134

Table 13a. Means for Three-Cluster Solution (Agglomerative Hierarchical).

 

 

 

Clusters
l 2 3
(N=107) (N=62) (N=59)
MMPI F 50.59 61.79 80.78
MMPI K 56.95 55.23 49.47
MMPI D 50.12 60.13 63.27
MMPI Pd 58.16 68.40 71.95
MMPI Mf 45.19 52.53 48.89
MMPI Pa 54.22 63.98 72.14
MMPI Pt 49.72 57.74 67.37
MMPI Ma 51.60 50.73 60.17

 

Footnote. Cluster 1 is composed of sex offenders whose t-scores are within normal ranges, which means
they could be classiﬁed as having psychological proﬁles within normal limits (Megargee, 2006). Cluster 2
is composed of sex offenders who have t-score elevations on scales Pd only; they may be characterized as
dishonest and manipulative (Megargee, 2006). Cluster 3 has t-score elevations on Scales F, Pd, Pa, and Pt.
Offenders with this proﬁle tend to be suspicious, hostile, narcissistic and aggressive (Megargee, 2006).

Table 13b. Means for Four-Cluster Solution (Agglomerative Hierarchical).

 

 

 

Clusters
l 2 3 4
(N=107) (N=62) (N=23) (N=36)
MMPI F 50.59 61 .79 77.43 82.92
MMPI K 56.95 55.23 49.48 49.47
MMPI D 50.12 60.13 57.65 66.86
MMPI Pd 58.16 68.40 74.39 70.39
MMPI Mf 45.19 52.53 44.61 51.61
MMPI Pa 54.22 63.98 62.35 78.39
MMPI Pt 49.72 57.74 60.96 71.47
MMPI Ma 51.60 50.73 71.35 53.03

 

Footnote. Clusters 1 and 2 of the three-cluster solution were repeated in the four-cluster solution. Cluster 3
of the three-cluster solution was divided into two separate clusters. Cluster 3 of the four-cluster solution is
characterized by elevations on Scales F, Pd, and Ma. These offenders may be impulsive, manipulative, and
aggressive (Megargee, 2006). Cluster 4 was characterized by elevations on scales F, D, Pd, Pa, and Pt.
Offenders with this proﬁle tend to be emotionally and behaviorally unstable.

135

Table 13c. Means for Nine-Cluster Solution (Agglomerative Hierarchical).

 

 

 

Clusters
l 2 3 4 5 6 7 8 9
(N=50) (N=29) (N=28) (N=29) (N=26) (N=7) (N=23) (N=31) (N=5L
F 54.14 46.52 48.46 63.90 55.73 75.57 77.43 82.58 85.00
K 55.54 63.31 52.89 58.97 54.54 42.26 49.48 50.81 41.20
D 53.08 52.90 41.96 63.28 60.00 47.57 57.65 63.42 88.20
Pd 54.16 61.59 61.75 64.14 73.85 65.86 74.39 68.03 85.00
Mf 41.74 51.10 45.21 52.38 50.54 60.57 44.61 50.90 56.00
Pa 50.32 61.10 54.07 63.69 64.54 63.14 62.35 76.52 90.00
Pt 49.58 53.28 46.29 63.55 54.42 46.00 60.96 69.42 84.20
Ma 48.88 48.28 59.89 46.07 55.81 51.14 71.35 51.97 59.60

 

Footnote. Clusters 1, 2, 3, and 4 are each characterized by t-scores within normal ranges. Cluster 5 is
characterized by a single elevation on Scale Pd with Scale Pa being just below clinical levels. These
offenders are characterized by hostility and narcissism (Megargee, 2006). Cluster 6 is comprised of sex
oﬁ‘enders with elevations on Scales F and Pd. These offenders may be manipulative and dishonest
(Megargee, 2006). Cluster 7 is characterized by elevations on Scales F, Pd, and Ma. These offenders may
be dishonest and impulsive (Megargee, 2006). Cluster 8 is comprised of sex offenders with elevations on
Scales F, Pd, Pa, and Pt. These offenders may be dishonest, mistrusting, and anxious (Megargee, 2006).
Cluster 9 is characterized by elevations on Scales F, D, Pd, Pa, and Pt. These offenders may be
emotionally unstable and prone to hostility (Megargee, 2006). Based on these results the four-cluster
solution may capture the heterogeneity of psychopathology in sex offenders better than the three-cluster
solution without being repetitive as in the nine-cluster solution. To further examine this hypothesis, the
four-cluster solution was tested for stability in the dataset.

136

Table 14. Final Iterative Cluster Analysis (Four-Cluster Solution).
Cluster
Manipulative

 

 

WNL Hostile Disturbed Sample

 

(N=86) (N=69) (N=36) (N=37) (N=228)
MMPI F 49.83 57.84 75.92 81.11 64.14
MMPI K 55.83 59.65 47.47 48.95 54.55
MMPI D 49.00 59.77 53.06 69.62 56.25
MMPI Pd 56.74 66.70 67.67 75.43 64.51
MMPI Mf 45.03 50.19 47.00 52.65 48.14
MMPI Pa 51 .97 64.52 61.11 78.49 61.51
MMPI Pt 47.87 58.01 55.89 74.14 56.47
MMPI Ma 52.15 48.68 63.42 56.46 53.58

 

Footnote 1. The means of the ﬁnal iterative solution are similar to those in the hierarchical solution, with
the exception that the Scale Ma of the Manipulative cluster is no longer clinically signiﬁcant. Please see
main text for conceptual descriptions.

Footnote 2. The sample means were included in this table for comparative purposes only. These means
were not included as a cluster in the analyses.

137

Table 15. Case Description of Cluster Assignment.

 

 

 

Hierarchical Final Iterative Hierarchical Final Iterative
Cluster Cluster Cluster Cluster
Case # Membership Membership Case # Membership Membership
1 Hostile Hostile 41 WNL WNL
2 Hostile Manipulative 42 Manipulative Manipulative
3 Disturbed Manipulative 43 WNL WNL
4 Hostile Hostile 44 WNL Hostile
5 WNL WNL 45 Manipulative Manipulative
6 Disturbed Manipulative 46 Manipulative Manipulative
7 WNL WNL 47 Hostile Manipulative
8 Manipulative Manipulative 48 WNL Hostile
9 WNL WNL 49 Hostile Hostile
10 WNL WNL 50 Disturbed Disturbed
1 1 WNL Hostile 51 WNL WNL
12 WNL WNL 52 WNL WNL
13 WNL WNL 53 WNL WNL
14 WNL WNL 54 WNL WNL
l 5 Disturbed Disturbed 55 WNL WNL
16 WNL WNL 56 WNL WNL
17 WNL WNL 57 Hostile Hostile
1 8 WNL Hostile 5 8 Manipulative Disturbed
1 9 Hostile Hostile 59 Hostile Hostile
20 WNL WNL 60 Disturbed Disturbed
2 1 WNL WNL 61 Disturbed Disturbed
22 Hostile Manipulative 62 WNL WNL
23 Hostile Hostile 63 WNL WNL
24 WNL WNL 64 WNL WNL
25 WNL WNL 65 Hostile Manipulative
26 Hostile Hostile 66 Hostile Hostile
27 Hostile Disturbed 67 WNL Hostile
28 Disturbed Disturbed 68 WNL WNL
29 WNL WNL 69 WNL WNL
30 Hostile Manipulative 70 WNL WNL
3 l Manipulative Disturbed 71 WNL Hostile
32 WNL Hostile 72 WNL WNL
33 WNL WNL 73 Hostile Hostile
34 WNL Hostile 74 Hostile Hostile
35 WNL WNL 75 Disturbed Disturbed
36 WNL WNL 76 WNL WNL
37 Disturbed Disturbed 77 WNL WNL
38 Hostile Hostile 78 Manipulative Disturbed
39 WNL WNL 79 WNL Manipulative
40 WNL WNL 80 WNL Hostile

 

 

138

Table 15 (cont’d).

 

 

Hierarchical Final Iterative
Cluster Cluster
Case # Membership MembershiL

81 WNL WNL
82 Manipulative Manipulative
83 Hostile Hostile
84 Disturbed Disturbed
85 Hostile Disturbed
86 WNL WNL
87 WNL WNL
88 Hostile Hostile
89 WNL WNL
90 Manipulative Manipulative
91 Manipulative Manipulative
92 WNL WNL
93 Hostile Hostile
94 Hostile Hostile
95 Disturbed Disturbed
96 WNL WNL
97 Disturbed Hostile
98 Hostile Hostile
99 Disturbed Disturbed
100 Hostile Hostile
101 Manipulative Manipulative
102 WNL WNL
103 WNL Manipulative
104 Manipulative Manipulative
105 WNL WNL
106 Hostile Hostile
107 Manipulative Disturbed
l 08 Disturbed Disturbed
1 09 WNL WNL
1 10 WNL WNL
1 1 1 WNL WNL
1 12 Disturbed Disturbed
l 13 Hostile Hostile
1 14 Hostile WNL
l 15 WNL WNL
1 l6 WNL WNL
1 17 Hostile Hostile
1 18 Manipulative Manipulative
l 19 Disturbed Disturbed
120 Disturbed Disturbed
121 Manipulative Disturbed
l22 WNL WNL

 

 

 

Hierarchical Final Iterative
Cluster Cluster
Case # Membership Membership

123 WNL WNL
124 WNL WNL
125 WNL WNL
126 WNL WNL
127 Hostile Hostile
128 Hostile Hostile
129 WNL WNL
130 Hostile Hostile
131 Hostile Hostile
132 WNL Hostile
133 WNL WNL
134 WNL WNL
135 Manipulative Manipulative
1 36 Disturbed Disturbed
137 WNL WNL
138 WNL WNL
139 WNL WNL
140 WNL WNL
141 WNL WNL
142 WNL WNL
143 I-Iostile Hostile
144 WNL WNL
145 Disturbed Disturbed
146 Hostile Hostile
147 WNL WNL
148 Hostile Hostile
149 WNL WNL
150 WNL WNL
151 WNL WNL
152 Hostile Hostile
153 Hostile Hostile
154 WNL WNL
155 WNL WNL
1 56 Disturbed Disturbed
157 Disturbed Disturbed
1 58 Disturbed Manipulative
159 Hostile Hostile
160 WNL WNL
161 Hostile Manipulative
162 WNL WNL
163 WNL WNL
164 WNL Manipulative

 

139

Table 15 (cont’d).

 

 

 

 

Hierarchical Final Iterative Hierarchical Final Iterative
Cluster Cluster Cluster Cluster
Case # Membership Membership Case # Membership Membership

165 Hostile Hostile 197 Disturbed Disturbed
l 66 WNL Hostile 1 98 Hostile Manipulative
167 WNL WNL 199 Disturbed Manipulative
168 Disturbed Hostile 200 WNL Hostile
169 Hostile Hostile 201 WNL WNL
l 70 Hostile Manipulative 202 WNL Hostile
171 Manipulative Disturbed 203 Disturbed Disturbed
1 72 Disturbed Disturbed 204 Disturbed Hostile
1 73 Hostile Hostile 205 Manipulative Manipulative
1 74 WNL WNL 206 Manipulative Manipulative
1 75 WNL WNL 207 Hostile Manipulative
1 76 WNL Hostile 208 Disturbed Disturbed
1 77 Manipulative Manipulative 209 WNL WNL
178 Hostile Hostile 210 Manipulative Manipulative
179 Disturbed Disturbed 21 1 Hostile Hostile
1 80 Hostile Manipulative 212 Disturbed Disturbed
181 Hostile Hostile 213 WNL Hostile
l 82 Hostile Disturbed 214 Hostile Hostile
.183 WNL Hostile 215 WNL Hostile
1 84 Hostile Hostile 216 WNL Hostile
1 85 Hostile Manipulative 21 7 Hostile Hostile
1 86 Hostile Hostile 21 8 Hostile Hostile
l 87 Manipulative Manipulative 219 WNL WNL
188 WNL WNL 220 WNL WNL
189 WNL WNL 221 WNL WNL
190 WNL Hostile 222 Disturbed Disturbed
1 91 Hostile Manipulative 223 Disturbed Hostile
l 92 WNL Hostile 224 Manipulative Manipulative
193 Hostile Hostile 225 Hostile Hostile
194 Disturbed Disturbed 226 WNL WNL
195 Disturbed Disturbed 227 Hostile WNL
196 Hostile Hostile 228 Disturbed Disturbed

 

 

Footnote. According to this table, 27% of cases changed from the Hostile cluster to another cluster.
Twenty-two percent of cases changed from the Disturbed cluster to another cluster. Twenty-one percent of
cases changed from the WNL cluster to another cluster and 26% of cases changed from the Manipulative
cluster to another cluster. A total of 76% of the sample did not change their cluster assignment from the

hierarchical solution to the iterative solution.

