Ppi
Ppi
by
Natalia L. Nikolova
M.A. (Psychology), Simon Fraser University, 2009
in the
Department of Psychology
Faculty of Arts and Social Sciences
Kevin S. Douglas
Senior Supervisor
Associate Professor
Stephen D. Hart
Supervisor
Professor
Marlene M. Moretti
Supervisor
Professor
William Glackman
Internal Examiner
Associate Professor
School of Criminology
Martin Sellbom
External Examiner
Assistant Professor
Department of Psychology
University of Alabama
ii
Partial Copyright Licence
iii
Ethics Statement
The author, whose name appears on the title page of this work, has obtained, for the
research described in this work, either:
a. human research ethics approval from the Simon Fraser University Office of
Research Ethics,
or
b. advance approval of the animal care protocol from the University Animal Care
Committee of Simon Fraser University;
or
A copy of the approval letter has been filed at the Theses Office of the University Library
at the time of submission of this thesis or project.
The original application for approval and letter of approval are filed with the relevant
offices. Inquiries may be directed to those authorities.
iv
Dedication
v
Acknowledgements
I am extremely grateful to everyone who was involved and made the completion of this
project possible. First, I would like to express my sincere gratitude to Dr. Kevin S.
Douglas, my senior supervisor, for being such an amazing mentor. He was always
supportive and available to provide invaluable guidance and feedback, and was very
generous in sharing his insight and expertise. The criticism he provided was always
constructive and delivered gently, which I appreciated very much. I feel very fortunate to
have had the opportunity to work with him over the past several years, as it was a
remarkable experience and I learned a lot!
I would also like to thank Dr. Stephen D. Hart and Dr. Marlene M. Morretti who served as
supervisory committee members. Their support throughout the entire process meant a
lot, and their invaluable expertise, constructive advice, and feedback really helped
improve this project. Special thanks go to my external examiner, Dr. Martin Sellbom,
who travelled from Alabama to attend the defense in person, and provided thoughtful
questions and insightful feedback during the defense. Thank you also to my internal
examiner, Dr. William Glackman, for his well thought out questions regarding this
project. I am thankful to all who were present during my defense for creating such a kind
and collegial atmosphere, and for making it an unforgettable experience.
I would like to express my sincere thanks to my close friends and colleagues, Andrea
Gibas and Diane Strub, as well as my partner, John W. Lewin, for their invaluable
support and inspiration, and for donating their time so generously during the final editing
stages of this document. My special thanks also go to my family and close friends for
their emotional support, encouragement, and reassurance during the entire process, as
without it I am not entirely convinced I would have endured.
My keen appreciation goes to Anita Turner, Graduate Program Assistant, for always
being so responsive, knowledgeable, and willing to help, as well as for her assistance in
sorting through the administrative aspect of this process.
Finally, I would like to express my gratitude to the Ministry of Public Safety and Solicitor
General, British Columbia, for allowing me access to their data and resources, as
without their assistance this project would not have been possible.
vi
Table of Contents
Approval .......................................................................................................................... ii
Partial Copyright Licence ............................................................................................... iii
Abstract .......................................................................................................................... iv
Dedication ....................................................................................................................... v
Acknowledgements ........................................................................................................ vi
Table of Contents .......................................................................................................... vii
List of Tables ................................................................................................................... x
List of Figures................................................................................................................ xiii
List of Acronyms............................................................................................................xiv
1. Introduction .......................................................................................................... 1
1.1. Psychopathy and Social Deviance or Crime ........................................................... 2
1.2. Conceptualizations of Psychopathy ........................................................................ 4
1.3. Frequently Used Measures of Psychopathy ........................................................... 8
1.3.1. Interview-Based Measures/Observer-Ratings of Psychopathy .................... 8
Psychopathy Checklist–Revised ................................................................. 8
Comprehensive Assessment of Psychopathic Personality–
Institutional Rating Scale (CAPP–IRS) ................................................ 10
1.3.2. Self-Report Measures of Psychopathy ...................................................... 11
Levenson Primary and Secondary Psychopathy Scales (LPSP) ............... 15
Self-Report Psychopathy Scale (SRP) ...................................................... 16
1.3.3. Psychopathic Personality Inventory (PPI).................................................. 17
Psychopathic Personality Inventory: Short Form ....................................... 23
Psychopathic Personality Inventory–Revised (PPI–R) .............................. 24
1.4. Incremental Validity of Psychopathy ..................................................................... 26
1.5. Psychopathy and Gender ..................................................................................... 28
1.5.1. Why Study Psychopathy among Females? ............................................... 28
1.5.2. Differences in Symptoms and Symptom Severity ...................................... 29
1.5.3. Differences in Behavioral Manifestation, Clinical Presentation, and
Comorbidity............................................................................................... 33
1.5.4. Differences with Regard to Personality and Behavioral Correlates,
and Recidivism ......................................................................................... 36
1.5.5. Factor Structure of Commonly-Used Psychopathy Measures ................... 38
1.5.6. Suitability of Existing Measures to Assess Psychopathy across
Gender...................................................................................................... 39
2. Current Study...................................................................................................... 42
2.1. Research Questions and Hypotheses .................................................................. 43
2.1.1. Research Question 1: Reliability of the PPI–R .......................................... 43
2.1.2. Research Question 2: Predictive Validity of the PPI–R with Respect
to Recidivism and Type of the Offense(s) ................................................. 44
2.1.3. Research Question 3: Incremental Validity of the PPI–R with
Respect to Recidivism .............................................................................. 44
Incremental Validity Relative to General Self-Report Measures of
Personality .......................................................................................... 44
Incremental Validity Relative to the CRNA ................................................ 45
vii
2.1.4. Research Question 4: Moderating Effects of Gender on PPI–R-
Assessed Psychopathy ............................................................................. 45
Do PPI–R Scores Differ across Gender? .................................................. 45
Does Gender Moderate the Predictive Utility of the PPI–R with
Respect to Recidivism? ....................................................................... 46
Does Gender Moderate the Relationship between PPI–R-Assessed
Psychopathy and Type of Offense(s)?................................................. 46
2.2. Method ................................................................................................................. 46
2.2.1. Participants ............................................................................................... 46
2.2.2. Comparative Measures ............................................................................. 48
NEO–Five Factor Inventory (NEO–FFI) .................................................... 48
Personality Assessment Inventory (PAI) ................................................... 49
Community Risk/Needs Assessment (CRNA) ........................................... 50
2.2.3. Criterion Measure ..................................................................................... 51
2.2.4. Procedure ................................................................................................. 52
2.2.5. Ethics Approval ......................................................................................... 53
2.2.6. Data Manipulation and Data Analytic Issues ............................................. 53
2.2.7. Data Analyses ........................................................................................... 55
Research Question 1: Reliability of the PPI–R .......................................... 55
Research Question 2: Predictive Validity of the PPI–R with Respect
to Recidivism and Type of the Offense(s) ............................................ 57
Research Question 3: Incremental Validity of the PPI–R with Respect
to Recidivism ....................................................................................... 59
Incremental Validity Relative to General Self-Report Measures
of Personality................................................................................. 59
Incremental Validity Relative to the CRNA .......................................... 60
Research Question 4: Moderating Effects of Gender on PPI–R-
Assessed Psychopathy ....................................................................... 60
Do PPI–R Scores Differ across Gender?............................................. 60
Does Gender Moderate the Predictive Utility of the PPI–R with
Respect to Recidivism? ................................................................. 61
Does Gender Moderate the Relationship between PPI–R-
Assessed Psychopathy and Type of Offense(s)? .......................... 61
3. Results ................................................................................................................ 62
3.1. Sample Characteristics ......................................................................................... 62
3.2. Assessment of Assumptions................................................................................. 65
3.3. Descriptive Information ......................................................................................... 67
3.4. Research Question 1: Reliability of the PPI–R ...................................................... 68
3.4.1. Cronbach’s Alpha Coefficients/Stratified Alpha Coefficients ...................... 68
3.4.2. Corrected Item-To-Total Correlations and Mean Inter-Item
Correlations .............................................................................................. 68
3.5. Research Question 2: Predictive Validity of the PPI–R with Regard to
Recidivism and Type of Offense(s) ....................................................................... 69
3.5.1. Predictive Validity until Time to Re-Offend (Over the Entire Follow-
Up Period)................................................................................................. 69
3.5.2. Predictive Validity for a Set Follow-Up Period of 6 Months ........................ 76
3.6. Research Question 3: Incremental validity of the PPI–R with Regard to
Recidivism ............................................................................................................ 78
viii
3.6.1. Incremental Validity Relative to General Self-Report Personality
Inventories ................................................................................................ 78
3.6.2. Incremental Validity Relative to the CRNA ................................................ 81
3.7. Research Question 4: Moderating Effects of Gender on PPI–R-Assessed
Psychopathy ......................................................................................................... 85
3.7.1. Do PPI–R Scores Differ across Gender? .................................................. 85
3.7.2. Does Gender Moderate the Predictive Utility of the PPI–R with
Respect to Recidivism?............................................................................. 86
3.7.3. Does Gender Moderate the Relationship between PPI–R-Assessed
Psychopathy and Type of Offense(s)? ...................................................... 88
3.8. Supplemental Analyses ........................................................................................ 96
3.8.1. Invalid/Missing data................................................................................... 96
ix
List of Tables
Table 5. Areas Under the Curve (AUCs) for Receiver Operating Characteristics
Analysis for Recidivism and Type of Offense(s) ........................................... 70
Table 8. Utility of the PPI–R to Differentiate between the Type of Post Release
Offense(s) among Those Who Recidivated.................................................. 75
Table 10. Multivariable Logistic Regression -- Odds Ratios for the Associations
between PPI–R Scores and Recidivism over the 6-month Follow-Up
Period .......................................................................................................... 77
Table 11. Bivariate Correlations between the PPI–R, PAI, NEO–FFM Scores