140

Table 16a. MANOVA for Cluster Membership and MMPI scales.

 

 

 

df F Eta-Squared

Pillai’s Trace 24, 657 22.63*** 0.45

Wilks’ Lambda 24, 629.97 34.87*** 0.56

Hotelling’s Trace 24, 647 51.45*** 0.66

Roy’s Largest Root 8, 219 130.61*** 0.83
*p < .05.
I"“p < .01.
***p < .001.

Table 16b. Between-Subject Effects for MANOVA in Table 16a.

 

 

 

df F Eta-Squared
MMPI F 3, 224 125.85*** 0.63
MMPI K 3, 224 21.19*** 0.22
MMPI D 3, 224 57.64*** 0.44
MMPI Pd 3, 224 43.94*** 0.37
MMPI Mf 3, 224 8.87*** 0.11
MMPI Pa 3, 224 89.06**"‘ 0.54
MMPI Pt 3, 224 107.05*** 0.59
MMPI Ma 3, 224 22.38*** 0.23
*p < .05.
“p < .01.
mp < .001.

141

Table 17. Summary of Cross-Validation Analyses for Demographic, Offense, and
Clinical Variables.

 

 

 

Clusters Statistics
WNL Hostile Manipulative Disturbed

N (max=228) (N=86) (N=69) (N=36) (N=3?) 112 (df) F (dt)

Race 8 81% 68% 64% 73% 5.45

(% White) (3) '-

Educaﬁon a 71% 74% 70% 71% 3 .81 '-

(% High School or (59/ 83) (50/68) (23/33) (25/ 35) (6)

more)

Marital Status 3 41% 55% 67% 50% 3.33 --

(% Ever Married) (34/82) (34/62) (14/21) (17/34) (3)

Employment 57% 54% 53% 62% 0.91 --

(% Employed) (3)

Parenthood 52% 74% 50% 59% 9.14* --

(% With Children) (3)

Offender’s Age 31_46° 3933 b 31.69 c 34.88 -- 7.78***

(Mean, Std. Dev.) (9.90) (13.20) (11.34) (9.65) (3, 215)

Offense Type 47% 41% 39% 46% 12.12 --

(% With 411: Degree) (40/85) (28/68) (9)

Penetration 63% 64% 53% 67% 1 .75 --

(% Who Used It) (24/36) (3)

Alcohol at Offense 36% 40% 39% 38% 0.24 --

(% Who Used It) (27/68) (3)

Drugs at Offense 17% 19% 39% 28% 7.56 --

(% Who Used It) (13/67) (10/36) (3)

Victim Status 80% 82% 89% 74% 2.43 --

(% With Child (63/79) (54/66) (31/35) (26/35) (3)

Victims)

Victim Gender 98% 87% 94% 95% 7.34 --

(% With Female (83/85) (59/68) (32/34) (3)

Victims)

Relationship to 64% 69% 53% 58% 3.03

Victima (54/85) (47/68) (21/36) (3) --

(% Who Knew

Victim)

Supervision Status 62% 62% 57% 67% 0.68

(% On Probation) (52/84) (41/66) (20/35) (24/36) (3) --

Prior Adult Offenses 59% 59% 78% 72% 5.35

(% With Prior (45/76) (37/63) (25/32) (26/36) (3) --

Offenses)

Juvenile Record 22% 18% 41% 18% 7.26 --

(% With Record) (16/74) (1 1/62) (13/32) (6/34) (3)

Victim’s Age 13.63 14.14 12.26 14.22 -- 0.57
JMean, Std. Dev.) (4.99) (9.10) (6.1 Q (6.09) (3, 194)

 

a = Results unchanged when using three or more divisions.

1” c = Means with different superscripts signiﬁcantly differ from each other but not from those lacking
subscripts. Means with the same superscript do not differ ﬁ'om each other but do differ from a different
(marked) cluster. Means without superscripts do not signiﬁcantly differ from any of the other clusters for
that variable.

142

Table 17 (cont’d).

 

 

 

Clusters Statistics

WNL Hostile Manipulative Disturbed
N (max=228) (N=86) (N=69) (N=36) (N=37) X2 (df) F ((11)
Prison Time 33.21 28.80 31.24 33.77 -- 0.16
(Months) (48.98) (36.11) (38.95) (48.97) (3, 223)
(Mean, Std. Dev.)
Treatment 85% 81% 64% 76% 7.18 --
Completion - (3)
(% Who Completed)
Attendance Record 25.43b 24.84 2057C 22.84 -- 4.12**
(Max=30) (6.84) (7.84) (10.23) (9.56) (3. 223)
(Mean, Std. Dev.)
Pre-Treatment Score 13,16” 19.31 20,04° 19.09 -- 2.58
(Mean, Std. Dev.) (3.38) (4.04) (3.60 (3.54) (3, 223)
Post-Treatment 13.26 13.00 11.06 12.35 -- 0.89
Score2 (6.72) (7.51) (8.91) (8.00) (3, 223)
(Mean, Std. Dev.)
Clinical Change -2.06 -2.45 -1.40 -1.74 -- 0.64
300,61 (4.12) (4.02) (3.91) (3.24) (3, 223)
(Mean, Std. Dev.)
Pre-Denial Score 3.07 3.52 3.56 3.12 -- 1.70
(Mean, Std. Dev.) (1.40) (1.63) (1.64) (1.30) (3, 223)
Post.Denjal Score 2.18 2.36 1.69 1.59 -- 1.84
(Mean, Std. Dev.) (1.73) (2-12) (1 -91) (1-69) (3. 223)
Denial Change -0.34 -0.44 -0.40 -0.59 -- 0.28
Score1 (1.42) (1.60) (1.44) (1.00) (3, 223)
(Mean, Std. Dev.)
Motivation to Change 7.56b 7.64 8.31 391“ -- 3.02*
(Mean, Std. Dev.) (Z-SEL (2.59) (2.45) (2.23) (3, 223)

 

1” c = Means with different superscripts signiﬁcantly differ from each other but not from those lacking

subscripts. Means with the same superscript do not differ from each other but do differ from a different
(marked) cluster. Means without superscripts do not signiﬁcantly differ ﬁom any of the other clusters for
that variable.
I = These variables are scored such that higher scores indicate higher levels of psychopathology. Thus, the
negative values are indicative of a decrease in psychopathology or denial rather than a worse treatment
response.

= The Missing values covariate was signiﬁcant for these two variables (p < .05). This result is not
surprising given that these two variables had a large number of missing values.

*p < .05.
**p < .01.

143

Table 18a. Hierarchical Regression for Cluster Membership and Motivation to Change
Predicting Post-Treatment Scores.

 

 

Variable B SE B B
Motivation to Change -0.63 0.20 -0.21 **
Dummy Code—Hostile -0.22 1.20 -0.01
Dummy Code—Manipulative -1.73 1.48 -0.08
Dummy Code--Disturbed -0.06 1.48 -0.00

 

Note. R2 = 0.05 (p < .01).
Mp < .01.

Table 18b. Hierarchical Regression for Cluster Membership and Motivation to Change

Predicting Post-Denial Scores.

 

 

Variable B SE B 13
Motivation to Change -0.29 0.05 -0.38***
Dummy Code—Hostile 0.20 0.28 0.05
Dummy Code—Manipulative -0.27 0.35 -0.05
Dummy Code--Disturbed -0.20 0.35 -0.04

 

Note. R2 = 0.16 (p < .001).
***p<.001.

Table 18c. Hierarchical Regression for Cluster Membership and Motivation to Change

Predicting Clinical Change.

 

 

Variable B SE B B
Motivation to Change 0.09 0.11 0.06
Dummy Code—Hostile -0.40 0.64 -0.05
Dummy Code—Manipulative 0.59 0.78 0.06
Dummy Code--Disturbed 0.19 0.79 0.02

 

Note. R2 = 0.01(n.s.).

Table 18d. Hierarchical Regression for Cluster Membership and Motivation to Change

Predicting Denial Change.

 

 

Variable B SE B B
Motivation to Change -0.04 0.04 -0.07
Dummy Code—Hostile -0.10 0.23 -0.03
Dummy Code——Manipulative -0.03 0.28 -0.01
Dummy Code--Disturbed -0.20 0.29 -0.05

 

Note. 1?.2 = 0.01(n.s.).

144

APPENDIX B

Figures for Statistical Results

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l7-c1uster 9-cluster 4-c1uster 3-c1uster 2-c1uster l-cluster
solution solution solution solution solution solution
N=13
N=8 N=29
N=8
N=28 ‘ N=28 N=107 N=107
N=11
=22 N=50 ‘
N=17
=7 m =7 N=228
N=26 ~ N=26 =62 l - =62
=8
)l N=29
N=21 ‘
N=121
=8
=3 N=23 N A N=23
N=12
N=59
N=5 ‘ =5
N=12 N=36
N=3l
N=19

 

 

 

 

Figure 1. Diagram of Agglomerative Hierarchical Cluster Analysis (N=228).

Footnote. Based on the results from the hierarchical analyses, it was determined that the 17-cluster solution
produced clusters which contained N’s too small to detect effects during the cross-validation analyses.

Also the two-cluster solution did not seem to reﬂect enough of the heterogeneity of this sample of sex
offenders. Thus, the means from the nine-, four-, and three-cluster solutions were examined to determine
which of those solutions best ﬁt the data (see Tables 13a-13c). Because of the iterative results (see Tables
lOa-12b) which indicated that that four cluster solution was able to produce a conceptually meaningﬁil
cluster that the three-cluster solution could not, it was hypothesized that the four-cluster solution would
best ﬁt the data.

145

14—cluster 8-c1uster 5-cluster 4—cluster 3-cluster 2-cluster l-cluster
solution solution solution solution solution solution solution

     

 

N=47 I—pl N=47 ]—>[ N=47

 

 

Figure 2. Diagram of Agglomerative Hierarchical Cluster Analysis for Random Half of
Dataset (N=114).

 

 

 

Cluster
1 2 3 4

01:33) (N=47) (N=25) (N=3)
MMPI F 62.65 48.43 80.88 77.00
MMPI K 53.65 56.98 52.08 46.50
MMPI D 57.65 50.11 65.92 46.25
MMPI Pd 56.26 61.57 72.20 71.25
MMPI Mf 52.21 45.66 51.28 47.00
MMPI Pa 61.56 54.51 74.12 68.88
MMPI Pt 54.79 49.21 68.96 53.50
MMPI Ma 46.62 52.77 56.68 73.50

 

Footnote. This ﬁgure and Figure 3 summarize the split-half stability test used to test the reliability of the
four—cluster solution in the dataset. The full sample was randomly split in half and a hierarchical cluster
analysis was performed in each half with the goal of ﬁnding a four-cluster solution in each half of the data.
Figure 2 demonstrates that, similar to the full dataset, this random sample of 114 cases yields a four-cluster
solution. This result suggests that the four-cluster solution is stable. The four-cluster means are
highlighted in accordance with the ﬁndings of the hierarchical analysis.

146

1 5-clustcr 7-cluster 5-cluster 3—cluster 2-cluster l-cluster

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

solution solution solution solution solution solution
N=9
N=9 N=24 N=24 N=24
N=6
N=8
N=85
N=4 N= 1 7 N= 1 7
N=5
N=6l
N=l 7 N=17
N=12 N=44 N= 1 14
N=27
N=15
N=3 N=3 N=3
N=3
N=6 N=29 N=29
=3
N=9 N=26
N=6 N=20
N=5

Figure 3. Diagram of Agglomerative Hierarchical Cluster Analysis for Other Random
Half of Dataset (N=114).

 

 

 

Cluster
1 2 3 4 5

(N=44) (N=26) (N=17) (N=24) (N=3)
MMPI F 55.89 80.42 64.24 45.13 80.33
MMPI K 62.77 48.50 47.12 54.63 42.00
MMPI D 56.59 65.08 51.82 47.67 94.67
MMPI Pd 68.70 72.04 64.35 52.96 88.67
MMPI Mf 47.59 48.62 45.53 46.38 50.67
MMPI Pa 59.75 72.81 57.35 50.13 88.67
MMPI Pt 55.89 69.54 54.71 45.29 87.67
MMPI Ma 50.07 56.81 60.59 52.33 60.00

 

Footnote. A four-cluster solution was not found in this half of the dataset, which is not uncommon given
the small sample size (Rapkin & Luke, 1993). Thus, because the four-cluster solution maximizes
heterogeneity and still leaves enough power for cross validation, it was determined to best ﬁt the data for
this sample. The means ﬁ'om the four-cluster solution of the hierarchical analysis (see Table 13b) were
used as the initial cluster centers (i.e., starting points) for the ﬁnal iterative solution. This yielded the ﬁnal
four-cluster solution which was used in the analyses (means described in Table 14; depiction in Figure 4).

147

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Hierarchical Cluster Solution Iterative Cluster Solution
=84
WNL ‘ N=20 WNL
(N=107) N= (N=86)
N:
_N:_
Hostile , N=45 Hostile
(N=62) N=12 (N=69)
N=3
. . _N;_
Mampulatrve T Manipulative
N=l7 (N )
N=6
, .N;
Drsturbed ' N=4 Disturbed
=36 3——— =
(N ) N=4 (N 37)
N=28

 

 

 

Figure 5. Case Migration from Hierarchical Solution to Final Iterative Solution.

Footnote. In summary, from the hierarchical WNL cluster, 84 cases moved to the iterative WNL cluster;
20 cases moved to the iterative Hostile cluster; 3 moved to the iterative Manipulative cluster; and zero
cases moved to the iterative Disturbed cluster. From the hierarchical Hostile cluster, 2 cases moved to the
iterative WNL cluster; 45 cases moved to the iterative Hostile cluster; 12 moved to the iterative
Manipulative cluster; and 3 moved to the iterative Disturbed cluster. From the hierarchical Manipulative
cluster, zero cases moved to the iterative WNL or the iterative Hostile clusters; 17 cases moved to the
iterative Manipulative cluster; and 6 moved to the iterative Disturbed cluster. From the hierarchical
Disturbed cluster, zero cases moved to the iterative WNL cluster; 4 cases moved to the iterative Hostile
cluster; 4 moved to the iterative Manipulative cluster; and 28 moved to the iterative Disturbed cluster. In
each hierarchical cluster, approximately 25% of the cases (range: 21% to 27%; see footnote of Table 15)
moved to a different cluster in the iterative solution. Reasons for this switch are provided in the main text.

148

9 9 9 9

 

 

 

 

Pre- , Contemplation Action Maintenance
Contemplation

 

 

 

 

 

 

 

 

 

 

  

 

 

      
    

      

-0.66*”

     

0.96***

 

078*" 073*"

Motivation to
Change

Chi-square=2.97; df=2; p=0.23; GFI=O.99; CFI=1.00; RMSEA=0.05; BIC=46.40
Figure 6. SEM Analysis Testing Motivation to Change as a Single Latent Construct.

Footnote. Standardized estimates are reported.
***. p < .001.

149

 

1!. . .

.Pre-Contem-
plation

 

 

Contem-
plation

 

 

 

 

Action

 

 

 

 

‘ Maintenance

9 f f f

 

 

 

0.96* *
Motivation
to Change
0.78* *
0.73 *

 

Pre- ‘
Treatment ﬂ
0 20... Scores

0.29”

 

 

0.02

 

 

 

Post-
Treatment
Scores

 

 

Chi-square=15.l; df=8; p=0.06; GFI=O.98; CFI=O.99; RMSEA=0.06; BIC=85.71

Figure 7. SEM Analysis Validating Motivation to Change Using Treatment Variables.

Footnote. Standardized estimates are reported.

***. p < .001.

150

7!

 

if

APPENDIX C

Figures and Tables for Introduction

 

   
  
   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Stages of Change
Maintenance
DB: pros outweigh cons
PC: counter-conditioning;
stimulus control;
contingency management;
Action
DB: pros outweigh
cons
PC: self-liberation
Preparation
DB: cons outweigh
pros
PC: self re-evaluation
Contemplation
DB: cons outweigh pros
PC: consciousness raising;
social re-evaluation
Pre-
Contemplation

 

 

 

Figure Cl. Trans—Theoretical Model of Behavior Change

Footnote. PC=Processes of change; DB=Decisional balance.

151

Relapse

-A‘_— _. ‘-

 

Table Cl. MMPI-2 Validity Scales (Megargee, 2006).

 

 

Validity Scale Description Mean Cut-off
(Std. Dev.)

(In)Frequency (F) Captures invalid proﬁles due to poor 59.49 2 100
reading and/or comprehension; also (22.70)
detects exaggerating symptoms

Lie (L) Detects those presenting themselves in 57.26 2 80
an overly positive manner (1 1.81)

Correction (K) Empirically derived assessment of 51.49 2 80
symptom minimization and positive (11.65)

impression management

 

Table C2. Description of MMPI-2 clinical scales (Butcher & Williams, 2000).

 

 

Subscale DescriLtion

Hypochondriasis (scale Hs) Individual endorses vague somatic symptoms

Depression (scale D) Individual’s subjective experience of dysthymia and
associated symptoms (e. g., hopelessness)

Hysteria (scale Hy) Measure of speciﬁc somatic symptoms, and
socialization

Psychopathic Deviant Measure of maladjustment, family/authority

(scale Pd) concerns, alienation (ﬁom self and society), and
boredom

Masculine/Feminine Assesses interests, hobbies, and other personality

(scale Mt) traits that may be corrrrnon among men and women

Paranoia (scale Pa) Assesses the degree to which an individual endorses

being morally-driven, untrusting of others, and
interpersonally aware

Psychasthenia (scale Pt) Measures abnormal fears, trouble concentration,
feelings of guilt, and self-criticism
Schizophrenia (scale Sc) Assesses the degree to which an individual endorses

bizarre thought processes, social isolation, poor
judgment, and an unstable self-identity

Hypomania (scale Ma) Measures mood elevation and instability, agitation,
and racing droughts

Social Introversion (scale Si) Measure of social anxiety, isolation, and low self-
esteem

152

 

tg~v_“ ha! - _ .'
.1

Table C3. Processes of Change

Experiential Factor
Consciousness raising

Self-reevaluation

Social reevaluation

Self-liberation
Dramatic relief

Behavioral Factor
Social liberation

Counter-conditioning
Stimulus control

Contingency
management
‘ Helping relationships

Processes of Change
Description
learning more about the problematic behavior and why it is
problematic
assessing the extent to which the problematic behavior ﬁts
with one’s own values and sense of self
assessing how the problematic behavior affects those in the
individual’s environment

recognizing one’s self-efﬁcacy

profound affective arousal regarding the problematic
behavior

Description
choosing to alter the environment

recognizing and altering the triggers for the behavior
altering the environment in such a way that the trigger is
less likely to occur

being rewarded for making changes

establishing emotionally intimate relationships with people
who can support them

 

153

APPENDIX D

Additional Research Information on TTM

Processes of change

Processes of Change is another central construct of the TTM (represented as PC
in Figure A; Prochaska & DiClemente, 1982, 1983; Prochaska & Prochaska, 1999). The
authors propose ten components which they believe are essential to actually creating and
supporting change in individuals. The Processes of Change construct includes ﬁve
experiential factors (which are the cognitive and affective factors) and ﬁve behavioral
factors. Both of these categories of factors are described in Table C3 of Appendix C.

The authors purport, and research supports, that the cognitive and affective
processes described here are used earlier in treatment during the initial stages of
motivation while the behavioral components are used in later stages of treatment and
change. The cognitive and affective factors need to be established before the behavioral
components can be implemented (Prochaska & DiClemente, 1983; Prochaska &
Prochaska, 1999).

Assessing Processes of Change. No research was found concerning the
measurement tools for the Processes of Change. This lack of research was not surprising
given that these change factors are difﬁcult to operationalize and directly assess (Tierney
& McCabe, 2005). However, if the therapist judges the client to be progressing through
stages, then it is likely that the client is successfully using at least some aspects of the

processes of change.

154

Decisional Balance

The ﬁnal construct of the TTM involves the concept of Decisional Balance
(represented as DB in Figure A; Prochaska & DiClemente, 1982, 1983; Prochaska &
Prochaska, 1999). This construct represents the relationship between individuals’
understanding of the positive and negative consequences of the maladaptive behavior and
their readiness to change. Decisional Balance involves the degree to which an individual
can balance the advantages and disadvantages of continuing the maladaptive behavior.
This construct is used throughout the course of treatment and in each stage of
motivational development (Tierney & McCabe, 2005). However, the ratio of advantages
to disadvantages changes signiﬁcantly as the individual progresses from one stage to the
next

Assessing decisional balance. Decisional Balance construct is assessed through
clinician interview (Prochaska & Prochaska, 1999; Tierney & McCabe, 2005). The
therapist assists the client in developing lists of advantages and disadvantages to
changing their maladaptive behavior in the beginning of treatment. As the client
continues to enhance his motivation, he and the clinician develop lists of the advantages
and disadvantages of challenging himself to continue to change. If the client is
progressing well, the advantages will outweigh the disadvantages of change (Prochaska

& Prochaska, 1999; Tierney & McCabe, 2005).

155

APPENDIX E

Psychopathology and Victim Status

In addition to implications for different levels of motivation to change, different
psychopathology characteristics can also be associated with different offense
characteristics, most notably, victim status. These offense characteristics may provide
additional indices affecting motivation to change. Research comparing sex offenders
against adults and sex offenders against children on various forms of psychopathology
has shown consistent differences in the psychopathology components of these groups
(Ahlmeyer et al., 2003). In a sample of approximately 8000 criminal offenders (695 sex
offenders) who were administered the Millon Clinical Multiaxial Inventory-II, Ahlmeyer
et al. (2003) found that sex offenders against children (n=472) had an affective
psychopathology presentation (i.e., Cluster C personality disorders) while sex offenders
against adults (n=223) had a disruptive behavior psychopathology presentation (i.e.,
Cluster B personality disorders). In their sample, 41% of sex offenders against children
had high scores on the avoidant personality disorder scale (as opposed to 28% of sex
offenders against adults) and 30% had high scores on the dependent personality disorder
scale (as opposed to 17% of sex offenders against adults). Both of these disorders are
Cluster C personality disorders which characterize individuals as being timid, overly
anxious, and/or socially inept (APA, 2000).

On the other hand, compared to sex offenders against children, sex offenders
against adults scored higher on all Cluster B personality disorders (i.e., Histrionic,

Antisocial, Narcissistic, Borderline), though those results were not signiﬁcantly different

156

 

from sex offenders against children (Ahlmeyer et al., 2003). Cluster B personality
disorders typically characterize individuals as impulsive, callous, emotionally labile,
disregarding of others, and/or self-centered (APA, 2000). These individuals also
experience emotions, particularly anger, in a brief, intense yet superﬁcial manner (APA,
2000), which may contribute to their impulsivity and aggression. In the Ahlmeyer et al.
(2003) study, sex offenders against adults did not surpass sex offenders against children
on any of the scales assessing Cluster A, Cluster C, or Axis I disorders (except Mania and
Alcohol Dependence).

Empirical research has also indicated that sex offenders differ on mood disorders.
For example, Ahlmeyer et al. (2003) found that their sample of sex offenders against
children (n=472) had overall higher rates of affective disorders than did sex offenders
against adults. Thirty percent of the sex offenders against children sample were high on
dysthymia; 49% were high in anxiety disorders; and 6% on major depression. For sex
offenders against adults, their highest Axis I score was 44% in anxiety followed by 35%
in alcohol use disorders. Their rates on dysthymia were 18%; their rate of major
depression was 2%. Ahlmeyer et al. (2003) concluded that, in terms of overall
psychopathology, sex offenders against children were the most severely emotionally
disturbed (i.e., internalized) and sex offenders against adults were more similar to non-
sexual offenders who were more behaviorally disturbed (i.e., extemalized). These results

are consistent with the pattern of ﬁndings previously noted.

157

APPENDIX F

The URICA and Sex Offenders

In examining the use of the modiﬁed URICA with 36 incarcerated sex offenders
against children, internal reliability was established for convicted sex offenders against
children who were waiting to begin treatment as was test-retest reliability. Internal
reliability ranged from .81 to .90; test-retest reliability ranged ﬁ'om .69 to .84. The
following table provides the internal and test-retest reliability for each subscale:

Table F1: Reliability of URICA Subscales

 

 

Subscale Internal reliability Test-retest reliability
Pre-Contemplation 0.83 0.69
Contemplation 0.90 0.84
Action 0.81 0.80
Maintenance 0.89 0.70

 

In addition to those who were waiting to begin treatment, internal reliability (but
not test-retest reliability) was also established for 47 incarcerated sex offenders against
children who were already in treatment. These offenders were assessed at pre-treatment,
mid-treatment, and post-treatment. The following table summarizes the internal
reliability for this sample of sex offenders at each assessment period:

Table F2: Internal Reliability of URICA across Time

 

 

Subscale Pre-treatment Mid-treatment Post-treatment
(n=47) (n=3 3) (n=3 3)
Pre-Contemplation 0.80 0.81 0.89
Contemplation 0.84 0.78 0.85
Action 0.85 0.82 0.85
Maintenance 0.81 0.78 0.84

 

The construct validity of this scale was also assessed with both samples and
revealed ﬁndings consistent with theoretical expectations. The Contemplation, Action,

and Maintenance scales were all positively correlated with one another and negatively

158

correlated with Pre-Contemplation. This pattern of ﬁndings was the same in both sex
offenders against children and sex offenders against adults. These ﬁndings suggest that
those in the Pre-Contemplation stage have lower levels of motivation compared to those
in the other three stages.