and Recidivism ............................................................................................ 79
Table 12. Incremental Validity of the PPI–R Relative to the PAI vis-à-vis
Recidivism over the Entire Follow-Up Period ............................................... 80
Table 13. Incremental Validity of the PPI–R vis-à-vis Recidivism over the 6-
Month Follow-Up Period .............................................................................. 80
Table 14. Bivariate Correlations between the PPI–R Scores, CRNA Ratings,
and Recidivism (Entire Follow-Up/6-Month Follow-Up Period) ..................... 81
Table 15. Incremental Validity of the PPI–R vis-à-vis Recidivism over the Entire
Follow-Up Period ......................................................................................... 83
Table 16. Incremental Validity of the PPI–R vis-à-vis Recidivism over the 6-
Month Follow-Up Period .............................................................................. 84
x
Table 17. Group Statistics by Gender ......................................................................... 85
Table 20. Moderation Analyses for Gender and the PPI–R Total Score vis-à-vis
Future Crime and Violence .......................................................................... 87
Table 21. Moderation Analyses for Gender and the SCI Factor Score vis-à-vis
Future Crime and Violence .......................................................................... 88
Table 22. Moderation Analyses for Gender and the PPI–R Total Score vis-à-vis
Gravity of the Offense (VIOLENT offenses) ................................................. 89
Table 23. Moderation Analyses for Gender and the PPI–R Total Score vis-à-vis
Gravity of the Offense (NON-VIOLENT offenses) ........................................ 89
Table 24. Moderation Analyses for Gender and the PPI–R Total Score vis-à-vis
Gravity of the Offense (OTHER offenses) .................................................... 89
Table 25. Moderation Analyses for Gender and the SCI Factor Score vis-à-vis
Gravity of the Offense (VIOLENT offenses) ................................................. 90
Table 26. Moderation Analyses for Gender and the SCI Factor Score vis-à-vis
Gravity of the Offense (NON-VIOLENT offenses) ........................................ 90
Table 27. Moderation Analyses for Gender and the SCI Factor Score vis-à-vis
Gravity of the Offense (OTHER offenses) .................................................... 90
Table 28. Moderation Analyses for Gender and the FD Factor Score vis-à-vis
Gravity of the Offense (VIOLENT offenses) ................................................. 91
Table 29. Moderation Analyses for Gender and the FD Factor Score vis-à-vis
Gravity of the Offense (NON-VIOLENT offenses) ........................................ 91
Table 30. Moderation Analyses for Gender and the FD Factor Score vis-à-vis
Gravity of the Offense (OTHER offenses) .................................................... 91
Table 31. Moderation Analyses for Gender and the C Factor Score vis-à-vis
Gravity of the Offense (VIOLENT offenses) ................................................. 92
Table 32. Moderation Analyses for Gender and the C Factor Score vis-à-vis
Gravity of the Offense (NON-VIOLENT offenses) ........................................ 92
Table 33. Moderation Analyses for Gender and the C Factor Score vis-à-vis
Gravity of the Offense (OTHER offenses) .................................................... 92
xi
Table 34. Moderation Analyses for Gender and the PPI–R Scores vis-à-vis
Utility of the PPI–R to Differentiate between the Type of Post-Release
Offense(s) (VIOLENT offenses) ................................................................... 94
Table 35. Moderation Analyses for Gender and the PPI–R Scores vis-à-vis
Utility of the PPI–R to Differentiate between the Type of Post-Release
Offense(s) (NON-VIOLENT offenses) .......................................................... 95
Table 36. Moderation Analyses for Gender and the PPI–R Scores vis-à-vis
Utility of the PPI–R to Differentiate between the Type of Post-Release
Offense(s) (OTHER offenses) ...................................................................... 95
Table 37. Bivariate Correlations between PPI–R Total, Factor and Scale scores
and Overall Recidivism ................................................................................ 97
Table 39. Prospective Validity of the PPI–R vis-à-vis Type of Post Release
Offense(s) .................................................................................................... 98
Table 40. Univariate Logistic Regression -- Odds Ratios for the Associations
between PPI–R Scores and Recidivism over the 6-Month Follow-Up
Period ........................................................................................................ 102
Table 41. Multivariable Logistic Regression -- Odds Ratios for the Associations
between PPI–R Scores and Recidivism during the 6-Month Follow-up
Period ........................................................................................................ 103
xii
List of Figures
xiii
List of Acronyms
xiv
LSRP Levenson Self-Report Psychopathy
MCMI–II Millon Clinical Multiaxial Inventory–II
MCMI–III Millon Clinical Multiaxial Inventory–III
MC–SDS Marlowe-Crowne Social Desirability Scale
ME Macheavelian Egocentricity scale
MIC Mean inter-item correlations
MMPI–2 Minnesota Multiphasic Personaltiy Inventory–2
MPQ Multidimensional Personality Questionnaire
NAS Novaco Anger Scale
NEO–FFI NEO–Five Factor Inventory
NEO–PI–R NEO Personality Inventory–Revised
PAI Personality Assessment Inventory
PCL Psychopathy Checklist
PCL:SV Psychopathy Checklist: Screening Version
PCL:YV Psychopathy Checklist: Youth Version
PCL–R Psychopathy Checklist–Revised
PDQ–R Personality Diagnostic Questionnaire–Revised
PH Proportional hazards
PICTS Psychological Inventory of Criminal Thinking Styles
PPI Psychopathic Personality Inventory
PPI:SF Psychopathic Personality Inventory: Short Form
PPI–R Psychopathic Personality Inventory–Revised
RN Rebellious Nonconformity scale
ROC Receiver operating characteristic analysis
SCI Self-Centered Impulsivity factor
SIR Statistical Information for Recidivism scale
SOI Social Influence scale
SRP Self-Report Psychopathy scale
SRP–II Self-Report Psychopathy scale–II
SRP–III Self-Report Psychopathy scale–III
STI Stress Immunity scale
SUD Substance use disorders
VRAG Violence Risk Appraisal Guide
xv
1. Introduction
The construct of psychopathy has been studied extensively over the past few
decades, and not surprisingly it is of interest to various branches of the criminal justice
system, as incarcerated individuals who score high on measures of psychopathy are
more prone to engage in institutional infractions, as well as crime and violence, following
release from custody (Edens, Poythress, Lilienfeld, & Patrick, 2008). The majority of
these studies were conducted using a criminal sample and psychopathy was
operationalized using the Psychopathy Checklist–Revised ([PCL–R; Hare, 1991, 2003];
Patrick, Edens, Poythress, Lilienfeld, & Benning, 2006). Nevertheless, over the past
several years there has been an emergence of new measures of psychopathy, some of
which are self-reports. The purpose of this study is: (1) to evaluate the internal
consistency of the Psychopathic Personality Inventory-Revised (PPI–R; Lilienfeld &
Widows, 2005), (2) to examine its predictive validity vis-à-vis recidivism, (3) to evaluate
its incremental validity in terms of predicting recidivism relative to two general measures
of personality: The Personality Assessment Inventory (PAI; Morey, 1991) and the NEO–
Five Factor Inventory (NEO–FFI; Costa & McCrae, 1992b), as well as relative to the
Community Risk/Needs Assessment (CRNA; Glackman, n.d.), and finally (4) to evaluate
and compare the performance of the PPI–R across gender, as well as to examine the
moderating effects of gender on the predictive utility of the PPI–R vis-à-vis crime and
violence, and the type of post release offense(s). In the following pages, a discussion of
the link between psychopathy and criminality or violence will be presented, followed by a
review of prominent models of psychopathy. Further, existing measures of psychopathy
will be reviewed (with a particular emphasis on the PPI–R as it is the focus of this study),
followed by a discussion on the incremental validity of psychopathy, and an overview of
psychopathy among women. Finally, the purpose, goals, and expected outcomes of the
current study will be discussed in detail.
1
1.1. Psychopathy and Social Deviance or Crime
Various branches of the criminal justice system need to make decisions about
people’s freedom (i.e., who is safe to be released from custody, what level of
management will be needed, etc.). This essentially requires relatively accurate
prediction of violence (Douglas, Vincent, & Edens, 2006). Diagnosing psychopathy in
making such predictions was shown to be promising in early research by Hart, Kropp,
and Hare (1988), which in turn sparked research on the association between
psychopathy and violence.
2
tools, Yang, Wong, and Coid (2010) only found support for the predictive validity of
Factor 2 of the PCL–R vis-à-vis violence at a moderate level, while the predictive
accuracy of Factor 1 was only at a chance level among men.
Similar trends were found in a meta-analytic study of the predictive validity of the
PCL–R vis-à-vis institutional adjustment and recidivism (Walters, 2003). Specifically,
Factor 2 was moderately correlated with institutional adjustment and recidivism, while
the associations between Factor 1 and those outcomes were less robust. Stronger
associations for Factor 2 and institutional misconduct were also reported in a meta-
analytic review by Guy, Edens, Anthony, and Douglas (2005). Nevertheless, Kennealy,
Skeem, Walters, and Camp (2010) conducted a meta-analysis to determine whether
core interpersonal and affective traits of psychopathy interacted with disinhibition and
antisocial behavior in the prediction of violence. Their results did not provide support for
an interaction between the Interpersonal-Affective and the Social-Deviance scales of the
PCL–R, and revealed that the Social-Deviance scale is more uniquely predictive of
violence.
It is worth pointing out that there is often notable heterogeneity across studies
included in meta-analyses, some of which is attributable to design variation (i.e.,
postdictive vs. predictive), as well as variation in the measurement of psychopathy (i.e.,
PCL measures vs. self-reports), or measurement of recidivism and testing outcomes
(i.e., operational definitions of outcomes; see Douglas et al., 2006). Such heterogeneity
may be among the contributing factors to detected differences in outcome.
Overall, these findings provide ample support for the relationship between
psychopathy and crime or violence, and as Douglas et al. (2006) point out “the field may
have reached an asymptote in terms of novel information to be gleaned from studies that
simply address the question whether an association exists between psychopathy and
recidivism” (pg. 548). Therefore a shift in focus toward exploring other research
questions (e.g., predictive and incremental validity of psychopathy in various settings or
samples, further exploration of newer conceptualizations of psychopathy in terms of
prediction of crime and violence, etc.) may be needed in order to advance the literature
on psychopathy and social deviance or crime (Douglas et al., 2006). Thus, the goal of
the present study was to evaluate the incremental and predictive validity of a newer self-
3
report measure of psychopathy (i.e., the PPI–R), and to compare its validity across male
and female offenders. Exploring these particular research questions was deemed
important as of the writing of this manuscript there has been no research focusing on
these specific aspects of the psychometric properties of the PPI–R. In addition, given a
majority of the research on psychopathy has focused on men (see Kreis & Cooke, 2011;
John, 2009; Logan & Weizmann-Henelius, 2012; Stanford & Felthous, 2011; Wynn,
Høiseth, & Pettersen, 2012), and in consideration of the very limited research on gender
and the PPI–R with offender samples, evaluating the effects of gender was considered
essential.
The definition of psychopathy has evolved over the years—it is now defined as a
more extreme variant of Antisocial Personality Disorder (APD; American Psychiatric
Association, 2000, p. 703), which focuses primarily on behavioral deficits and
maladaptive lifestyle. Specifically, psychopathy is defined as a personality disorder
characterized by affective, interpersonal, and behavior deficits. Those deficits are
evident through unstable and impulsive behavior and lifestyle, a tendency to violate
social norms, as well as egocentricity, grandiosity, callousness, lack of empathy,
remorse or guilt, shallow affect, manipulativeness, irresponsibility and short-temper
4
(Hare, 2006). Psychopathy is considered to be stable across interpersonal and social
domains, as well as across the life span.
The three- and the four-factor operational models are quite similar as they
encompass overlapping parts of the same construct. In fact, the main difference
between them is reflected in the way they conceive the relationship between
psychopathy and antisocial or socially deviant behavior (Douglas et al., 2006). Namely,
whereas behavioral deviance is conceptualized as a mere consequence of psychopathy
according to the three-factor model, it is actually understood to be an inherent
characteristic of the disorder according to the four-factor model (Cooke & Michie, 2001;
5
Cooke, Michie, Hart, & Clark, 2004; Hare, 2003). Despite this disagreement between
scholars, as discussed in the previous section, there is a large body of literature that has
found support for a robust association between psychopathy and antisocial/socially
deviant behavior (see Cooke & Michie, 2001; Cooke et al., 2004; Douglas et al., 2006;
Hare & Neumann, 2006; Hare & Neumann, 2010; Skeem & Cooke, 2010a; Widiger,
2006), which carries practical implications for the fields of law enforcement, criminal
justice, and clinical-forensic psychology.
The triarchic model is another model of psychopathy that was recently proposed
by Patrick, Fowles, and Krueger (2009). Unlike the two-, three-, and four-factor
operational models, which are based on examinations of the factor structure of various
psychopathy measures, the triarchic model was developed based on a review of
historical conceptualizations of psychopathy together with empirical findings about the
most prominent assessment measures of psychopathy. It is an integrative model,
according to which psychopathy includes three phenotypic constructs: Disinhibition,
boldness, and meanness. It incorporates not only historical and contemporary
conceptualizations of psychopathy, but also broader personality, psychopathology, and
neurobiological concepts (Skeem, Polaschek, Patrick, & Lilienfeld, 2011).
6
Overall, the conceptualization of psychopathy and its association with crime and
violence have been well established. In addition, there are a few well researched
operational models of psychopathy, which will likely continue to be revised or fine tuned
as new empirical evidence emerges. In order to take advantage of this knowledge,
however, we must have measures that allow for accurate and reliable assessment of
psychopathy. It is a common practice in law enforcement as well as in the field of
clinical-forensic psychology to take into account past social deviance when attempting to
estimate the risk for future social deviance. Nevertheless, Cooke et al. (2004) pointed
out an inherent problem with this practice. Although an individual’s potential and
capability to engage in particular acts can very well be assessed based on their past
behavior, whether or not he/she will engage in the same patterns of behavior in the
future cannot be established with certainty. One way to avoid such tautology would be
to employ measures that independently assess psychopathic personality features and
socially deviant behavior. It is also important to avoid mono-operation bias in assessing
psychopathy (i.e., relying on a single measure), as well as doing so in a limited number
of settings, as that may lead to serious construct underrepresentation (Cooke, Michie, &
Hart, 2006). This of course does not imply that the field is limited due to a lack of
available psychopathy measures and therefore there is a need to focus on the
development of new measures. Quite the contrary -- there are a number of measures
available, and as a way of shedding light on the current state of the practice of assessing
psychopathy, an overview of measures typically utilized in the field of clinical-forensic
psychology is presented next. Of note, creating an exhaustive catalogue of the existing
measures of psychopathy is beyond the scope of this dissertation. Therefore, only
measures which have been researched extensively and have been shown to have well-
established or promising psychometric properties, and are therefore typically used in
practice, were selected for review.
7
1.3. Frequently Used Measures of Psychopathy
Psychopathy Checklist–Revised
The PCL–R has two derivatives, the Screening Version (PCL:SV; Hart, Cox, &
Hare, 1995) and the Youth Version (PCL:YV; Forth, Kosson, & Hare, 2003), which are
psychometrically, empirically, and conceptually similar to the PCL–R (Forth et al., 2003;
Hare & Neumann, 2006). However, they are outside of the scope of this dissertation
and will not be reviewed in detail. In terms of the psychometric properties of the PCL–R,
original validation data included in the 2nd edition of its manual (Hare, 2003) reveals
high internal consistency, and interrater reliability. Nevertheless, a recent study raised
concerns about the PCL–R Factor 1 score, which was found to have exceedingly poor
interrater reliability (Edens, Boccaccini, & Johnson, 2010). Descriptive and validity data
for different offender groups, as well as forensic psychiatric patients, is also available in
the manual, and are considered to be adequate. It is worth pointing out that the two
factors were moderately correlated, which could be interpreted as support for their
convergent validity, but could also raise questions about their discriminant validity.
8
Even though the PCL–R was designed strictly for the assessment of the clinical
construct of psychopathy, its utility for assessing risk for recidivism and other applied
purposes has been evaluated extensively. While some studies show support for its
utility in assessing recidivism risk (see Dolan & Doyle, 2000; Hare & Neumann, 2006;
Olver, Neumann, Wong, & Hare, 2012), others have raised concerns or have shown that
the PCL–R is outperformed by other measures in that regard (Gendreau, Goggin, &
Smith, 2002; Guy, Douglas, & Hendry, 2010; Skeem & Cooke, 2010b; Yang et al.,
2010). Similarly, the predictive validity of the PCL–R vis-à-vis crime and antisocial
behavior has been examined in male and female samples, with support shown by some
scholars (e.g., Campbell, French, & Gendreau, 2009; Dolan & Völlm, 2009; Walters,
2003; Yang et al., 2010), while others have raised questions in that respect (McKeown,
2010; Nicholls & Petrila, 2005; Strand & Belfrage, 2005).
When the PCL–R is disaggregated into its two factors, the prevailing literature
indicates that Factor 2 (encompassing the Antisocial and Lifestyle facets) is a stronger
predictor of recidivism and violence than Factor 1 (encompassing the Interpersonal and
Affective facets; see Gonsalves, Scalora, & Huss, 2009; Kennealy et al., 2010; Olver et
al., 2012; Walters, 2003). Further, with regard to its incremental validity a recent study
(Walters, Wilson, & Glover, 2011) revealed that when the Antisocial and Lifestyle facets
were combined into a single factor, it displayed incremental validity relative to the
Interpersonal and Affective facets. In addition, the Antisocial summed composite score
predicted recidivism significantly better than the summed composite score of the other
three facets (i.e., Interpersonal, Affective, or Lifestyle). These findings were in line with
two previous studies on general and violent recidivism (see Walters & Heilbrun, 2010;
Walters, Knight, Grann, & Dahle, 2008). Nevertheless, in another study, the composite
of the Interpersonal, Affective, and Lifestyle facets failed to predict general and violent
recidivism once age and criminal history were taken into account (Walters, 2011). All in
all, while extensive support for the psychometric properties of the PCL–R exists, some
contradicting findings have been recently reported. In addition, even though the PCL–R
was not originally developed as a risk assessment instrument, it has been extensively
used in that capacity. Yet, there appears to be a lack of agreement in the field in terms
of its utility to assess risk for recidivism, which has raised questions and concerns in that
regard.
9
Comprehensive Assessment of Psychopathic Personality–Institutional
Rating Scale (CAPP–IRS)
The primary difference between the two versions is the method of administration.
The Life-Time version is to be administered by trained professionals, and ratings are
based on an interview. The Staff Rating Scale, on the other hand, is intended for use by
paraprofessionals at secured facilities (i.e., correctional staff members), and therefore
imposes fewer demands for training prior to its use. Ratings of psychopathic symptoms
are based on observations over a six-month time frame during which the participant has
been residing at the facility. Due to the fact that it covers six domains of psychopathy,
the CAPP is optimized for use in a variety of settings—at secured treatment facilities as
well as in the community—instead of being limited for use in a single setting. In addition,
the CAPP is designed to assess not only lifetime severity of symptoms, but also changes
in severity over discrete periods of time, and as such can be useful when the temporal
stability of symptom severity is of interest.
Considering the recent development of this measure, currently there are only a
couple of published studies on the CAPP–IRS. Nevertheless, in combination with data
from unpublished studies, there is preliminary support for what appear to be promising
psychometric properties of the CAPP–IRS. Specifically, there is preliminary support for
its high internal consistency, and good to excellent interrater reliability (see Corrado,
Watkinson, Hart, & Cohen, 2006; Corrado, Watkinson, Hart, Lussier, & Cohen, 2007;
McCormick, Corrado, Hart, & Cohen, 2008). Preliminary support for the content and
construct validity of the instrument has also been reported (see Cooke, Hart, Logan, &
Kreis, 2009; Cooke, Logan, Kreis & Hoff, 2009; Kreis, & Cooke, 2011; Nikolova, 2009).
Specifically, Kreis & Cooke (2011) found that the interpersonal domains of the CAPP
10
(i.e., Attachment, Dominance, and Self) were most salient of psychopathy, while the
Cognitive domain was the least salient. Further, there was support for the association
between violence and some of the items on the CAPP–IRS (i.e., “suspicious” and “lack
of emotional depth”), as well as preliminary support for the discriminant and convergent
validity of the CAPP–IRS. Namely, higher severity ratings of CAPP–IRS symptoms were
given to patients with a diagnosis of a Cluster B personality disorder (see Corrado et al.,
2006; Corrado et al., 2007; McCormick et al., 2008; Strub, Kreis, & Hart, 2008;
Watkinson, Corrado, Lussier, Hart, & Cohen, 2007). Finally, a recent study with forensic
psychiatric patients in Denmark revealed support for the predictive accuracy of the
CAPP in terms of violent recidivism, as well as for its concurrent validity with the PCL:SV
in that regard (Pedersen, Kunz, Rasmussen, & Elsass, 2010).
11
sources of information. Further, psychopathic individuals also often lack insight into their
functioning, and may be simply unaware of the extent and scope of their psychological
problems (i.e., not realizing the impact of their actions on others). Therefore, external
observers might be better at providing an accurate account of the functioning of a
psychopathic individual. Furthermore, given their inability to experience certain
emotional states, for instance guilt or empathy, psychopathic individuals may not be able
to report the lack of such emotions. Alternatively, if they report experiencing emotions,
their self-report may be confounded due to their lack of understanding of their true
emotions (e.g., reporting feeling remorseful when in fact they regret being incarcerated).
Another major criticism of self-report measures of psychopathy pertains to the fact that
many of those instruments are saturated with items assessing negative emotionality, yet,
negative emotionality should be fairly independent of psychopathy. As a result, the
discriminant validity of such instruments is lower, making it difficult to distinguish
psychopathy from other conditions that are marked by antisocial behavior.
Lilienfeld and Fowler (2006) also provide an overview of the advantages of self-
report measures of psychopathy. First, self-reports may be particularly useful in
providing information on subjective emotional states and traits. Specifically, emotional
states and traits are best known by the person experiencing them; all others must rely on
information directly provided to them by the respondent, or on behavioral inferences of
the emotional states or traits of the respondent based on observations and interactions
with them, which are ultimately prone to subjective bias. Thus, the quality of the
information captured via self-reports, assuming it is accurate, might be superior to that
based on interview-based measures. Further, even though psychopathic individuals are
likely unreliable in reporting the lack of emotions such as guilt, empathy, and fear (as
they may be simply unaware of this deficit), they may experience anger and alienation
more frequently than those who do not show psychopathic features, and may be capable
of providing reliable accounts of such emotions.
Self-reports are also quite economical (Lilienfeld & Fowler, 2006). They are often
brief and relatively easy to complete, and do not require any specialized training in order
to be administered. Further, when response style indicators are incorporated in the self-
report measures, they can assess response styles systematically, which is not the case
for interview based measures. Assessing response styles is quite relevant when
12
working with highly psychopathic individuals given engaging in pathological lying and
being conning and manipulative are among the core characteristics of the disorder
(Lilienfeld & Fowler, 2006). In fact, a recent meta-analytic review of the literature on the
associations between self-reported psychopathy and distorted response styles revealed
support for the validity of self-report psychopathy measures, thus tapering concerns
regarding positive response bias associated with self-report measures (Ray et al., 2012).
Finally, interrater reliability has no relevance for self-reports as they are completed by
respondents and require no judgment calls by interviewers or observers (Lilienfeld &
Fowler, 2006). This is particularly relevant here, as the assessment of psychopathy
includes the evaluation of various affective and emotional states. This in turn requires
significant clinical inference on the part of the interviewers, which is often quite
challenging and likely to negatively affect inter-rater reliability. As Lilienfeld and Fowler
(2006) point out “because validity is limited by the square root of reliability (Meehl, 1986),
the subjectivity inherent to interview-based measures will, ceteris paribus, constrain their
validity” (pg. 108). Lastly, it is important to highlight that the information obtained by self-
report measures, if provided accurately, might overlap substantially, or even fully
correspond to information obtained through an interview, given the domains of
functioning assessed by both types of measures (i.e., the self-report measure and the
interview) are adequately represented in both (see Jones & Miller, 2012, and Miller,
Jones & Lynam, 2011, regarding the convergence between self and informant reports of
psychopathy).