Finally, this study also considered the inﬂuence of social desirability on the
participants’ responses and, overall, found that it was not a factor in their answers. Some
variation in socially desirable responding was evident but the authors concluded that
these variations were not signiﬁcant enough to warrant concern. An important ﬁnding
from the Tierney and McCabe (2004) study is that the majority of the offenders were
involved with the Pre-Contemplation process even though they had volunteered to
participate in treatment. They surmised that this ﬁnding represents external motivations
to attend treatment (e. g., more privileges in the housing units in prison), which is a

common assumption in the sex offender literature.

159

APPENDIX G

Intervention Strategies Used in CBT with Sex Offenders

Behavior therapy. The ultimate goal of behavior therapy with sex offenders is to
bring under control maladaptive sexual urges and behaviors. This intervention strategy
arises from several theories suggesting that sex offenders cannot control their sexual
impulses and use sexual gratiﬁcation to fulﬁll a variety of emotional and psychological
needs (Ward & Beech, 2006). This intervention ironically does not directly address the
“maladaptive sexual behaviors” of rape and assault; instead, the focus is on behaviors that
may increase the risk of rape or assault of an individual, such as downloading
inappropriate sexual material on the internet and using the material to masturbate.
Behavioral interventions addressing these impulses vary and include masturbatory
reconditioning techniques, electrical aversion, and covert sensitization, to name only a
few (Abel et al., 1992). Retraining sex offenders to more appropriately feel and satisfy
their sexual urges promotes responsible sexuality and reduces one of the driving forces
behind sexual aggression.

Prosocial skills training. The goal of prosocial skills training is to increase those
behaviors which may help sex offenders to better function in society and create a sense of
self-efﬁcacy. Sex offenders often have difﬁculty initiating and continuing conversations,
expressing their emotions in a healthy manner, and engage effectively with others (Abel
et al., 1992). These deﬁcits decrease their social bonds with others and increase the risk
for offending as they search for a means to achieve intimacy and fulﬁll their emotional

needs (Ward & Beech, 2006). These techniques typically include social skills training,

160

which involves teaching offenders, who typically lack proper social skills, general
strategies to initiate and maintain conversations and positive interactions with others
(mainly those of the opposite sex?). Also included in prosocial skills training is
assertiveness training, which involves developing the skills to successfully express both
positive and negative emotions and thoughts (Abel et al., 1992), potentially reducing the
frustration that often precedes sexual offending (Ward & Beech, 2006). Finally in this
training, sex offenders, who typically are naive to matters of the human anatomy, engage
in sex education/sexual dysfunction treatment which teaches them the basic knowledge of
sexual health and the process to obtaining healthy sexual relationships (Abel et al., 1992).
Together, these elements ready the sex offender for release into the community with new
ways of interacting with others and new knowledge of the appropriateness of their
thoughts and behaviors, increasing the likelihood of positive interactions with others and
decreasing the need to offend.

Restructuring distorted cognitions and developing victim empathy. Cognitive
distortions (e. g., sex with children is not harmful; women always play hard to get) are
often used as a justiﬁcation for sex offenders’ actions and help to neutralize their negative
affect after committing a sex crime (Hanson, Cox, & Woszczyna, 1991). Because these
cognitions perpetuate their offense behaviors, the goal of cognitive restructuring is to
identify and refrarne these distorted cognitions while creating more adaptive belief
systems (Abel et al., 1992). Clinicians generally hold that using CBT skills to modify
these distorted cognitions will result in decreased recidivism as it allows for their sexual
offenses to become more ego-dystonic. Similarly, the development of victim empathy is

viewed as a method for reducing recidivism as it makes offenders more aware of the

161

consequences of their actions, which they often deny or minimize. Intervention strategies
for this aspect of treatment include sex offenders watching videotapes of sexual assault
victims talk about their trauma and/or sex offenders writing the details of the assault ﬁ'om
the perspective of their victim (Abel et al., 1992).

Relapse prevention. This ﬁnal aspect of the treatment is one of the most
important as it teaches offenders how to anticipate and cope with potential relapses. The
speciﬁc components of relapse prevention are extensive and thus are only brieﬂy
described here. In brief, this technique includes a variety of psychoeducational (e.g.,
linking emotions to behavior; sex education), behavioral (e. g., avoidance strategies;
programmed coping mechanisms), and cognitive (e.g., identify and analyze maladaptive
thoughts) elements that teach the offenders how to recognize triggers of a relapse (e. g. an
affect shift) and intervene on their own behalf (e. g., via effective coping mechanisms).

Mastery of this step is one of the key elements of reducing recidivism (Abel et al., 1992).

162

APPENDIX H

Motivational Interviewing

Motivational interviewing parallels the TTM described earlier. Both approaches
assume that clients have varying levels of motivation to change when they begin
treatment and this motivation continues to oscillate throughout the treatment process
(Arkowitz & Miller, 2008). Further, both approaches recognize and accept ambivalence
as a central part of the change process that needs to be addressed throughout treatment in
order to continue the client’s progression (Arkowitz & Miller, 2008).

Motivational interviewing is composed of four basic principles or assumptions.
The ﬁrst principle is expressing empathy for the client (Arkowitz & Miller, 2008). For
the therapist, this would involve trying to understand the client’s point of view and the
circumstances which may be inﬂuencing the client’s behaviors. While it may seem
obvious that a therapist needs to have empathy for a client, expressing empathy may be
difﬁcult to do when therapists are working with sex offenders (Mann, 1996 as cited in
Ginsburg et al., 2002). When working with offender populations it can sometimes be
difﬁcult to ﬁnd a balance between expressing empathy for the client and condoning their
actions, especially when the client is pulling for the therapist to condone their actions.
Thus, therapists may err in favor of being less empathic so as to not convey the wrong
message to the offender or they may have a bias against the sex offender because of their
offense, making empathy building difﬁcult (Farbring & Johnson, 2002). However, using
empathy in motivational interviewing does not involve condoning the clients’ offenses

but instead assumes that using empathy in a nonjudgrnental manner creates a safe

163

environment in which the client can begin thinking critically about his behavior
(Arkowitz & Miller, 2008; Mann, 1996 as cited in Ginsburg et al., 2002).

The second assumption of motivational interviewing is that motivation can be
inﬂuenced through creating cognitive dissonance between the client’s values and
behaviors (Arkowitz & Miller, 2008). As the therapist gathers more information about
the client, she is able to generate discrepancies between the clients’ values (e.g., hurting
children is wrong) and their behaviors (e. g., sexually assaulting a child). She can then
reﬂect those discrepancies back to the client in an empathic, nonjudgrnental manner. It is
assumed in motivational interviewing that the cognitive dissonance created in these
situations will motivate the client to want to align his behaviors with his values and begin
to change (Arkowitz & Miller, 2008).

The third principle of motivational interviewing involves the therapist tolerating
resistance and ambivalence about change (Arkowitz & Miller, 2008). The assumption of
motivational interviewing is that ambivalence represents the client vacillating between
changing and not changing. Thus, if the therapist has even a small indication that the
client may be interested in change then she can use other therapeutic techniques (e.g.,
worksheets on advantages and disadvantages of change) to engage the client in
conversations about change. The ﬁnal principle of motivational interviewing is
encouraging self-efﬁcacy (Arkowitz & Miller, 2008). Encouraging self-efﬁcacy is
especially important when considering the culture of a criminal, which oﬁen involves
being constantly reminded on their poor decision-making as well as being told what they
should and should not do (Mann, 1996 as cited in Ginsburg et al., 2002). These

controlled boundaries were implemented in order to keep the public safe and perhaps

164

unintentionally communicates to the offender that he has no control in his life anymore.
This feeling can negatively affect clients’ self—efﬁcacy and self-esteem, as they may
believe they are not able to take care of themselves or change their own behaviors (Mann,
1996 as cited in Ginsburg et al., 2002).

One advantage of these approaches and motivational interviewing more generally
is that it is an individual, client-centered approach. This is an advantage because group
therapy is often the format chosen for sex offenders (Barker & Beech, 1993). In group
therapy, sex offenders may not feel as open to discussing their thoughts and concerns
about change because of pressure from other offenders to maintain a tough exterior
(Ginsburg et al., 2002). Also, sex offenders have a tendency to deny their offenses as a
way of reducing the cognitive dissonance between their values and behaviors (Mann,
1996 as cited in Ginsburg et al., 2002). This type of denial interferes with the
development of motivation to change but cannot be addressed in much depth during the
typical group therapy process because of the need to not individualize group therapy
(Mann, 1996 as cited in Ginsburg et al., 2002). Thus having an individual therapy
approach that can be used to help the offender focus only on motivation to change may be
beneﬁcial to his ability to engage in other forms of therapy.

Another advantage of motivational interviewing is that it may be used as its own,
individual treatment or be used in conjunction with other treatment approach. Meta-
analytic results of non-criminal populations have found that using motivational
interviewing as a pre—treatrnent approach has led to larger effect sizes in treatment
responses and longer-lasting change results (Arkowitz & Miller, 2008). Though this

study was not conducted in sex offenders speciﬁcally, it could be the case that treatment

165

responses would be enhanced if motivational interviewing was combined with cognitive-
behavioral therapy, which is the preferred method of treatment for sex offenders currently
(Barker & Beech, 1993). The typical cognitive behavioral treatment program with sex
offenders begins with directly challenging the sex offenders’ denial of and justiﬁcation
for the offense (Beech & Fisher, 2002), which may cause some offenders to become
defensive and respond negatively to the approach (Kear-Cowell & Pollack, 1997). They
may actually disengage from treatment and, if they continue in the program, may only do
so because of the negative consequences they may face if they discontinue treatment
(Ginsburg et al., 2002). Having a pre-treatment where motivation is enhanced is
postulated to improve the sex offenders’ understanding of themselves and the treatment
process and makes them more receptive to the challenges associated with change (Mann,

1996 as cited in Ginsburg et al., 2002).

166

APPENDIX I

Sampling Procedures

Ten ﬁles from each of the years proposed in for analysis (i.e., 2000-2005) were
randomly selected for a total of 60 ﬁles explored. To ensure randomness, a storage box
containing the ﬁles was randomly chosen from among the piles of boxes and 5 sex
offender ﬁles were chosen randomly from that box. Each ﬁle was examined and
categorized as usable or unusable. Usable ﬁles were those which contained all of the
measures needed for the study and may or may not have had items missing on those
measures. Unusable ﬁles were those which had at least one measure completely missing
from the ﬁle. In addition to the actual measures that were needed for the study (e. g.,
URICA), additional information on the type of crime committed (and a description of the
crime if available) and victim status was also collected. This information was included
into the decision of determining a ﬁle as usable or not (i.e., if the type of crime or victim
status information was missing, then the ﬁle was counted as missing, regardless of if the
remaining measures were available).

The following table summarizes the results from this data sampling:

Table 11: Summary of Usable Files

 

 

Usable ﬁles

Utility rate (deﬁned as all data present) 50% N=3 52
Contact oﬂenders 75% (of 352) 264
Adult victims 25% (of contact offenders) 66
Child victims 75% (of contact offenders) 198
Non-contact oﬂenders 25% (of 352) 88

 

167

Table 12: Utility Rate of Files Per Year

 

 

Utility rate per year
(i.e., % of ﬁles per year with no missing data)

2000 20% N=20
2001 40% 42

2002 60% 74

2003 40% 48

2004 70% 92

2005 70% 83

 

Table I3: Estimates of Missing Data
Degree of missing data projected N’s
Missing 1 measure 33% N=116
Missing 2 measures 37% 130
Missing 3 measures 27% 95
Completely missing 3% l l

 

 

 

Overall, if the results from this sampling are extrapolated to the entire 703 sex
offenders enrolled in the program from 2000-2005, then the 50% utility rate would yield
approximately 352 offenders with complete ﬁles (albeit, some may have data missing at
an item level). If only contact offenders are used in this study, then the sample size
would be 264. If those missing only the URICA were included, then the total N would
raise. It is not known how many URICA’s are missing at this time; thus, the exact
increase in N if the URICA’s are imputed cannot be calculated at present.