Nevertheless, some of the other self-report measures used for the assessment of
psychopathy (e.g., the Psychopathic Deviate scale of the Minnesota Multiphasic
Personality Inventory–2 [MMPI–2; Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer,
1989], and the Socialization scale of the California Psychological Inventory [CPI; Gough
& Bradley, 1996]) have been found to exhibit low or modest intercorrelation, indicating
they are likely not measuring the same construct (Lilienfeld & Fowler, 2006). Another
issue with the self-report assessment of psychopathy pertains to the fact that due to
method covariance (i.e., sole reliance on self-reports), the correlations between
psychopathy questionnaires may be inflated. Finally, many of the measures most
commonly used to assess psychopathy (i.e., Psychopathic Deviate scale of the MMPI–2,
Socialization scale of the CPI, Antisocial scale of the Millon Clinical Multiaxial Inventory–
II [MCMI–II; Millon, 1987], and Antisocial scale of the Personality Assessment Inventory
[PAI; Morey, 1991]) appear to measure non-specific behavioral deviance, rather than
core interpersonal or affective features of psychopathy, as they relate preferentially to
Factor 2 of the PCL–R (Lilienfeld & Fowler, 2006). In light of these limitations, the use of
general measures of personality for the assessment of psychopathy, even though
efficient in terms of time and resources, does not appear to be optimal.
14
Levenson Primary and Secondary Psychopathy Scales (LPSP)
15
Self-Report Psychopathy Scale (SRP)
The SRP is a 29-item scale developed by Hare and colleagues to resemble the
factor structure of the PCL (see Hare, 1985). The original version of the measure
correlated only modesty with the PCL, and did not provide sufficient content coverage for
some of the core personality characteristics of psychopathy (i.e., callousness,
dishonesty, superficial charm). As a result, Hare, Harpur, and Hemphill (1989)
developed a 60-item updated version of the instrument, the SRP–II. Items are scored on
a 5-point Likert scale, ranging from 1 = Disagree Strongly to 5 = Agree Strongly. The
SRP–II contains two factors resembling the factor structure of the PCL–R – the first
factor captures personality and affective aspects of psychopathy, while the second one
focuses on impulsivity and antisocial lifestyle. Nevertheless, confirmatory factor analysis
revealed that the best fitting model for the measure was the original four-factor model
proposed by the authors (Lester, Salekin, & Sellbom, 2013). Yet, others have reported
that the factor structure of the SRP-II does not parallel that of the PCL-R based on
exploratory factor analyses in non-forensic samples (Williams & Paulhus, 2004). Taken
together, these findings highlight the need for further exploration. Further, a 31-item
abridged version of the instrument has been used in research settings (e.g., Paulhus &
Williams, 2002; Salekin, Trobst, & Krioukova, 2001), though questions have been raised
as to whether it adequately captures the structure of the PCL–R (see Lester et al.,
2013).
The internal consistency and construct validity of the SRP–II have been
evaluated with various samples and appear to be promising (see Lester et al., 2013;
Derefinko & Lynam, 2006; Williams & Paulhus, 2004). There is support for the
convergent validity of the SRP–II with respect to self-report measures of narcissism,
social deviance, as well as reversed conscientiousness, agreeableness, anxiety, and
empathy (see Lester et al., 2013; Lilienfeld & Fowler, 2006; Zágon & Jackson, 1994).
Further, the total score of the SRP–II correlated positively with other self-report
measures used to assess psychopathy (e.g., MMPI–2 Antisocial Practices content scale,
the Antisocial scale of the PAI, PPI,and the Five Factor Psychopathy Index [FFP, Miller,
Lynam, Widiger, & Leukefeld, 2001]; see Derefinko & Lynam, 2006; Lilienfeld & Fowler,
2006; Salekin et al., 2001), as well as with conceptually relevant correlates such as
maladaptive attachment, riskier driving attitudes and behavior, and engaging in
16
antisocial conduct (Lester et al., 2013). However, the discriminant validity of the SRP–II
with respect to generalized criminality requires further clarification (see Lilienfeld &
Fowler, 2006).
Finally, the SRP–II was further revised to its most recent version, the SPR–III
(Paulhus, Neumann, & Hare, in press). As part of the revision process exploratory and
confirmatory factor analyses were employed in two different samples to ensure that the
factor structure of an earlier version of the SPR-III adequately captured the four factors
of psychopathy identified in the PCL–R (Williams, Paulhus, & Hare, 2007). Results
revealed four distinct but intercorrelated factors, similar to those of the PCL–R (see also
Mahmut, Menictas, Stevenson, & Homewood, 2011; Neal & Sellbom, 2012). Further,
evaluation of the psychometric properties of the SRP–III revealed appropriate construct
validity with related personality or psychopathy measures (Neal & Sellbom, 2012; ).
Good convergent and discriminant validity for the instrument have also been reported in
a community sample (Mahmut et al., 2011). In addition, an acceptable level of internal
consistency has been obtained in a college sample, where the measure has also
evidenced promising criterion-related, convergent, and discriminant validity with regard
to predicting scores on conceptually relevant external criteria (e.g., drug use, thrill
seeking, aggression, irresponsibility, impulsiveness, fraud and theft, callous affect, as
well as lack of dependability, honesty, empathy, emotional distress, negative
emotionality, social avoidance, and shyness) (Neal & Sellbom, 2012).
17
in their samples. Finally, it is worth pointing out that Lilienfeld and Andrew’s choice to
construct a self-report measure (as opposed to an interview-based instrument) was
driven by the desire to develop a measure suitable for large-scale studies outside of
prison settings.
The PPI contains 187 items, scored on a 4-point Likert scale (i.e., 1 = False to 4
= True). It yields a total score representing global psychopathy, as well as eight lower-
order facet scores—Social Potency, Fearlessness, Stress Immunity, Machiavellian
Egocentricity, Impulsive Noncomformity, Blame Externalization, Carefree
Nonplanfulness, and Coldheartedness. Three validity scales were also incorporated in
the PPI and were intended to detect response styles known to be common among
psychopathic individuals (i.e., positive impression management, malingering, and
careless or random responding).
Studies on the factor structure of the PPI have revealed two factors: Factor I (i.e.,
Fearless Dominance [FD]) included Social Potency, Fearlessness, and Stress Immunity,
while Machiavellian Egocentricity, Impulsive Noncomformity, Blame Externalization, and
Carefree Nonplanfulness loaded on Factor II (i.e., Impulsive Antisociality [IA], which was
18
subsequently labeled Self-Centered Impulsivity [SCI]; see Benning, Patrick, Hicks,
Blonigen, & Krueger, 2003). Of note, Coldheartedness did not load to either factor.
Factor I and II of the PPI appear to be orthogonal, which is different than the PCL–R,
where the two factors are moderately correlated (Lilienfeld & Fowler, 2006; Marcus,
Fulton, & Edens, 2012). This suggests that the PPI does not assess psychopathy as a
unitary construct.
Discrepant findings about the factor structure of the PPI have been reported in an
incarcerated male sample. Specifically, Neumann, Malterer and Newman (2008) were
unable to replicate the two-factor solution of Benning et al., 2003, using exploratory and
confirmatory factor analyses. Instead they found evidence for a three-factor solution,
where Impulsive Noncomformity, Blame Externalization, Machiavellian Egocentricity,
and Fearlessness loaded on one factor, Carefree Nonplanfulness and Stress Immunity
loaded on a second factor, and Coldheartedness and Carefree Nonplanfulness loaded
together on a third factor.
The psychometric properties of the PPI have been examined in a few non-
criminal samples and there is support for its high internal consistency, test-retest
reliability, as well as convergent validity with other measures of psychopathy and
antisocial behavior, or theoretically related concepts such as empathy, sensation
seeking or driving anger (see Blonigen, Carlson, Krueger, & Patrick, 2003; Lilienfeld,
1990; Lilienfeld & Andrews, 1996; Woltil, 2011;). Further, there is evidence that the PPI
can be useful in assessing APD as much as psychopathy, which raises questions about
its discriminant validity from general measures of social deviancy (see Lilienfeld &
Fowler, 2006). Despite that, there is support for the discriminant validity of the PPI from
measures of depression, negative emotionality, and schizotypy (Benning et al., 2003;
Lilienfeld & Andrews, 1996).
19
anxiety in a non-clinical and non-forensic sample (i.e., student sample) (Falkenbach,
Poythress, Falki, & Manchak, 2007).
Even though the PPI was developed for use with non-criminal populations, its
psychometric properties have been examined with correctional samples. Results reveal
negative association with measures of empathy and positive correlations with measures
of aggression (Sandoval, Hancock, Poythress, Edens, & Lilienfeld, 2000). Further,
meaningful associations between the PPI and some of the scales of the PAI, which are
theoretically consistent with the conceptualization of psychopathy, have also been
reported. Namely, the PPI correlated positively with the Antisocial, Aggression,
Dominance, and Borderline scales of the PAI (Edens, Poythress, & Watkins, 2001).
Positive correlations between the PPI and physical or nonaggressive disciplinary
infractions were also reported in the same study. Furthermore, results from a study on
the criterion-related validity and the relationship of the PPI with the Behavior Activation
System (BAS), Behavior Inhibition System (BIS), and the Five-Factor Model (FFM) of
personality revealed that Factor I (i.e., FD) was related to low BIS, high BAS, expert
prototype psychopathy scores, and primary psychopathy (Ross, Benning, Patrick,
Thompson, & Thurston, 2009). Factor II (i.e., IA or SCI) was associated with high BAS
and other psychopathy measures. Further, high Extraversion and Openness, and low
Agreeableness and Neuroticism predicted Factor I, while high Neuroticism and low
Agreeableness and Conscientiousness predicted Factor II. The authors pointed out that
even though low Agreeableness was related to both factors, it was considered to be a
function of the manifestation of psychopathy, given the two factors relate differentially
with other FFM traits (see Ross et al., 2009).
Further, the association between the PPI and the PCL–R was examined in a
sample of 50 male offenders, yielding moderately high correlations between the
measures (Poythress, Edens, & Lilienfeld, 1998). Most importantly the correlations were
higher with PCL–R Factor 1 (r =.54) than with Factor 2 (r = .40), which may suggest that
the PPI is the first self-repot measure to correlate substantially with the core
characteristics of psychopathy, at least as measured by the PCL–R (Lilienfeld & Fowler,
2006). Of note, those findings were not replicated in a study conducted by Lilienfeld and
Skeem (2004), where the associations with Factor 2 of the PCL–R were stronger than
those of Factor 1. Similar results were reported by Malterer, Lilienfeld, Neumann, and
20
Newman (2010), which raises questions regarding the convergent validity of the PPI
factors with the PCL–R/SV. In addition, the concurrent validity of the PPI relative to the
PCL:SV and its associations with past violence were examined in a sample of insanity
acquittees (Kruh et al., 2005). Results revealed significant correlations between the PPI
total, and several subscale scores (i.e., Social Potency, Machiavellian Egocentricity, and
Impulse Noncomformity), and the total and factor scores of the PCL:SV, yielding support
for the concurrent validity of the PPI in this sample. In the same study, the postdictive
utility of the PPI was comparable to that of the PCL:SV with regard to past violence.