With regard to the representativeness of the complete data ﬁles relative to the rest
of the ﬁles, the following table of correlations indicates that missingness does not form
patterns in the data based on offense characteristics or motivation to change scores: none

of these correlations are signiﬁcant.

168

Table I4: Correlations of Missing Data

 

 

 

Missingrress
(N=55) Supervision status Offense type Victim status
URICA -.058 .057 -.037
MMPI -.025 .030 -.162
Attendance .012 -.179 .049
Completion -.058 -.142 -.275
Personal history .017 -.237 .063

 

Total sample size. At the time of this study’s proposal approval, there are 250
offenders in the dataset who meet the criteria for the study (i.e., contact offenders with
complete data). These data were obtained through a similar procedure used to obtain the
sample data just described. First, the MMPI-2 data were obtained for the offenders
enrolled in the treatment facility from 2000-2005. The client numbers associated with
these data were then matched to the client ﬁles at the treatment facility which housed the
rest of the data. If an offender whose MMPI-2 data are complete also was a contact
offender, completed the URICA, and had the other offense characteristics data in their
ﬁles, their client number was given a checkrnark on a computer print out of the client
numbers. This process was continued until 250 offenders had checkmarks by their
names. There are still more ﬁles to go through; thus, this 250 will increase by about 15
or 20 offenders for the ﬁnal analyses. However, due to time constraints, all of the

available ﬁles could not be explored.

169

 

APPENDIX J

Description of MMPI Scales Used in Present Study

Utility of MMPI-2 in present study

The two MMPI-2 components of interest for this study were the validity and
clinical scales. The research regarding each is brieﬂy described next. Included with the
description of each scale and the rationale for its use in the study is the rate of occurrence
of clinically signiﬁcant elevations of that scale (i.e. t-scores greater than 65; Megargee,
2006) as well as its correlation with content scale TRT. Content scale TRT is a measure
of negative attitudes towards treatment and difﬁculty changing behavior patterns l
(Butcher & Williams, 2000). Therefore, the higher this t-score, the more difﬁculty these
individuals have with accepting help and making difﬁcult changes in their lives. It is a
scale which appears to parallel the motivation to change construct explored in this study.
Including the correlations with this scale will assist in strengthening the rationale for
including each scale in the analyses.
MMPI-2 validity scales

With sex offender populations, the MMPI-2’s validity scales have shown
convergent validity with other measures of minimization, denial, exaggeration, and
cognitive distortions such as the Multiphasic Sex Inventory (Haywood, Grossman,
Kravitz, & Wasyliw, 1994; Schlank, 1995). These results suggest that the MMPI-2
validity scales had valid use in the current sample of sex offenders, assuming that their
validity scales did not indicate an invalid proﬁle, in which case the ﬁle would have to be

removed. (Details regarding how validity was determined are provided in the Data Entry

170

section below). Of the eight possible validity scales, only three scales were consistently
present in all of the archival data collected for the study—Scales F, L, and K. Of these,
Scales F and K were included in the cluster analyses, while Scale L was used to help
determine the validity of the individual’s MMPI proﬁles. Scale F was used to detect
negative impression management (i.e., malingering or exaggeration of symptoms;
Megargee, 2006). Scale K was used to detect self-deception, which is a type of positive
impression management (i.e., unconscious and protective minimization of symptoms;
Megargee, 2006). Research suggests that sex offenders may engage in denial and
exaggeration of symptoms to fulﬁll self-serving intentions (Lanyon, 2001), which may
make their psychological presentations difﬁcult to interpret.

Sca_le_F_, Scale F was chosen as the exaggeration measure for this study because it
appeared to be the most relevant validity scale which could assess externalizing behaviors
such as the exaggeration of symptoms (Butcher & Williams, 2000). It also was one of
the validity scales which clustered in previous research (e. g., Duthie & McIvor, 1990;
Kalichman etal., 1989, 1992). In criminal samples, 49% had t-score elevations from 65-
80 on scale F; 46% had t-score elevations above 80 (Megargee, 2006). Symptom
exaggeration could be seen as a form of manipulation or self-serving behavior (Lanyon,
2001). Research supports the notion that scale F may be an externalizing variable and
has shown that, in college students, scale F positively correlated with scales Pd, Ma, and
ANG of the MMPI-2 (O’Laughlin & Schill, 1994). In criminal samples, scale F
correlates positively with TRT in the .80 range (Megargee, 2006). Therefore, as

symptom exaggeration increases, negative attitudes about treatment and lack of

171

motivation to change also increase. This correlation suggests that motivation to change
may be less genuine in offenders with elevations on this scale.

Span—e1; Scale K detects positive impression management that is more defensive
than instrumental (Megargee, 2006). Part of its original purpose was to detect the
suppression of psychological problems in those who had genuine psychological problems
(Butcher & Williams, 2000). In criminal samples, 20% had t-score elevations between
65 and 80 for scale K, placing them in clinically signiﬁcant yet interpretable ranges
(Megargee, 2006). Less than 1% of these criminals had t-score elevations above 80.
This scale correlates negatively with TRT in the .70 range (Megargee, 2006). Therefore,
as symptom suppression increases, negative attitudes toward treatment decrease
presumably because the offender does not think that he has many problems to change
since he is suppressing his symptoms. While it may appear that this would result in
higher levels of motivation to change, high scores on scale K may actually manifest as
higher Pre-Contemplation scores, which assesses the self-perceived need to change. The
type of minimization that scale K detects may signiﬁcantly inﬂuence a participant’s
actual desire to change his behavior, which is why it seemed appropriate to include in the
present study.

Excluded validity scales. Scale L was excluded from the cluster analyses because
it was not as consistently elevated in the previously reviewed MMPI clustering studies as
Scales F and K. Scale L was used in determining proﬁle validity as it is a measure of
positive impression management (Butcher & Williams, 2000; Megargee, 2006). It also
can reﬂect a nay-saying response set on the questionnaire that is not associated with

intentional deception. Examining this score in conjunction with other validity scores

172

 

(i.e., TRIN; see Table C1 of Appendix C) helps to deternrine which tendency an elevated
L is reﬂecting: positive impression management or non-acquiescence. Unfortunately, the
TRIN score was not available for all offenders in this sample; thus, the L score had to be
interpreted as a single scale using that guide that t—scores greater than or equal to 80 are
not interpretable in correctional samples (Megargee, 2006).
MMPI-2 clinical scales

The 10 clinical scales of the MMPI-2 are each conceptually associated with a
speciﬁc domain of psychopathology (Butcher & Williams, 2000). These scales are used
to determine the proﬁle and level of psychopathology an individual is experiencing. In
the present study, the clinical scales of interest were D, Pd, Mf, Pa, Pt, Sc, and Ma.

m This scale (a measure of depressive symptoms; Butcher & Williams,
2000) was included in accordance with the literature suggesting that depression rates in
sex offenders (35%; Stinson et al., 2005) are higher than in the general population (6.7%;
Kessler et al., 2005). Scale D has been shown to positively correlate with other
internalizing measures (e. g., Beck Depression Inventory) in a sample of college students
(Ben-Porath et al., 1993) and is a common feature of criminal offenders (Megargee,
Mercer & Carbonell, 1999). In fact, scales Pd and D are the second most commonly
elevated pair of scales in criminal samples (Megargee et al., 1999). For scale D alone,
20% of criminal offenders show t-score elevations greater than or equal to 65 (Megargee,
2006). This scale also correlates in the .60 range with content scale TRT (Megargee,
2006). The pessimistic cognitive style often associated with depression (Beck, 1995)
may explain the relatively high correlation between scales D and TRT as these

individuals may think they will not succeed in treatment or beneﬁt from it.

173

Scale Pd. Scale Pd is an assessment of antisocial characteristics and behaviors
(Megargee, 2006). The most commonly elevated pair of MMPI-2 subscales in criminal
populations is scales Pd and Ma, though it only occurs in 13% of criminal populations
(Megargee et al., 1999). For scale Pd alone, 35% of male criminals have elevations at or
above a t-score of 65 (Megargee, 2006). This scale was chosen for this study not only
because it captures antisocial behaviors but also because of its consistent presence in both
sexual and nonsexual offender samples (Butcher & Williams, 2000). It also is positively
correlated (in .50 range) with content scale TRT (Megargee, 2006). In the present study,
individuals with higher t-scores on scale Pd may have lower levels of motivation to
change, consistent with this correlation analysis.

Scale Mf. This scale was originally developed to detect homosexual tendencies in
individuals (Megargee, 2000). As such, it is coded differently in males and females with
high scores in males indicating more “feminine” tendencies and high scores in females
indicating more “masculine” tendencies. In male criminal samples, only 3% of them had
t-score elevations at or above 65 on this scale, indicating that only a few of them have
“feminine” tendencies. Instead, 40% of the male criminal sample has t-scores below 40
(Megargee, 2006): For males, t-scores below 40 on this scale are indicative of aggressive
behavior, callousness, and stereotypical masculine interests. Duthie and McIvor (1991)
and Kalichman et al. (1992) both found clusters of sex offenders in their studies with
elevations on Scale Mf, indicating low levels of masculine behaviors. In criminal
samples, Mf did not correlate (positively or negatively) with scale TRT or any other of
the MMPI-2 scales (Megargee, 2006). However, one could hypothesize that sex

offenders would have low t-scores on this measure (i.e., more masculine traits) if they are

174

 

sexually abusing women and thus asserting a masculine power over the women. They
also could have scores within normal ranges as in one of the MMPI clustering studies
(Kalichman et a1, 1989). As for sex offenders against children, they may have elevations
on this scale (i.e., low masculine characteristics) consistent with previous studies (Duthie
& McIvor, 1991; Kalichman et al., 1992) or scores within normal ranges as with Shealy
et al. (1990).

Scale Pa. Scale Pa includes content which assesses not only distrust of others but
also more disturbed thought patterns and beliefs (Megargee, 2006). This scale has been
found to be one of the most commonly elevated scales in criminal samples (Megargee et
al., 1999), with 25% of criminal offenders having elevations at or above a t-score of 65
(Megargee, 2006). This scale has correlations in the .70 range with scale TRT
(Megargee, 2006), indicating that some feature of this trait (e.g., perhaps a mistrust of the
therapist) is interfering with these individuals’ desire or willingness to change their
behavior.

Scale Pt. This scale is a measure of intense anxiety about a variety of situations
and circumstances (Butcher & Williams, 2000). Quayle et al. (2006) has found that sex
offenders tend to experience higher levels of anxiety, which is consistent with ﬁndings
from other studies (e.g., Stinson et al., 2005). Because of these ﬁndings, scale Pt was
included, as it is a measure of persistent generalized anxiety. In male criminal samples,
21% had t-score elevations greater than or equal to 65 (Megargee, 2006). This scale
correlates in the .70 range with scale TRT (Megargee, 2006), suggesting that anxiety may

be a hindrance to engaging in treatment.

175

 

Scale Sc. Scale Sc is a measure of disturbed thought processes, social isolation,
and poor judgment (Butcher & Williams, 2000). Early research with the original MMPI
found that scales Pa and Sc were frequently elevated in sex offenders against adults
(Erikson et al., 1987). However, the research seeking to use this scale to distinguish sex
offenders ﬁom nonsexual offenders has been inconsistent at best (Kalichman et al.,
1989). It is still possible that some sex offenders have social isolation and disturbed
thought processes. Indeed, the rate for thought disorders is four times as high in sex
offenders (4%) as in nonsexual offenders (1%; Ahlmeyer et al., 2003). In criminal
offenders, 23% had t-score elevations above 65; this scale was also correlated with scale
TRT in the .70 range (Megargee, 2006). Though there is some doubt about its
applicability in the current study due to the mixed sample of sex offenders against adults
and sex offenders against children, because of the potential severity of this scale in this
population, it was included in the analyses.