Furthermore, Poythress et al. (2010) compared the PPI and the LPSP in a large
offender sample relative to the PCL–R. They found not only better convergent and
discriminant validity of the PPI, but results also revealed that the PPI had more
consistent incremental validity in predicting PCL–R scores relative to the LPSP, and its
patterns of associations with measures of external criterion variables were more similar
to the patterns observed for the PCL–R. The concurrent and discriminant validity of the
PPI were also examined in a male sample of maximum security federal inmates
(Rosner, 2004). Specifically, the PPI’s relationship to corresponding scales on the PAI
and the MCMI–III was evaluated, and there was support for the convergent validity of the
PPI as a measure of psychopathic personality features. Nevertheless, results were
inconsistent with regard to its discriminant validity from measures of schizophrenia-
spectrum disorder and mania.
The psychometric properties of the PPI have also been examined in female
samples. For instance, the PPI was negatively associated with the CPI Socialization
scale, and positively with the PAI Antisocial scale in a female correctional sample
(Chapman, Gremore, & Farmer, 2003). These findings provide support for the
convergent validity of the PPI, but raise questions about its discriminant validity as the
Antisocial scale of the PAI measures primarily antisocial behavior. The findings are
similar to those obtained from an undergraduate female sample (Hamburger, Lilienfeld,
& Hogben, 1996), which raises questions about the criterion validity of the instrument.
However, Berardino, Meloy, Sherman, and Jacobs (2005) found support for the
concurrent and discriminant validity of the PPI in a sample of incarcerated women, and
concluded that the PPI assesses the antisocial as well as the interpersonal-affect facets
of psychopathy.
21
Further, Marcus et al. (2012) conducted a meta-analytic review of studies
examining the theoretically important correlates of the two PPI factors. The authors
found that each factor predicted distinct correlates, which were sometimes opposite from
one another. Particularly, while SCI was highly correlated with PCL–R (Factor 2) and
LPSP, FD was modestly correlated (at best) with PCL–R (Factor 1), and the Primary
Psychopathy Scale of the LPSP. In addition, while the SCI was associated with
measures of antisocial personality, the FD was not. Instead, FD was correlated with
positive traits and emotions, negatively related with neuroticism, and unrelated with
constraint. Finally, results revealed that the FD factor was not highly correlated with
most other measures of psychopathy, which raised questions regarding the role of FD in
the conceptualization of psychopathy. Specifically, they found that FD was not
particularly maladaptive, and those who score high on FD are unlikely to appear
especially pathological. These findings were interpreted as a limitation of the PPI and
it’s two factor model by Neumann, Uzieblo, Crombez, and Hare (2013). Similar findings
were reported in another meta-analytic study (Miller & Lynam, 2012). Specifically, the
authors reported strong convergent associations between the SCI factor and other
psychopathy measures (i.e., PCL/PCL–R, LSRP, and SRP), as well as expected
correlations between SCI factor scores and theoretically important correlates (e.g.,
interpersonal antagonism, empathy, impulsivity, APD). Conversely, they expressed
skepticism that high scores on the FD factor alone could be used to identify individuals
as psychopathic. Taken together these findings raise the question of whether PPI-
assessed psychopathy is a unitary construct or whether it is comprised of “relatively
independent lower-order facets”, which has been previously suggested as a possibility
by Lilienfeld and Widows (2005; see Marcus et al., 2012).
Notably, Marcus et al.’s (2012) meta-analysis has been criticized for excluding
studies that have quantified the PPI factors by including personality variables (e.g.,
subscales of the Multidimensional Personality Questionnaire [MPQ; Tellegen, 1982]),
and other psychopathy measures (Patrick, Venables, & Drislane, 2012). The meta-
analysis has also been criticized for categorizing personality-criterion variables around
broad personality dimensions (i.e., Positive Emotionality, Negative Emotionality, and
Constraint), as such an approach may attenuate or obscure significant associations
between the PPI factors and relevant correlates (Patrick et al., 2012). Further, in
response to Marcus et al.’s meta-analysis, Blonigen (2013) emphasized the theoretical
22
and conceptual relevance of the FD factor to psychopathy, which did not appear to be
highlighted by Marcus et al. Finally, Benning (2013) offered a conceptual explanation for
the heterogeneity of effect sizes reported in the meta-analysis, and along with Lilienfeld
(2013) suggested that the results from the meta-analytic study reflect the nature of
psychopathy, and are not indicative of psychometric limitations of the PPI or
methodological issues of the study.
Finally, the predictive validity of the PPI has been examined in correctional
samples, and results provide support for the predictive utility of the instrument.
Specifically, the PPI total and factor scores were predictive of the number of infractions,
and Factor I was more strongly associated with nonaggressive misconduct, while Factor
II was more strongly correlated with aggressive misconduct (Edens et al., 2008).
Support for the predictive utility of the PPI with regard to the veracity of drug reports was
also evident in a sample of adult males seeking admission to an inpatient drug treatment
facility (Abrams, 2005). Overall, its good psychometric properties suggest that the utility
of the PPI in criminal justice settings is promising.
23
brevity and efficiency in research settings appeared to be the primary benefits
associated with its use (Kastner, Sellbom, & Lilienfeld, 2012). However, Smith,
McCarthy and Anderson (2000, as cited in Smith, Edens, & Vaughn, 2011), warned
against a potential construct underrepresentation and possibly an atypical factor
structure due to the way the measure was constructed. In fact, Wilson et al. (1999)
found that Impulsive Nonconformity and Coldheartedness loaded on Factor I, while
Factor II was comprised of the original subscales with the exception of Impulsive
Nonconformity. Further, the psychometric properties of the PPI:SF were evaluated
against the full length PPI in a college and a prison sample, and results revealed
comparable discriminant validity for both instruments (Kastner et al., 2012).
Nevertheless, the construct validity and internal consistency of the PPI:SF were found to
be lower than that of the PPI, and the PPI:SF also evidenced weaker correlations with
conceptually relevant criteria (e.g., egocentricity, aggressiveness, and aggression).
Concerns about the internal consistency of several of the PPI:SF subscales (e.g.,
Carefree Nonplanfulness, Impulsive Nonconformity, and Social Potency) were also
raised by Smith et al. (2011), who evaluated the measure in three samples:
Undergraduate, foster care, and juvenile justice. Yet, the expected associations of the
PPI:SF scales and external correlates (e.g., egocentricity, hypochondriasis, deceptive
practices, extraversion, depressive symptoms, hostility, etc.) were found to be generally
consistent with prior studies evaluating the PPI. The authors concluded that their
findings provide some evidence that the subscales combined to create Factor I and
Factor II of the PPI:SF are tapping distinct constructs (Smith et al., 2011).
The PPI was recently revised (Lilienfeld & Widows, 2005) in order to lower its
reading level and remove items with questionable psychometric properties, or items that
were culturally specific. It now contains 154 items and its structure is essentially
identical to that of the PPI. It produces a total score, as well as eight content scores
(i.e., Machiavellian Egocentricity, Social Influence, Coldheartedness, Carefree
Nonplanfulness, Fearlessness, Blame Externalization, Rebellious Nonconformity, and
Stress Immunity), and three factor scores (i.e., Self-Centered Impulsivity, Fearless
Dominance, and Coldheartedness, the latter of which is the same as the content score).
This version of the instrument contains four validity scales: Virtuous Responding (VR),
24
Deviant Responding (DR), and two Inconsistent Responding (IR) scales (i.e., a 15-item
or a 40-item scale). Items are scored on a 4-point Likert scale (i.e., False, Mostly False,
Mostly True, or True). Notably, support for the need and utility of the VR and DR validity
scales of the PPI-R has been shown in a recent study, where the scales evidenced utility
in differentiating between dissimulators and genuine test-takers, and mean elevations
and the psychometric validity of the scale scores appeared to be grossly affected by
response bias (Anderson, Sellbom, Wygant, & Edens, 2012).
Finally, the construct validity of the PPI–R was examined in a sample of forensic
psychiatric patients, and results revealed that the SCI factor predicted anger, hostility,
impulsivity, drug abuse, overall psychiatric symptoms, and violence risk (Edens &
McDermott, 2010). The FD factor predicted anger, depression, low anxiety and alcohol
abuse or dependence. Further, the association between impulsivity and the PPI–R was
examined, and results revealed positive correlations for the total score and the SCI
25
factor (Ray, Poythress, Weir, & Rickelm, 2009). Fearless Dominance, in contrast,
appeared to be strongly associated with sensation seeking, weakly associated with lack
of premeditation, and negatively with lack of perseverance. Given the PPI–R is a
relatively new measure, there were no published studies on its predictive validity with
regard to recidivism as of the writing of this manuscript.
While many studies have evaluated the relationship between psychopathy and
crime or violence, the incremental validity of psychopathy over and above other
important risk factors such as demographic variables, criminal history, substance abuse
or mental disorders, remains under-investigated. Some important findings have
emerged, however, based on the available research. For instance, in their review of
existing meta-analyses on psychopathy and criminal recidivism, Douglas et al. (2006)
reported that there is support for the predictive incremental utility of psychopathy as
measured by the PCL–R vis-à-vis crime and violence, and that appears to be the case
even after controlling for factors such as substance abuse, criminal history, and
demographics. Along the same lines, Hemphill, Hare, et al. (1998) and Hemphill,
Templeman, et al. (1998) reported that the PCL–R made significant contributions toward
predicting recidivism not only beyond demographic characteristics and criminal history,
but also beyond personality disorder diagnoses. Similar results were also reported in a
more recent meta-analysis conducted with male and female juvenile offenders, where
psychopathy was found to be significantly associated with general and violent recidivism
even after covariates of psychopathy were controlled for (Edens, Campbell, & Weir,
2007; Olver, Stockdale, & Wormith, 2009). That was also the case in a meta-analysis
conducted with adults (Campbell et al., 2009).
26
and informant-reported psychopathy scores were related to EBs, and provided some
degree of incremental validity. Specifically, they found that self-reported psychopathy
scores were somewhat more useful than informant reports of psychopathy as they
generally provided more unique information regarding EBs, and in particular substance
abuse and gambling, while informant-reported psychopathy proved useful in predicting
intimate partner violence.
Taken together these findings highlight the need for further research on this
topic. Given the robust association between psychopathy and social deviance or
criminality, it is reasonable to assume that measures of psychopathy should be preferred
over general measures of personality when faced with the task of estimating risk for
27
future offending. Nevertheless, there is not enough empirical evidence to support that
assumption.