Scale Ma. Scale Ma is a measure of hypomania symptoms (e.g., impulsivity;
Butcher & Williams, 2000). Thirteen percent of criminal offenders have t—score
elevations at or above 65 on this scale (Megargee, 2006). It is most commonly elevated
with scale Pd and the two of them together are the most commonly occurring two-point
MMPI-2 elevation in criminal offenders (Megargee etal., 1999) and second most
common two-point elevation in sex offenders with nonsexual criminal histories (Erickson
et al., 1987). It is theorized that elevations on scales Pd and Ma together are indicative of
sex offenders being impulsive and risk-takers (Butcher & Williams, 2000). These
personality qualities may place them in situations where they are more likely to commit

crimes. Surprisingly, this scale was not correlated above .50 with scale TRT (Megargee,

176

2006). Given the notion that the individuals with higher t-scores on Scale Ma enjoy the
impulsivity and stimulation associated with hypomania symptoms (e. g., Butcher &
Williams, 2000), one would expect that they would be resistant to change. The low
correlation with Scale TRT suggests that, in the present study, people with higher scale
Ma t-scores may have a pattern of motivation signifying indifference to change.

Excluded clinical scales. Scales Hs, Hy, and Si were not included in the analyses
for this study. The central reason for their exclusion is that these variables did not
consistently emerge as signiﬁcant in previous clustering studies, thus, indicating that
these variables may not be key domains in sex offenders’ pathologies.

The following chart summarizes the use of the MMPI-2 scales in the cluster

 

 

analyses:
MMPI-2 subscales
Validity scales F, K
Excluded validity scales L
Clinical scales D, Pd, Mf, Pa, Pt, Sc, Ma

Excluded clinical scales Hs, Hy, Si

177

 

APPENDIX K

Cluster Analysis: Background

Hierarchical Cluster Analysis

Hierarchical procedures involve creating clusters based on the similarities and/or
differences between individual cases (Hair & Black, 2000). There are two types of
hierarchical procedures: divisive and agglomerative. In the divisive procedure, the whole
sample is consider one cluster and then each individual case is separated from the cluster
to form smaller clusters based on the high within-group similarity and low between-group
differences. The agglomerative approach works in the opposite direction. Here, each
case is considered a cluster and these clusters are grouped with each other based on
within-group and between-group means. With both divisive and agglomerative
approaches, clusters are nested rather than mutually exclusive groups.

In the hierarchical method, several approaches could be used to determine how
the cases should be “linked” with one another as the clusters are formed (Rapkin & Luke,
1993). These approaches include single linkage, complete linkage, average linkage, and
Ward’s method. The ﬁrst three linkage methods focus on linking the cases based on their
similarity of one case to another case in a particular cluster. The similarity can be
deﬁned as the single case being most similar to a case in a cluster (i.e., single linkage),
least similar to a case in a cluster (i.e., complete linkage), or the average similarity being
the single case and all of the other cases in the cluster (i.e., average linkage; Rapkin &
Luke, 1993). In the ﬁnal method, Ward’s method, the focus in assigning cases to a

cluster is on maximizing between- group heterogeneity and minimizing within-group

178

heterogeneity rather than on matching a single case to the overall cluster. Therefore, the
resulting clusters are more balanced and better answer the research question (Rapkin &
Luke, 1993). Since Ward’s method will result in the most homogenous clusters (within-
group), this method was used in the present study to determine proximity in the clusters.

The hierarchical method has both advantages and disadvantages for use in this
study. One suggested advantage of the hierarchical approach is that each individual case
is considered a cluster and each of these clusters is grouped based on its individual
characteristics (Hair & Black, 2000). This more person-centered approach more closely
ﬁts the general beneﬁt of cluster analysis as being an exploration of natural relationships.
One potential drawback to the hierarchical approach is that it is sensitive to the presence
of outliers and inappropriately-chosen clustering variables (i.e., irrelevant clustering
variables) in the data, which could lead to the false creation of clusters (Hair & Black,
2000). Because of this, it may be necessary to run the analyses several times and, each
time, delete undesirable cases which may be offsetting the results. Because of its
strengths and weaknesses it may be necessary to use another type of approach in order to
obtain the best results.
Iterative Cluster Analysis

The iterative approach involves assigning cases to a particular cluster based on the
researcher’s predetermined speciﬁcation of the number of clusters and the “cluster seed”
(i.e., the mean of all cluster variables for each individual randomly assigned in the ﬁrst
clustering attempt) expected in the sample (Hair & Black, 2000; Luke, 2005; Rapkin &
Luke, 1993). The iterative approach then involves matching the case’s mean values on

the clustering variables to the most similar cluster seed until the best clusters are created

179

 

based on the number of pre-speciﬁed clusters. Any of the four above linkage methods
can be employed here in order to assign cases to the appropriate clusters. The added
advantage of this approach is that it allows cases to be reassigned to clusters (Rapkin &
Luke, 1993). The cluster seeds are recalculated when new cases are added, after which
cases may be reassigned to different clusters if their means better match the new mean of
another cluster. The result is a set of more homogeneous (within-group) clusters (Rapkin
& Luke, 1993). While the advantage to this approach is that it allows cases to be
reassigned to other clusters, the disadvantage of this approach is that the researcher has to
pre-select the clusters (Hair & Black, 2000; Rapkin & Luke, 1993), which can be difﬁcult
if little is known theoretically about how the sample would cluster, as in the present
study.
Combined Approach

It is possible to use both types of cluster analyses in a given study (Hair & Black,
2000). The hierarchical procedure is used ﬁrst to establish the appropriate number of
clusters for the data, the exact cluster seeds for each of those clusters, and identify the
outliers. The iterative approach is then used to assign each case to the appropriate cluster
based on its means on the clustering variables. The beneﬁt of this approach is that it
combines the advantages of both approaches and the advantages of one approach
counters the disadvantages of the other approach (Hair & Black, 2000). For example, the
major disadvantage of the iterative approach is that the researchers must carefully choose
the approach for determining the clustering seed, and one disadvantage of the hierarchical
approach is that the clusters are nested in each other rather than mutually exclusive.

When these approaches are combined, the hierarchical approach creates the cluster seeds

180

for the researcher based on the best ﬁt of the data (which addresses the disadvantage of
the iterative approach) and the iterative approach takes the cluster seed information to
create mutually exclusive clusters which ﬁt the data the best (which addresses the

disadvantage of the hierarchical approach).

181

 

APPENDIX L

Post-Hoc Analyses for Predictors of Motivation to Change

A ﬁnal set of regression analyses was run to identify which factors in this study
predicted motivation to change. Four regression analyses were run. The ﬁrst regression
included the individual MMPI clustering variables (rather than the cluster membership
categories) predicting motivation to change. Results indicated that scales K and Mf were
signiﬁcant predictors of motivation to change but none of the other MMPI scales were
(see Table J1). The model accounted for 11% of the variance in motivation to change.
These results are particularly interesting considering that these are the two variables
which did not seem to contribute to the structure of the clusters and were considered for
removal from the analyses. The beta for scale K was negative, indicating that as
symptom suppression decreases, motivation to change increases. The directionality of
this scale is appropriate and suggests that as offenders begin to become more familiar
with their psychological problems and their psychological defenses are beginning to
dissolve they become more willing to work on their problems. The beta for scale Mf

23 motivation to

was positive, indicating that as stereotypical feminine tendencies increase,
change increases. Males with high scores on Mf tend to be more emotionally sensitive
(Megargee, 2006). This sensitivity may help the offender to be more emotionally attune
to their own feelings of guilt and shame for the crimes that they have committed. Guilt

and shame have been hypothesized to be important factors in helping clients make and

sustain behavioral change in their lives (Tierney & McCabe, 2005).

 

23 Scale Mf is scored such that higher scores indicate more stereotypical feminine behaviors in males.
Lower scores, on the other hand, indicate more stereotypical masculine behaviors in males (Butcher &
Williams, 2000).

182

 

The second regression run used the offense characteristics as predictors of
motivation to change. Results indicated that victim’s age and supervision status were
both signiﬁcant predictors of motivation to change (see Table 12). This model accounted
for 35% of the variance in motivation to change. The beta for victim’s age was negative,
indicating that as the victim’s age decreases, motivation to change increases. This
ﬁnding may indicate that offenders with child victims have a higher sense of guilty for
their actions and are more motivated to change. It is also possible that offenders with
child victims feel societal pressures to stop offending. Thus they may self-report higher
levels of motivation to change but may not have an actual genuine interest in changing.
Supervision status was also a signiﬁcant predictor of motivation to change. The beta was
positive. Since supervision status is dichotomized (i.e., probation or parole), this result
indicates that those offenders who are on parole are more motivated to change. This
ﬁnding is interesting because the offenders on parole constitute a minority in the dataset
(N =84) but they have spent a signiﬁcantly longer amount of time in jail or prison
(mean=72.83 months) than the probationers (mean=7.1 1 months; F (l, 219) =24l .41;
p<.001). Since many of these offenders begin this treatment program shortly after they
have been released from prison, it is possible that they are reporting being motivated to
change because they have a new appreciation for their freedom and not necessarily
because they want to follow societal norms.

The third cluster tested which clinical variables were predictors of motivation to
change (see Table J3). Results indicated that only the post-denial score was a signiﬁcant
predictor of motivation to change. This model accounted for 21% of the variance in

motivation to change. The beta for denial was negative, indicating that as denial

183

decreases, motivation not change increases. Tierney and McCabe (2005) proposed that
sex offenders’ use of cognitive distortions to justify their behaviors may keep them from
developing high levels of motivation to change because they do not necessarily view their
behaviors as negative. This regression result offers some support for this notion.

The ﬁnal regression analysis combined the statistically signiﬁcant MMPI scales
(i.e., scales K and Mt), offense characteristics (i.e., victim’s age and supervision status),
and clinical variables (i.e., post-denial score) into one regression predicting motivation to
change. These variables were combined to determine which types of variables accounted
for the most variance in motivation to change. Therefore the variables were entered as
blocks with scales K and Mf in the ﬁrst block; supervision status and victim age in the
second block; and post-denial score in the third block. Results indicated that each
variable maintained its signiﬁcance in the ﬁnal model, which accounted for 40% of the
variance in motivation to change (see Table 14). The results also indicated that the model
gained the most variance when the offense characteristics were added to the model
(27%). The MMPI scales together accounted for approximately 8% of the variance and
the post-denial score accounted for approximately 6%. These ﬁve variables seem to be
the most robust in predicting motivation to change in this sample of sex offenders.
However, the results are mixed regarding whether these characteristics predict genuine

motivation to change or are reﬂective of underlying ulterior motives of the clients.

184

Table Ll. Regression for Psychopathology Predicting Motivation to Change.

 

 

 

Variable B SE B B
MMPI F 0.02 0.01 0.12
MMPI K -0.04 0.02 -0.14*
MMPI D 0.01 0.02 0.02
MMPI Pd 0.03 0.02 0.12
MMPI Mf 0.06 0.02 0.21 **
MMPI Pa -0.02 0.02 -0.0
MMPI Pt 0.01 0.02 0.02
MMPI Ma -0.02 0.02 -0.09
Note. R2 = 0.11.
*p<.05.
**p<.01.

Table L2. Regression for Offense Characteristics Predicting Motivation to Change.

 

 

 

Variable B SE B B

Offense Type -0. 14 0.18 -0.06
Penetration —0. 1 8 0.36 -0.03
Alcohol at Offense 0.69 0.39 0.13
Drugs at Offense 0.17 0.43 0.03
Victim Status 0.35 0.62 0.05
Victim Gender -0.50 0.03 -0.05
Victim Age -0.07 0.03 -0.19*
Relationship to Victim 0.09 0.11 0.06
Supervision Status 2.43 0.57 0.46***
Prison Time 0.00 0.01 -0.01
Prior Adult Offenses 0.48 0.37 0.09
Juvenile Record -0.23 0.43 -0.04

Note. R2 = 0.35.

*p<.05.
***p<.001.

Table L3. Regression for Clinical Variables Predicting Motivation to Change.

 

 

 

Variable B SE B B
Attendance Record -0.02 0.03 -0.05
Requirements Complete -1 .62 2.55 -0.26
Pre-Treatrnent Score 0.03 0.14 0.04
Post-Treatment Score 0.16 0.16 0.46
Clinical Change 0.01 0.16 0.01
Pre-Denial Score -0.28 0.34 -0.17
Post-Denial Score -0.73 0.37 -0.54*
Denial Change 0.01 0.38 0.01

Note. R2 = 0.21.
*p<.05.

185

Table L4. Modiﬁed Regression for Psychopathology, Offense Characteristics, and

Clinical Variables Predicting Motivation to Change.

 

 

 

Variable B SE B B
MMPI K -0.04 0.02 -0.15*
MMPI Mf 0.05 0.02 O.16**
Victim Age -0.06 0.02 -0.15*
Supervision Status 2.18 0.32 0.42***
Post-Denial Score -0.33 0.08 -0.25***
Note. R2 = 0.40.
*p<.05.
**p<.01.
***p<.001.

186

REFERENCES

Adams, D. B. (2002). Summary of state sex offender registries, 2001. Bureau of Justice
Statistics Fact Sheet Retrieved April 14, 2006.

Ahlmeyer, S., Kleinsasser, D., Stoner, J ., & Retzlaff, P. (2003). Psychopathology of
incarcerated sex offenders. Journal of Personality Disorders, 1 7(4), 306-318.

Ahrens, C. E., Dean, K., Rozee, P.D., & McKenzie, M. (2008). Understanding and
preventing rape. In Psychology of Women: A Handbook of Issues and Theories
(2” Ed. ). F. L. Denmark & M. A. Paludi (Eds). Westport, CT: Praeger
Publishers/Greenwood Publication Group, pp. 509-554.

American Psychiatric Association. (2000). Diagnostic and Statistical Manual of Mental
Disorders, Fourth Edition, Text Revision. Washington, DC: American
Psychiatric Association.

Anderson, W. P. & Kunce, J. T. (1979). Sex offenders: Three personality types. Journal
of Clinical Psychology, 35, 671-676.

Arkowitz, H., Westra, H. A., Miller, W. R., & Rollnick, S. (2008). Motivational
Interviewing in the Treatment of Psychological Problems. New York, NY: The
Guilford Press.

Bard, L. A., Carter, D. L., Cerce, D. D., Knight, R. A., & et al. (1987). A descriptive
study of rapists and child molesters: Developmental, clinical, and criminal
characteristics. Behavioral Science & the Law, 5(2), 203-220.

Baer, R. A., & Miller, J. (2002). Underreporting of psychopathology on the MMPI-2: A
meta-analytic review. Psychological Assessment, 14(1), 16-26.

Barbaree, H. E., & Marshall, W. L. (1991). The role of male sexual arousal in rape: Six
models. Journal of Consulting and Clinical Psychology, 59, 621-630.

Barrett, M., Wilson, R. J ., & Long, C. (2003). Measuring motivation to change in sexual
offenders ﬁom institutional intake to community treatment. Sexual Abuse: A
Journal of Research and Treatment, 15(4), 269-283.

Bassarath, L. (2001). Neuroimaging studies of antisocial behavior. Canadian Journal of
Psychiatry, 46(8), 728-732.

Ben-Porath, Y. S., McCully, E., & Almagor, M. (1993). Incremental validity of the

MMPI-2 content scales in the assessment of personality and psychopathology by
self-report. Journal of Personality Assessment, 61(3), 557-575.

187

Beyko, M. J ., & Wong, S. C. P. (2005). Predictors of treatment attrition as indicators for
program improvement not offender shortcomings: A study of sex offender
treatment attrition. Sexual Abuse: A Journal of Research and Treatment, 1 7(4),
375-3 89.

Blanchard, K. A., Morgenstem, J ., Morgan, T. J ., Labouvie, E., & Bux, D. A. (2003).
Motivational subtypes and continuous measures of readiness for change:
Concurrent and predictive validity. Psychology of Addictive Behavior, 1 7(1), 56-
65.

Boulet, J ., & Boss, M. W. (1991). Reliability and validity of the Brief Symptom
Inventory. Psychological Assessment, 3(3), 433-437.

Bowman, K. M. (1938). The challenge of sex offenders: Psychiatric aspects of the
problem. Mental Hygiene, 22, 10-20.

Brogan, M. M., Prochaska, J. O., & Prochaska, J. M. (1999). Predicting termination and
continuation status in psychotherapy using the Trans-theoretical model.
Psychotherapy, 36, 105-113.

Brown, S. (2005). Treating Sex Oﬂenders: An Introduction to Sex Offender Treatment
Programmes. United Kingdom: Willan Publishing.

Butcher, J. N., Dahlstrom, W. G., Graham, J. R., Tellegan, A., & Kaemmer, B. (1989).
MMPI-2: Minnesota Multiphasic Personality Inventory-2: Manual for
administration, scoring, and interpretation. Minneapolis: University of Minnesota
Press.

Butcher, J. N., & Williams, C. L. (2000). Essentials of MMPI-2 and MMPI-A
Interpretation (2 Ed.). Minneapolis: University of Minnesota Press.

Carson, R. C. (1969). Interpretive Manual to the MMPI. In J. N. Butcher (Ed.), MMPI:
Research Developments and Clinical Applications (pp. 279-296). New York:
McGraw-Hill.

Chisholm, S. M., Crowther, J. H., & Ben-Porath, Y. S. (1997). Selected MMPI-2 scales
ability to predict premature termination and outcomes from psychotherapy.
Journal of Personality Assessment, 69(1), 127-144.

Craig, L. A., Browne, K. D., Beech, A., & Stringer, I. (2006). Differences in personality

and risk characteristics in sex, violent, and general offenders. Criminal Behavior
and Mental Health, 16, 183-194.

188

Craig, R. J., & Olson, R. E. (2003). Predicting the outcome of methadone maintenance
treatment with the negative treatment indicators content scale from the MMPI-2.
Psychological Reports, 93, 1056-1058.

Curnoe, S., & Langevin, R. (2002). Personality and deviant sexual fantasies: An
examination of the MMPIs of sex offenders. Journal of Clinical Psychology,
58(7), 803-815.

Curry, S. J ., Wagner, E. H., & Grothaus, L. C. (1991). Evaluation of intrinsic and
extrinsic motivation interventions with a self-help smoking cessation program.
Journal of Consulting and Clinical Psychology, 59, 318-324.

Dennison, S. M., Stough, C., & Birgden, A. (2001). The Big 5 dimensional personality
approach to understanding sex offenders. Psychology, Crime & Law, 7, 243-261.

Derogatis, L. R., & Melisaratos, N. (1983). The Brief Symptom Inventory: An
introductory report. Psychological Medicine, 13(3), 595-605.

Drapeau, M., Komer, C. A., Brunet, L., Granger, L., Caspar, F., Despland, J .-N., et al.
(2003). A psychodynamic look at pedophile sex offenders in treatment. Archives
of Psychiatry and Psychotherapy, 5(3), 31-42.

Duthie, B. & McIvor, D. L. (1990). A new system for cluster-coding child molester
MMPI proﬁle types. Criminal Justice and Behavior, 1 7(2), 199-214.

Eaton County Prosecuting Attorney. (2006). Michigan Legal Glossary. Retrieved ﬁom
http://www.eatoncountv.org/prosecutor/deﬁne.htm#csc.

Erickson, W. D., Luxenberg, M. G., Walbeck, N. H., & Seely, R. K. (1987). Frequency
of MMPI two-point code types among sex offenders. Journal of Consulting and
Clinical Psychology, 55(4), 566-570.

Finkelhor, D. (1984). Child Sexual Abuse: New Theory and Research. New York: The
Free Press.

Firestone, P., Nunes, K. L., Moulden, H., Broom, 1., & Bradford, J. M. (2005). Hostility
and recidivism in sexual offenders. Archives of Sexual Behavior, 34(3), 277-283.

Fisher, D. D., Beech, A., & Browne, K. D. (1999). Comparison of sex offenders to
nonoffenders on selected psychological measures. International Journal of
Oﬂender Therapy and Comparative Criminology, 43(4), 473-491.

Garos, S., Bleckley, M. K., Beggan, J. K., & Frizzell, J. (2004). Intrapsychic conﬂict and

deviant sexual behavior in sex offenders. Journal of Oﬂender Rehabilitation,
40(1/2), 23-40.

189

Geen, R. G. (1998). Aggression and antisocial behavior. In D. Gilbert, S. Fiske & G.
Lindzey (Eds.), The Handbook of Social Psychology (4 ed., Vol. 2, pp. 317-347).
Boston: The McGraw-Hill Company, Inc.

Gingrich, T. N., & Campbell, J. B. (1995). Personality characteristics of sexual offenders.
Sexual Addiction & Compulsivity, 2(1), 54-61.

Graham, K. R. (1994). The sexual addiction model in treatment of incarcerated offenders:
A study on recidivism. Sexual Addiction & Compulsivity, 1(3), 278-283.

Greenfeld, L. (1998). Sex offenses and offenders: An analysis of data on rape and sexual
assault. Retrieved. ﬁom http://www.rainn.org/docs/statistics/soo.pdf.

Groth, N. A., Burgess, A. W., & Holmstrom, L. L. (1977). Rape: Power, anger, and
sexuality. American Journal of Psychiatry, 134(1 1), 1239-1243.

Gudjonsson, G. H., & Sigurdsson, J. F. (2000). Diﬂerences and similarities between
violent offenders and sex offenders. Child Abuse & Neglect, 24(3), 363-3 72.

Hair Jr., J. F ., & Black, W. C. (2000). Cluster Analysis. In L. G. Grimm & P. R. Yarnold
(Eds), Reading and Understanding More Multivariate Statistics (pp. 147-205).
Washington, DC: American Psychological Association.

Hall, G. C. N., & Hirschman, R. (1991). Toward a theory of sexual aggression: A
quadripartite model. Journal of Consulting and Clinical Psychology, 59, 662-669.

Hall, G. C. N., & Hirschman, R. (1992). Sexual aggression against children: A
conceptual perspective of etiology. Criminal Justice and Behavior, 19, 8-23.

Hanson, R. K., & Morton-Bourgon, K. E. (2005). The characteristics of persistent sexual
offenders: A meta-analysis of recidivism studies. Journal of Consulting and
Clinical Psychology, 73(6), 1154-1 163.

Haywood, T. W., Grossman, L. S., Kravitz, H. M., & Wasyliw, O. E. (1994). Proﬁling
psychological distortion in alleged child molesters. Psychological Reports, 75,
915-927.

Hill, A., Briken, P., Kraus, C., Strohm, K., & Berner, W. (2003). Differential
pharmacological treatment of paraphilias and sex offenders. International Journal
of Oﬂender Therapy and Comparative Criminology, 47(4), 407-421.

Howells, K., Day, A., & Wright, S. (2004). Affect, emotions and sex offending.
Psychology, Crime & Law, 10(2), 179-195.

Hudson, S. M., Ward, T., & McCormack, J. C. (1999). Offense pathways in sexual
offenders. Journal of Interpersonal Violence, 14(8), 779-798.

190

Kalichman, S. C., Dwyer, M., Henderson, M. C., & Hoffman, L. (1992). Psychological
and sexual functioning among outpatient sexual offenders against children: A
Minnesota Multiphasic Personality Inventory (MMPI) cluster analytic study.
Journal of Psychopathology and Behavioral Assessment, 14(3), 259-276.

Kalichman, S. C., Shealy, L., & Craig, M. E. (1990). The use of the MMPI in predicting
treatment participation among incarcerated adult rapists. Journal of Psychology &
Human Sexuality, 3(2), 105-119.

Kalichman, S. C., Szyrnanowski, D., McKee, G., Taylor, J ., & Craig, M. E. (1989).
Cluster analytically derived MMPI proﬁle subgroups of incarcerated adult rapists.
Journal of Clinical Psychology, 45(1), 149-155.

Kanfer, F. H. (1975). Self-management models. In F. Kanfer & A. Goldstein (Eds),
Helping People Change. New York: Pergamon Press.

Kear-Colwell, J ., & Pollock, P. (1997). Motivation or confrontation: Which approach to
the child sex offender? Criminal Justice and Behavior, 24(1), 20-33.

Kessler, R. C., Berglund, P., Derrrler, O., Jin, R., Merikangas, K. R., & Walters, E.
(2005). Lifetime prevalence and age-of-onset distributions of DSM-IV disorders
in the national comorbidity survey replication. Arch Gen Psychiatry, 62, 593-768.

Lally, S. J. (2003). What tests are acceptable for use in forensic evaluations? A survey of
experts. Professional Psychology: Research and Practice, 34(5), 491-498.

Lalumiere, M. L., Chambers, L. J ., Quinsey, V. L., & Seto, M. C. (1996). A test of the
male deprivation hypothesis of sexual coercion. Ethnology and Sociobiology, 1 7,
299-318.

Langevin, R., Wright, P., & Handy, L. (1990a). Use of the MMPI and its derived scales
with sex offenders 1. Reliability and validity studies. Annals of Sex Research, 3,
245-291.

Langevin, R., Wright, P., & Handy, L. (1990b). Use of the MMPI and its derived scales
with sex offenders II. Reliability and criterion validity. Annals of Sex Research, 3,
453-486.

Lanyon, R. I. (2001). Psychological assessment procedures in sex offending. Professional
Psychology: Research and Practice, 32(3), 253-260.

Lee, J. K. P., Pattison, P., Jackson, H. J ., & Ward, T. (2001). The general, common, and

speciﬁc features of psychopathology for different types of paraphilias. Criminal
Justice and Behavior, 28(2), 227-256.