Forouzan and Cooke (2005) pointed out that gender differences in psychopathy
are also notable with regard to the degree of symptom severity, as well as the meaning
of particular behavioral indicators of the disorder. Specifically, the authors indicated that
interpersonal characteristics such as glibness, superficial charm, and grandiose self-
worth among women are only apparent in extreme cases. In addition, the authors noted
29
that indicators of psychopathy, such as promiscuous sexual behavior, might have
different meaning for males versus females. Specifically, for females it might be
underpinned by a desire to exploit, and therefore, could be used as a manipulation tactic
to obtain financial, social, or narcissistic gain, and may be reflective of a parasitic or
impersonal lifestyle. Conversely, for men it may be underpinned by sensation seeking or
may reflect mating effort. Social norms may also affect how psychopathic traits are
interpreted across gender. For instance, while financial dependency may be socially
acceptable for females, it is looked down upon for males, and it is considered to be
parasitic (Forouzan & Cooke, 2005; Logan & Weizmann-Henelius, 2012). In addition,
females with psychopathic traits may rely on different tactics than psychopathic males to
achieve the same goals. For instance, given their lesser upper body strength, physical
aggression is less likely to yield the same results for women as for men, and therefore
women may resort to manipulation, flirtation, or coercion to reach their goals (Nicholls &
Petrila, 2005). As a result of these differences, psychopathy among women may not be
captured by assessment instruments designed for men.
Even though the etiology of psychopathy is not fully understood, research thus
far supports the notion that a number of factors (e.g., physiological, social, cultural,
environmental, etc.) come into play in the shaping of gender differences in general, and
psychopathy in particular (for a discussion, see Kreis, 2009; Nicholls et al., 2007; Verona
& Vitale, 2006). For instance, hormonal and neurochemical differences such as levels of
androgens and dysregulated serotonin (5-hydroxytryptamine [5-HT]) have been linked
with differential prevalence on antisocial behavior and aggression across gender
(Verona & Vitale, 2006). Specifically, abnormally high levels of androgens in girls seem
to be linked to male-specific behavior during childhood (i.e., girls tend to be tomboys),
which may be overridden by socialization and other developmental influences over time.
In addition, male hormones have been linked to aggression and other types of antisocial
behavior in both men and women, even though findings have not been entirely
consistent (see Kreis, 2009; Nicholls et al., 2007; Verona & Vitale, 2006). Also, while
lower levels of 5-HT have been associated with aggression and impulsivity in men (i.e.,
externalizing psychopathology), they have been linked with mood disorders such as
anxiety and depression in women (i.e., internalizing psychopathology; Verona & Vitale,
2006).
30
Further, developmental socialization has also been highlighted as a reason for
gender differences in the manifestation of psychopathy. For instance, assuming strong
and dominant roles, as well as having high self-regard and mastery, is typically accepted
and expected of men rather than women (Nicholls et al., 2007; Perri & Lichtenwald,
2010). In a similar fashion, culture may influence the degree of gender differences in the
expression of personality traits in general, as well as psychopathy (Kreis, 2009; Nicholls
et al., 2007). Specifically, cultural factors may play a role through modeling the
manifestation of pathologies and by influencing the frequency of occurrence of
psychopathology. They may also be influential in terms of shaping specific ways of
coping with problems, which typically differ across gender. In addition, cultural norms
may dictate whether specific conditions or behaviors are seen as pathological, and may
shape people’s beliefs and reactions to different disorders (Kreis, 2009).
As proposed by Kreis (2009) and Kreis and Cooke (2011), the differences in the
symptoms of psychopathy across gender also make sense theoretically from an
evolutionary psychological theory perspective (Campbell, 2002 as cited in Kreis, 2009),
which postulates that males and females differ with regard to domains relevant to
reproductive success. Namely, these domains include aggression, dominance,
emotionality, interpersonal skills, and risk-taking. According to Triver (1972) those
differences are due to the different reproductive strategies used by men versus women,
which are closely linked to the differential parental investment needed across gender to
ensure reproductive success (as cited in Pratto, Sidanius, & Levin, 2006). Specifically,
31
as men can father many children with a minimal biological investment, they should prefer
short-term mating relations with multiple fertile partners. In addition, considering men
are unlikely to be interested in investing in children who are not theirs, they tend to be
more jealous than women, find ways to protect women’s sexuality from other men, and
at times may expropriate other men’s resources for their own families. In contrast, as
women invest more of their time and resources to ensure survival of their offspring, they
should prefer a long-term mating strategy with a partner who is stable and could provide
resources. As a result, women tend to be more selective of their mates and the mating
conditions than men (Kreis, 2009; Pratto et al., 2006). In light of this, gender differences
in psychopathic characteristics as described at the beginning of this section appear
fitting within the framework of evolutionary psychological theory given the roles of males
and females in the reproductive process (for a discussion, see Kreis, 2009).
32
1.5.3. Differences in Behavioral Manifestation, Clinical
Presentation, and Comorbidity
Despite some mixed findings, it has been hypothesized that the behavioral
manifestation of psychopathy varies across gender, where a more antisocial pattern has
been noted among males, and more of a histrionic pattern among females (compare and
contrast Cloninger, 1978; Lilienfeld, Van Valkenburg, Larntz, & Akiskal, 1986; Marsee,
Silverthorn & Frick, 2005; and Czar, Dahlen, Bullock, & Nicholson, 2011; Schmeelk,
Sylvers, & Lilienfeld, 2008). Specifically, males tend to be more likely to express
psychopathy through physical aggression or by engaging in conning behavior (Bobadilla,
Wampler, & Taylor, 2012; Forouzan & Cooke, 2005; Logan & Weizmann-Henelius,
2012; Strand & Belfrage, 2005; Verona & Vitale, 2006). In contrast, psychopathy among
females is more likely (though not exclusively) to be manifested via relational aggression
(i.e., behaviors intended to harm others’ relationships such as back-stabbing, gossiping,
rumor spreading, lying, and deceitfulness), as females have been found to be
significantly more relationally aggressive than males starting in childhood (see Bobadilla
et al., 2012; Crick & Grotpeter, 1995; Logan, 2009; Logan & Weizmann-Henelius, 2012;
Marsee et al., 2005). It is important to highlight that results have been somewhat
different in high risk samples, where females have been found to be both relationally and
physically aggressive, while males appeared to be more physically than relationally
aggressive (Penney & Moretti, 2007; Stickle, Marini, & Thomas, 2012). Even so, these
findings provide support for the notion of gender differences in the manifestation of
psychopathy, with relational aggression being more typical of females, and physical
aggression being more characteristic of males.
33
Personality Disorder (APD), which captures the behavioral aspects of psychopathy, were
more likely to be irresponsible as parents, to be physically violent toward their children or
partners, and to engage in prostitution. Such tendencies are often considered a violation
of their societal role as females (e.g., shallow expression of nurturance, lack of
selflessness; see Verona & Vitale, 2006). As a result of these gender differences,
female antisocial behavior might be less likely to be reflected on official criminal records,
and existing assessment instruments may not capture aggression demonstrated by
them.
Results from studies examining the base rates of psychopathy, APD and
Conduct Disorder (CD) among women are also somewhat inconsistent. While some
studies with correctional samples using observer-based measures (e.g., PCL, PCL–R
and PCL:SV) reported overlapping base rates for females and males (e.g., Louth, Hare,
& Linden, 1998; Strand & Belfrage, 2001; Strachan, 1993; Tien, Lamb, Bond, Gillstrom,
& Paris, 1993), others have reported substantially lower base rates for females (Dolan &
Völlm, 2009; Goldstein et al., 1996; Loucks, 1995; Neary, 1990; Nicholls et al., 2005;
Salekin, Rogers, & Sewell, 1997; Vitale, Smith, Brinkley, & Newman, 2002; Warren et
al., 2003) relative to the typical base rates reported for males (i.e. 15%–30% for males
vs. 9%–23% for females; Vitale et al., 2002). Further, findings from studies conducted
with non-criminal samples also differ—while some report higher mean scores for males
versus females (e.g., Weiler & Widom, 1996; Forth, Brown, Hart, & Hare, 1996), others
have found no gender difference in PCL–R scores (e.g., Cooney, Kadden, & Litt, 1990;
Stafford & Cornell, 2003).
35
1.5.4. Differences with Regard to Personality and Behavioral
Correlates, and Recidivism
Gender differences have also been reported with regard to the personality and
behavioral correlates of psychopathy. Specifically, personality correlates such as
experience of emotion, empathy, lack of attachment, self-control and socialization,
interpersonal dominance, as well as arrogance, calculation and narcissism, are relatively
consistent across gender (e.g., Forth et al., 1996; Kreis, 2009; Miller, Watts, et al., 2011;
O’Connor, 2003; Patrick, 1994; Rutherford, Cacciola, Alterman, & McKay, 1996;
Salekin, Rogers, Ustad, & Sewell, 1998; Strachan, 1993; Vitale et al., 2002; Verona,
Patrick & Joiner, 2001; Zágon & Jackson, 1994). Some exceptions have been noted by
Miller, Watts, et al. (2011), who found that the antisocial factor of psychopathy was
correlated more strongly with impulse-related tendencies among males, and with
openness to experience among females. In addition, Kreis (2009) pointed out that
women may present as less dominant and may engage in more impression
management than their male counterparts.
36
scores were not predictive of future offending among adolescent females (Odgers et al.,
2005; Vincent, Odgers, McCormick, & Corrado, 2008), and even though Odgers et al.
(2005) reported that there was an association between deficient affective experiences
and aggression, that effect was negated once victimization experiences were taken into
account.
Contextual factors related to the violence being committed were found to vary
across gender as well. Specifically, men were more likely than women to have been
engaging in substance use and less likely to have been adhering to prescribed
psychotropic medication prior to engaging in acts of violence (Robbins et al., 2003).
Further, males have been reported to show more anger, regardless of the context, while
females tend to experience a wider range of negative affects such as anger, sadness,
fear, or guilt more frequently and more intensely (see Forouzan & Cooke, 2005). As
pointed out by Skeem et al. (2011), one possible explanation for such findings is
reflected in the multifactorial threshold models (Cloninger, Christiansen, Reich, &
Gottesman, 1978) according to which for biological reasons (i.e., lower testosterone), as
well as social norms prohibiting against over-aggression in females, a greater diathesis
and associated traits are required for females to manifest psychopathy. Therefore, when
37
psychopathy is detected among females with assessment measures designed for males,
we might be dealing with individuals who have particularly heightened predisposition for
disinhibited behavior (Skeem et al., 2011).
Further, Sevecke, Pukrop, Kosson, and Krischer (2009) used the PCL:YV in a
sample of German adolescents and found satisfactory fit for the three-factor model for
males, but none of the models (i.e., two-, three- , or four-factor) were satisfactory for
females. Likewise, O’Connor (2003) did not find support for the goodness of fit of the
two-factor model of psychopathy among females. Nevertheless, other authors have
reported support for the three- or four-factor models of psychopathy across gender. For
instance, Warren et al. (2003) examined the goodness of fit of the two-, three-, and four-
factor model of psychopathy in a female prison sample, and found the best fit of the data
was achieved with the three-factor model, although the four-factor presented a good fit
as well. Similarly, Dolan and Völlm (2009), Strand and Belfrage (2005), and Weizmann-
Henelius et al. (2010) showed support for the three-factor model of the PCL–R/PCL:SV
38
in female populations. The goodness of fit of the data for the three-factor model using
the PCL:SV was also confirmed in a study by Skeem, Mulvey, & Grisso (2003), using a
mixed gender sample, who reported that their findings were consistent across gender.
Further, there is a lack of consensus as to whether the raw scores of the PCL–
R/SV reflect the same level of psychopathy across gender. As a result, some studies
have used traditional cut-off scores on the PCL–R (i.e., score of 30), while others have
lowered the score. This has led to significant variations of base rates in psychopathy
among women across samples (see Jackson, Rogers, Neumann, & Lambert, 2002;
39
Warren et al., 2003), and calls for further research and caution in interpreting the
available literature (Falkenbach, 2008; Forouzan & Cooke, 2005; Logan, 2009).
Similar to the PCL–R, results from studies on the CAPP–IRS indicated that it
may capture psychopathy in women, although some adjustments may be necessary
(see Kreis & Cooke, 2008). Specifically, Kreis & Cooke (2011) reported support for the
content validity of the CAPP–IRS across gender, and suggested that at symptom and
domain levels of the CAPP, psychopathic men and women share key similarities, but
there are also important gender differences, which were likely a matter of degree. With
regard to gender differences, men were rated as higher in the Attachment, Dominance,
Behavioral, and Cognitive domains, and at the symptom level, “self-aggrandizing,”
“sense of invulnerability,” “self-centered,” “domineering,” “reckless,” “disruptive,”
“aggressive,” “lacks anxiety,” and “unempathic” were more relevant to psychopathy in
men. As for women, symptoms such as “manipulative,” “lacks emotional stability,” and
“unstable self-concept” were found to be more relevant to psychopathy. The authors
concluded that although some symptoms or domains were found to be more prominent
in psychopathic men versus psychopathic women (and vice versa), they were not
specific to psychopathy in only one gender.
All in all, this suggests that although gender differences should not be ignored in
the assessment of psychopathy, existing measures are likely to have at least some utility
in assessing psychopathy in women, although some modification of the measures may
be necessary to ensure assessment accuracy. Particular caution in the interpretation of
findings would be needed in areas where gender differences are likely to be detected.
As noted earlier, from evolutionary psychology theory perspective gender differences
would only be expected in domains relevant to successful reproduction such as
aggression, dominance, risk-taking, interpersonal skills, and emotionality (Kreis, 2009;
Kreis & Cooke, 2011). Gender differences in the manifestation of psychopathy would
also be expected in those same domains from social dominance theory (i.e., aggression
and dominance domains; Sidanius & Pratto, 1999) and relational theory perspectives
(i.e., risk-taking, interpersonal skills, and emotionality domains; Miller, 1976, as cited in
Convington, 2007). Specifically, in consideration of their respective roles in the
reproductive process, in comparison to men, women would be expected to be less
dominant and less risk-taking. In addition, relative to men, women would be less
40
physically aggressive, but rather relationally aggressive. Nevertheless, they would be
capable of being on par with men with regard to physical aggression if need be or if they
would personally gain from it. Further, women would have especially well-developed
interpersonal skills, and would typically use them as a primary manipulative or exploitive
strategy. It is also likely that they would be less status-seeking and overtly grandiose
than men. Finally, differences in the emotional responses of men versus women would
be expected in terms of intensity, frequency, and external expression of emotions, with a
relatively higher degree for all among women, which they may use as a manipulative or
exploitive tactic (for a discussion, see Kreis, 2009). As noted by Kreis (2009) given the
etiology of psychopathy is yet to be fully understood, the explanation offered in terms of
expected gender differences is speculative in nature. Even so, the evolutionary
psychological theory, social dominance theory, and relational theory provide a useful
framework in understanding where to expect differences across gender.
41
2. Current Study
This study was designed to further the validation of the PPI–R. Establishing the
psychometric soundness of a measure that is fairly new is in and of itself crucial. In
addition, given the PPI–R was developed for use with non-criminal samples it is
important to determine whether or not it could be used with criminal populations,
especially since the base rates of psychopathy among criminals are relatively high.
Further, although there are a few studies that have examined some of the psychometric
properties of the PPI–R, there is currently no research on the predictive validity of the
measure vis-à-vis crime and violence. It is worth pointing out that the PPI was
developed with the intention of creating “a relatively ‘pure’ measure” (Lilienfeld &
Andrews, 1996, p. 491) of psychopathy, which emphasizes personality traits, rather than
behavior aspects of it. Therefore they did not include many items on criminality. They
reasoned their decision for doing so by stating:
The lack of emphasis on criminally related items in the PPI–R does not change
the fact that there is a well-established association between psychopathy and social
deviance or criminality. Thus, there is no reason to assume that as a measure designed
to assess the core traits of psychopathy the PPI–R would have no utility in identifying an
association with social deviance or criminality. Considering that this association is of
interest in criminal justice settings, as it defines the role of psychopathy in violence risk
assessment, evaluating the predictive validity of the PPI–R with regard to future crime
and violence was deemed important.
42
Examining the effects of gender on the relationship between psychopathy and
future crime or violence is also important given that the literature on psychopathy in
women, although expanding in terms of the number of studies being published, is still
rather sparse when it comes to systematic investigations on the topic. Even so, thus far
existing research and theory suggest that gender differences in psychopathy are likely to
be detected and therefore should be explored. The importance of evaluating the
incremental validity of the PPI–R becomes apparent in light of its potential applications in
practical settings. Namely, being able to assess and diagnose psychopathy properly is
very important in correctional settings. Unfortunately, measures of personality such as
the MMPI–2, CPI, and the MCMI–II have not been particularly useful in assessing
psychopathy, as they appear to assess general tendencies toward socially deviant or
criminal behavior, and are therefore low in specificity. This highlights the need to
evaluate the PPI–R as it was specifically designed to assess psychopathy. The decision
to evaluate the PPI–R as opposed another measure of psychopathy was driven by its
promising psychometric properties emerging in the literature, and also the well-
established properties of its predecessor, the PPI.
To fulfill the goals of the present study, four research questions were examined:
43
2.1.2. Research Question 2:
Predictive Validity of the PPI–R with Respect to Recidivism
and Type of the Offense(s)
It was hypothesized that results from this study will provide support for the
predictive validity of the PPI–R in terms of crime and violence following release from
custody. Specifically, in consideration of the existing literature on the relationship
between psychopathy and crime, it was expected that individuals who scored higher on
the PPI–R will be more likely to engage in crime and violence subsequent to their
release from custody. In addition, the predictive utility of the SCI and FD factors of the
PPI–R were examined separately. In accordance with prior research it was expected
that the SCI factor, which corresponds with Factor 2 of the PCL–R, will be a stronger
predictor of future crime and violence than the FD factor, which corresponds with Factor
1 of the PCL–R.
44
Johnson, 2010; Buffington-Vollum, Edens, Johnson, & Johnson, 2002; Salekin et al.,
1997; Walters & Duncan, 2005). Similarly, the Neuroticism (N), Conscientiousness (C),
and Agreeableness (A) scales of the NEO–FFI were selected for use in the analyses, as
research reveals support for the relationship between violence and low
conscientiousness and agreeableness, and high neuroticism (e.g., Caprara,
Barbaranelli, & Zimbardo,1996; Miller & Lynam, 2003; Miller, Lynam, & Leukefeld, 2003;
Skeem, Miller, Mulvey, Tiemann, & Monahan, 2005).
Even though results from existing studies on the base rates of psychopathy
across gender are inconclusive (i.e., some have reported significantly higher scores for
males while others have found no significant differences across gender), most of the
studies that have used self-report measures to assess psychopathy have reported
gender differences in scores, with males scoring higher (e.g., Lilienfeld & Hess, 2001;
Miller, Watts, et al., 2011; Wilson et al., 1999; Zágon, & Jackson, 1994). Therefore it
was expected that males participating in this study will score higher on the PPI–R than
their female counterparts.
45
Does Gender Moderate the Predictive Utility of the PPI–R with Respect to
Recidivism?
Finally, the literature is also mixed with regard to the strength of the relationship
between psychopathy and different types of recidivism (i.e., general vs. violent) for
males versus females (see Grimes, Lee, & Salekin, 2011). Nevertheless, given the
manifestation of psychopathy appears to differ across gender, where physical
aggression tends to occur more frequently among males, while relational aggression
appears to be more characteristic of females (see Robbins et al., 2003; Verona & Vitale,
2006), it was expected that males with high scores on the PPI–R will engage in more
severe offense(s) than females.
2.2. Method
2.2.1. Participants
Participants were recruited from two correctional facilities in the Greater
Vancouver area, British Columbia, Canada, where they were serving less than a two-
year sentence. Given there are relatively few studies focusing on psychopathy in
women, the goal was to have a sample that was relatively equally split in terms of
gender. Study participants were recruited if they were between the ages of 19 and 50,
able to communicate in English, had no chart diagnosis of mental retardation, and were
not exhibiting acute psychotic symptoms (e.g., delusions, hallucinations, disorganized
cognition or behavior, etc.), which could have impeded data collection. The language
46
requirement was implemented as the researchers collecting data were monolingual for
English. In addition, the age restriction was implemented in order to reduce the
possibility of having participants suffering from age-related issues (e.g., various forms of
dementia and other degenerative brain disorders), and also in consideration of the fact
that base rates of violence decline with age (i.e., a trend which has been well-noted in
criminology; see Hagan, 2011; Plass, 2007). While addressing issues of psychopathy
among the elderly is important, it is beyond the scope of this study and warrants a
separate evaluation. There were no restrictions on participation based on attributes
such as race, ethnicity, culture, religion, sexual orientation, or disability (other than
mental retardation). The goal was to recruit a total of 120 participants, equally split
between genders.
A few sources of information were used to estimate the desired sample size,
using an alpha level of .05 and power of .80, which has become an accepted level of
power needed for a worthwhile investigation (Aiken & West, 1991). First, general
guidelines for determining the minimum sample size needed to obtain reasonably stable
estimates in logistic or survival models were taken into consideration. According to
these guidelines a minimum of 10 to 15 (Babyak, 2004; Peduzzi, Concato, Feinstein, &
Holford, 1995; Peduzzi, Concato, Kemper, Holford, & Feinstein, 1996) or 20
observations/events per candidate predictor were necessary to allow good estimates
(Harrell, 1997; Harrell, 2012; see also Harrell, Lee, Califf, Pryor, & Rosati, 1984).
Next, web-based statistical calculators for a-priori sample size determination (i.e.,
Free Statistics Calculators; Soper, 2011) were used. Soper’s statistical calculators have
been used broadly and have been referenced in a number of published articles across
different disciplines (e.g., Beltran, Segura, & Bello, 2012; Dattalo, 2009;
Diseth, Danielsen, & Samdal, 2012; Jensen, Fraser, Shankardass, Epstein, & Khera,
2009; Marambe, Vermunt, & Boshuizen, 2012; Soper, Demirkan, Goul, & St. Louis,
2012; Wattal, Schuff, Mandviwalla, & Williams, 2010; Yeom & Heidrich, 2009, among
others). Based on preliminary statistical power estimations using Soper’s calculators for
multiple regression including three to four predictors or hierarchical multiple regression
analysis including two to six predictors, which were used in this study, the needed
sample size to detect moderate effect size departures from the null distribution was 76
to 84, and 55 to 79 respectively (Soper, 2011). As for detecting moderate effect size
47
differences in the PPI–R scores across gender, the needed sample was estimated to be
51 participants per group using Soper’s calculator for independent samples t-tests for a
one-tailed hypothesis. The sample size needed for correlational analysis was estimated
using G*Power software (Erdfelder, Faul, & Buchner,1996), which yielded a desired N of
82. These sample size estimations were consistent with guidelines offered by Cohen
(1992) in determining desired sample sizes for commonly used statistical analyses.