191

Levenson, J. S. & Cotter, L. P. (2005). The effects of Megan’s Law on sex offender
reintegration. Journal of Contemporary Criminal Justice, 21(1), 49-66.

Looman, J ., Dickie, I., & Abracen, J. (2005). Responsivity issues in the treatment of
sexual offenders. Trauma, Violence, & Abuse, 6(4), 330-353.

Luke, D. A. (2005). Getting the big picture in community science: Methods that capture
context. American Journal of Community Psychology, 35(3/4), 185-200.

Marques, J. K., Day, D. M., Nelson, C., & West, M. A. (1994). Effects of cogrritive-
behavioral treatment on sex offender recidivism: Preliminary results of a
longitudinal study. Criminal Justice and Behavior, 21(1), 28-54.

Marques, J. K., Nelson, C., Alarcon, J ., & Day, D. M. (2000). Preventing relapse in sex
offenders: What we learned from the SOTEP's experimental treatment program.
In D. R. Laws, S. M. Hudson & T. Ward (Eds.), Remaking relapse prevention
with sex oﬂenders: A sourcebook (pp. 321 -340). California: Sage.

Marshall, W. L., Anderson, D., & Champagne, F. (1997). Self-esteem and its relationship
to sexual offending. Psychology, Crime & Law, 3, 161-186.

Marshall, W. L., Cripps, B, Anderson, D., & Cortoni, F. A. (1999). Self-esteem and
coping strategies in child molesters. Journal of Interpersonal Violence, 14(9),
955-962.

Marshall, W. L., & Barbaree, H. E. (1990). An integrated theory of the etiology of sexual
offending. In W. L. Marshall, D. R. Laws & H. E. Barbaree (Eds.), Handbook of
Sexual Assault: Issues, Theories, and Treatment of the Oﬁ’ender (pp. 257-275).
New York: Plenum.

McConnaughy, E. A., Prochaska, J. O., & Velicer, W. F. (1983). Stages of change in
psychotherapy: Measurement and sample proﬁles. Psychotherapy: Theory,
Research, and Practice, 20, 368-375.

McCrady, B. S. (2001). Alcohol use disorders. In D. H. Barlow (Ed.), Clinical Handbook
of Psychological Disorders (3 ed., pp. 376-433). New York: The Guilford Press.

McMurran, M., Tyler, P., Hogue, T., Cooper, K., Dunseath, W., & McDaid, D. (1998).
Measuring motivation to change in offenders. Psychology, Crime & Law, 4, 43-
50.

Megargee, E. I. (1994). Using the Megargee MMPI-based classiﬁcation system with the
MMPI-2’s of male prison inmates. Psychological Assessment, 6(4), 327-334.

Megargee, E. I. (2006). Using the MMPI-2 in Criminal Justice and Correctional Settings.
Minnesota: University of Minnesota Press.

192

Megargee, E. I., Mercer, S. J ., & Carbonell, J. L. (1999). MMPI-2 with male and female
state and federal prison inmates. Psychological Assessment, 11(2), 177-185.

Michigan Department of Corrections. (2007). Policy Directive: Supervision of
probationers and parolees (Vol. 06.04.130, pp. 1-7).

Michigan Legislature. (1931). The Michigan Penal Code: Act 328 of 1931 (Vol.
LXXVI).

Miller, R. D. (2003). Chemical castration of sex offenders: Treatment or punishment? In
B. J. Winick & J. O. L. Fond (Eds.), Protecting Society from Sexually Dangerous
Offenders: Law, Justice, and Therapy (pp. 249-263). Washington, DC:
American Psychological Association.

Miller, H. A. & Rollnick, S. (2002). Motivational Interviewing: Preparing People for
Change (2” Ed. ). New York, NY: The Guilford Press.

Miller, H. A., & Young, G. R. (1998). Prison segregation: Administrative detention
remedy or mental health problem? Criminal Behavior and Mental Health, 7(1),
85-94.

Mofﬁtt, T. E., Caspi, A., Harrington, H., & Milne, B. (2002). Males on the life-course
persistent and adolescence-limited antisocial pathways: Follow-up at age 26.
Development & Psychopathology, 14, 179-206.

Murphy, W. D., Coleman, E. M., Haynes, M. R., & Stalgartis, S. (1979). Etiological
theories of coercive sexual behavior and their relationship to prevention.

Neidigh, L., & Tomiko, R. (1991). The coping strategies of child sexual abusers. Journal
of Sex Education & T herapy, 1 7(2), 103-110.

Nichols, H., & Molinder, I. (1984). Manual for the Multiphasic Sex Inventory. Tacoma:
Crime and Victim Psychology Specialists.

O'Laughlin, S., & Schill, T. (1994). The relationship between self-monitored aggression
and the MMPI-2, F, 4, 9 composite and anger content scale scores. Psychological
Reports, 74, 733-734.

Prentky, R. A., & Burgess, A. W. (1993). Hypothetical biological substrates of a fantasy-
based drive mechanism for repetitive sexual aggression. In A. W. Burgess (Ed.),
Rape and Sexual Assault 111 (pp. 23 5-256). New York: Garland.

Prochaska, J. M., DiClemente, C. C., & Norcross, J. C. (1992). In search of how people

change: Applications to addictive behaviors. American Psychologist, 4 7(9), 1102-
1 1 14.

193

Prochaska, J. O., & DiClemente, C. C. (1982). Transtheoretical therapy: Toward a more
integrative model of change. Psychotherapy: Theory, Research, and Practice, 19,
276-288.

Prochaska, J. O., & Prochaska, J. M. (1999). Why don't continents move? Why don't
people change? Journal of Psychotherapy Integration, 9(1), 83-102.

Project MATCH Research Group. (1997). Matching alcoholism treatments to client
heterogeneity: Project MATCH posttreatment drinking outcomes. Journal of
Studies on Alcohol, 58, 7-29.

Project MATCH Research Group. (1998). Matching alcoholism treatments to client
heterogeneity: Project MATCH three-year drinking outcomes. Alcoholism:
Clinical and Experimental Research, 22, 1300-131 1.

Quackenbush, R. E. (2003). The role of theory in the assessment of sex offenders.
Journal of Child Sexual Abuse, 12(3-4), 77-102.

Quayle, E., Vaughan, M., & Taylor, M. (2006). Sex offenders, internet child abuse
images, and emotional avoidance: The importance of values. Aggression and
Violent Behavior, 1 I , 1-1 1.

Rapkin, B. D. & Luke, D. A. (1993). Cluster analysis in community research:
Epistemology and practice. American Journal of Community Psychology, 21(2),
247-277.

Rokach, A. (2000). Offence type and coping with loneliness. Psychological Studies,
45(1&2), 1-8.

Ryan, K. M. (2004). Further evidence for a cognitive component of rape. Aggression and
Violent Behavior, 9, 579-604.

Sarkar, N. N. & Sarkar, R. (2006). Sexual assault: A crime against women in the
perspective of science, society, and law. International Medical Journal, 13(1),
31-36.

Schwartz, B. (1997). Theories of sex offenders. In B. Schwartz & H. Cellini (Eds.), The
Sex Offender: New Insights, Treatment Innovations, and Legal Developments (pp.
2-32). Kingston, NJ: Civic Research Institute.

Schwartz, M. F ., & Masters, W. H. (1993). Integration of trauma-based, cognitive
behavioral, systemic and addiction approaches for treatment of hypersexual pair-
bonding disorder. In P. J. Barnes (Ed.), Sexual Addiction and Compulsivity (Vol.
1, pp. 57-76). New York: Brunner Mazel.

194

Serran, G., Fernandez, Y., Marshall, W. L., & Mann, R. E. (2003). Process issues in
treatment: Application of sexual offender programs. Professional Psychology,
Research and Practice, 34(4), 368-374.

Seto, M. C., & Barbaree, H. E. (1999). Psychopathology, treatment behavior, and sex
offender recidivism. Journal of Interpersonal Violence, 14, 1235-1248.

Shealy, L., Kalichman, S. C., Henderson, M. C., Szyrnanowski, D., & McKee, G. (1991).
MMPI proﬁle subtypes of incarcerated sex offenders against children. Violence
and Victims, 6(3), 201-212.

Shine, J ., McClosky, H., & Newton, M. (2002). Self-esteem and sex offending. The
Journal of Sexual Aggression, 8(1), 51-61.

Smallbone, S. W., & Milne, L. (2000). Associations between trait anger and aggression
used in the commission of sexual offenses. International Journal of Oﬂender
Therapy and Comparative Criminology, 44(5), 606-617.

Smith, M. H. (1924). The mental conditions found in certain sex offenders. Lancet, 206,
643-646.

Sommers, M. S. (2007). Deﬁning patterns of genital injury from sexual assault. Trauma,
Violence, & Abuse, 8(3), 270-280.

Sommers, M. S., Schafer, J ., Zink, T., Hutson, L., & Hillard, P. (2001). Injury patterns in
women resulting from sexual assault. Trauma, Violence, & Abuse, 2(3), 240-258.

Sommers, M.S., et al. (2006). The effects of age and ethnicity on physical injury from
rape. Clinical Scholarship for the Care of Women, Childbearing Families, &
Newborns, 35(2), 199-207.

Steinley, D. (2003). Local optima in K-means clustering: What you don’t know may hurt
you. Psychological Methods, 8(3), 294-304.

Stinson, J. D., Becker, J. V., & Tromp, S. (2005). A preliminary study on ﬁndings of

psychopathy and affective disorders in adult sex offenders. International Journal
of Law and Psychiatry, 28, 63 7-649.

Strassberg, D. S. (1997). A cross-national validity study of four MMPI-2 content scales.
Journal of Personality Assessment, 69(3), 596-606.

Terry, K. J ., & Mitchell, E. W. (2001). Motivation and sex offender treatment efﬁcacy:

Leading a horse to water and making it drink? International Journal of Offender
Therapy and Comparative Criminology, 45(6), 663-672.

195

Tierney, D. W., & McCabe, M. P. (2002). Motivation for behavior change among sex
offenders: A review of the literature. Clinical Psychology Review, 22, 113-129.

Tierney, D. W., & McCabe, M. P. (2004). The assessment of motivation for behaviour
change among sex offenders against children: An investigation of the utility of the
Stages of Change Questionnaire. Journal of Sexual Aggression, 10(2), 23 7-249.

Tierney, D. W., & McCabe, M. P. (2005). The utility of the Trans-Theoretical Model of
Behavior Change in the treatment of sex offenders. Sexual Abuse: A Journal of
Research and Treatment, 1 7(2), 153-170.

Tjaden, P., & Thoennes, N. (1998). Full report of the prevalence, incidence, and
consequences of violence against women: Research report. Retrieved. from
http://www.rainn.org/docs/statistics/fullnvawsurvev.pdf?PHPSESSID=6c89660e6
cc455a59b88e8b2b93c26e0.

 

Valliant, P. M., & Blasutti, B. (1992). Personality differences of sex offenders referred
for treatment. Psychological Reports, 71, 1067-1074.

Vandiver, D. M., & Walker, J. T. (2002). Female sex offenders: An overview and
analysis of 40 cases. Criminal Justice Review, 27(2), 284-300.

Velasquez, R. J ., Callahan, W. J ., & Carrillo, R. (1989). MMPI proﬁles of Hispanic-
American inpatient and outpatient sex offenders. Psychological Reports, 65(3),
1055-1058.

Velicer, W. F., Prochaska, J. O., Bellis, J. M., DiClemente, C. C., Rossi, J. S., Fava, J. L.,
et al. (1993). An expert system intervention for smoking cessation. Addictive
Behaviors, 18(269-290). '

Ward, T., & Beech, A. (2006). An integrated theory of sexual offending. Aggression and
Violent Behavior, 11, 44-63.

Ward, T., & Gannon, T. A. (2006). Rehabilitation, etiology, and self-regulation: The
comprehensive good lives model of treatment for sexual offenders. Aggression
and Violent Behavior, 11, 77-94.

Ward, T., & Hudson, S. M. (1998). A model of the relapse process in sexual offenders.
Journal of Interpersonal Violence, 13(6), 700-725.

Ward, T., & Hudson, S. M. (1998). The construction and development of theory in the
sexual offending area: A metatheoretical ﬁ'amework. Sexual Abuse: A Journal of

Research and Treatment, 10(1), 47-63.

Ward, T., & Siegert, R. J. (2002). Toward a comprehensive theory of child sexual abuse:
A theory knitting perspective. Psychology, Crime & Law, 9, 319-351.

196

11111

5