As for detecting interaction effects, according to Aiken and West (1991) the
required total sample size was 55 for moderate size departures from the null distribution,
assuming the predictors are error-free; the needed sample size would increase
substantially as the reliability of the predictors decreases. Specifically, as the reliability
of the predictors drops from 1.00 to .80, the required sample size would be slightly more
than double (Aiken & West, 1991). Per these guidelines, the reliability of the PPI–R
scores and gender were taken into consideration. Namely, while the reliability of the
PPI–R scores would be expected to be lower than 1.00, the internal consistency
coefficients reported in the literature were typically around or above .80. Of note, as
described in the Result section, alpha coefficients for the PPI–R indices used in the
moderation analyses ranged from .837 to .939 in this study. As for gender, it was
expected that it would be measured without error. Therefore, a sample of 120
participants was considered adequate for the detection of moderate interaction effects if
they were present. All in all, based on these a-priori power estimations for regression,
correlational, independent samples t-tests, and moderation analyses, the specified N
(i.e., 120) was deemed acceptable and sufficient to yield meaningful results (Cohen,
1988, p.452).
The NEO-FFI (Costa & McCrae, 1992b) is a 60-item scale developed to measure
normal adult personality based on the Five-Factor Model (FFM) of personality, which
uses five domains (listed below) to define personality and account for characterological
differences among people. The NEO–FFI is the short form of the original 240-item
NEO–Personality Inventory Revised (NEO–PI–R), assessing features of neuroticism (N),
extraversion (E), openness to experience (O), agreeableness (A), and
48
conscientiousness (C). Items are scored on a 5-point Likert scale ranging from Strongly
Disagree to Strongly Agree. Research findings confirm its good psychometric
properties. Specifically, alpha coefficients calculated in the development sample ranged
from .74 (A) to .89 (N) (average α = .80) indicating good internal consistency (Costa &
McCrae, 1992b). There is evidence for satisfactory test-retest reliability ranging from .80
(A) to .87 (O) over a six-month period, and from .73 (A) to .86 (O) over a 30-month
period (Murray, Rawlings, Allen, & Trinder, 2003; see also Botwin & Juni, 1995). In
addition, the NEO–FFI has good convergent and discriminant validity with personality
measures such as MMPI–2 and MCMI–II (Zeiger, 1996). Research indicates that the
NEO–FFI has good correspondence with the NEO–PI–R (average scale r = .87), even
though its reliability and validity have been reported to be slightly lower than the NEO–
PI–R for which domain level reliabilities ranged from .86 to .95 (Botwin & Juni, 1995;
Skeem et al., 2005).
49
Community Risk/Needs Assessment (CRNA)
50
between 71% and 83%. Support for stability of the CRNA ratings, correct classification
of offenders’ recidivism status, and its overall utility as a classification tool for community
case management was reported by Glackman & Trytten (2000). Given its utility in
assessing risk for recidivism, the CRNA was used as a provisional “benchmark” for the
performance of the PPI–R in terms of predictive validity with regard to recidivism, and
the specific ratings employed in the analyses included: (a) Summary rating, (b)
Supervision Level, (c) Need rating, and (d) Risk rating. Information regarding the
number of prior court dispositions, the number of prior supervision failures, the number
of prior failures to comply, age at first arrest, and imprisonment after a conviction were
also collected and used to establish the descriptive characteristics of the sample.
51
2.2.4. Procedure
Staff at the correctional facilities helped to identify individuals suitable for the
study. As a general practice, inmates’ mental health is evaluated for psychiatric and
substance use issues upon admission. These evaluations are completed by trained staff
(i.e., typically nurses), and individuals exhibiting symptoms are flagged in the system as
special care in terms of medications or substance abuse treatment (e.g., methadone
treatment) may be required. Individuals flagged in the system as exhibiting acute mental
health symptoms (including acute substance use-related symptoms such as drug-
induced psychosis), which could interfere with data collection, were not recruited for the
study.
The assessments took place within two provincial jails in the Lower Mainland,
British Columbia. A random sample of the inmates identified by staff as potential
participants according to the selection criteria, who agreed to be approached for study
participation, were invited to take part in the study. They were scheduled for a session,
during which they were provided with informed consent, supplemented by a description
of the study and clarification of its requirements and purpose. Limits of confidentiality
were discussed and they were notified that their standing with the correctional system
would not be affected by their decision to participate (or not participate) in the study.
Participants who agreed to participate were then interviewed and asked to complete self-
report questionnaires. Self-report measures selected for use in this study assessed
psychopathic features and behavioral patterns of psychopathy (i.e., PPI–R), as well as
various aspects of personality in general (i.e., NEO–FFI and PAI). A semi-structured
interview protocol was used to gather demographic information, as well as information
about past antisocial behavior, mental health, treatment involvement, and social history.
Demographic information along with individuals’ responses on the PPI–R, PAI, and
NEO–FFI were used to answer research questions 1, 2, and 3. In addition, participants’
criminal records were reviewed within 6 to 26 months following their release from
custody to determine whether they had engaged in criminal acts subsequent to their
index offense as a way of addressing research questions 2, 3, and 4. Participants were
provided with a $20 honorarium for their participation.
53
among correctional samples (see Edens & Ruiz, 2005). This rule was applied in the
current study, and of the total sample 89 individuals (i.e., 45 males and 44 females)
produced valid profiles. Of note, only one of the research questions (i.e., research
question 3) required the joint use of the random responding rules for the PPI–R and the
PAI. There was not much overlap between the individuals who did not complete the
PPI–R versus the PAI, and no overlap between those who provided invalid responses on
one measure versus the other, which ultimately led to excluding 45% of the original
sample (i.e., 56 profiles) from the analyses for research question 3.
It is worth pointing out that the high prevalence of invalid PPI–R or PAI profiles is
not a phenomenon specific to this study alone, but it is a rather commonly encountered
issue (see Nikolova, Hendry, Douglas, Edens, & Lilienfeld, 2012). It appears to reflect
one of the main shortcomings of self-report measures (i.e., random responding), and is
often related to the study sample (i.e., incarcerated populations). Thus, despite the a-
priori power estimations, power was lower due to invalid/missing data. Specifically,
according to the sample size estimations described in the Method section, the reduced
sample size used in the analyses most likely resulted in reduced power for research
question 4 (i.e., moderating effects of gender). The final sample used in the remaining
research questions, however, was still greater than the minimum sample needed to
detect moderate effect size departures from the null distribution as determined based on
the a-priori power estimations, and therefore it is likely that there were no adverse effect
on power for those research questions. Nevertheless, in order to compensate for the
reduced power in research question 4, and to determine whether missing data had a
negative impact on the patterns of association that emerged in this study, some of the
main analyses were repeated with the entire sample, as well as after employing less
stringent exclusion criteria for the PPI–R protocols. Results are presented in the
Supplemental Analyses section of this document.
Further, due to missing data, scale/subscale scores on the NEO–FFI and the PAI
were prorated. As recommended by the authors of the NEO–FFI, Costa and McCrae
(1992b), protocols containing 10 or more missing values were deemed invalid and were
not used in the analyses. For those with nine or fewer items left blank, neutral
responses were used to replace missing responses. In addition, a conservative
approach was taken for any protocols for which there were more than four missing items
54
per domain—such protocols were considered invalid and were not included in the
analyses. Further, the PAI manual specifies that scale or subscale scores should not be
interpreted if 20% or more of item-level data on a scale or a subscale are missing
(Morey, 2007). Therefore, PAI scores were prorated based on the mean for completed
items only if less than 20% of the item-level data were missing per scale/subscale.
Protocols for which 20% or more of the item-level data were missing from specific
scales/subscales were not included in the analyses. Furthermore, scoring software was
used to calculate PPI–R scores. It contained algorithms for dealing with missing data as
specified in the PPI–R manual. Namely, similar to the PAI, if 20% or more of item-level
data were missing on a single scale, the relevant scale score could not be computed.
Additionally, if more than 30 items in total were missing from a given protocol, it could
not be scored and was considered invalid (Lilienfeld & Widows, 2005).
The internal consistency of the PPI–R in the current sample was evaluated by
calculating Cronbach’s alpha coefficients as well as mean inter-item correlations (MIC)
and corrected item-to-total correlations (CITC) for each of the PPI–R content scales.
According to Nunnally and Bernstein (1994, pg. 264) alpha coefficients of .70 and higher
are indicative of satisfactory reliability. MIC values ranging at minimum from .15 to .50
are considered to be an indicator for satisfactory internal consistency reliability (Clark &
Watson, 1995). CITC’s revealed whether the internal consistency reliability of the PPI–R
would improve if certain items are removed, and if so, by how much. It is worth pointing
out that as a measure of internal consistency alpha reflects the intercorrelatedness of a
set of items, rather than their homogeneity or unidimensionality (Cortina, 1993). Internal
consistency is necessary, yet, not sufficient for homogeneity, and therefore it is an
inadequate index of unidimensionality. For instance, alpha coefficients for different sets
of items can be equal even when the item intercorrelation matrices for those sets are
radically different from one another. It is also true that alpha can be high even if a set of
items is multidimensional given the items are relatively intercorrelated (Cortina, 1993).
Further, in addition to being a function of intercorrelatedness between a set of items,
alpha is a function of the number of items included in a test, and as the number of items
55
increases, alpha increases. For instance a scale with 20 or more items will have an
alpha coefficient greater than .70, even if the intercorrelations between the items are
small. Therefore, the number of items in a scale could mask the dimensionality of the
scale. In summary, even though alpha increases as a function of item intercorrelations,
and decreases as a function of multidimensionality, it “can be high in spite of low item
intercorrlations and multidimensionality” (Cortina, 1993, pg. 103). It is not a measure of
unidimensionality, and it would be wrong to use it as such as it could lead to incorrect
conclusions (e.g., assuming that a measure is comprised of a single factor when that is
not the case; Schmitt, 1996). Therefore, precision estimates of alpha coefficients were
also calculated, and were used to determine the dimensionality of each of the scales.
Precision estimates were measured in terms of the standard error of item
intercorrelations, which is a function of the variance of item intercorrelations (Cortina,
1993). Precision estimates values of 0 reflect unidimensionality for a given scale, while
values greater than 0 suggest departure from unidimensionality.
56
Research Question 2:
Predictive Validity of the PPI–R with Respect to Recidivism and Type of the
Offense(s)
Multivariate analyses were used to examine the predictive validity of the PPI–R.
Specifically, survival analyses were employed as they are suitable for uneven follow-up
periods, and take into account time to re-offense (Luke & Homan, 1998). This was
important in the current study given that the follow-up period used to determine whether
participants engaged in criminal acts following release from custody was variable in
length depending on the date of recruitment (i.e., ranging from 6 to over 26 months).
Separate sets of survival analyses using a Cox model were carried out to evaluate the
predictive validity of the PPI–R total, factor, and content scale scores with regard to
recidivism in general, as well as with respect to the type of post release offense(s). In
addition, survival analyses using a Cox model were carried out to determine whether the
PPI–R has predictive utility in terms of differentiating between the type of post release
offense(s) committed by the individuals who recidivated. These analyses were
conducted with a subsample that included only those who recidivated following their
release from custody for the index offense.
Of note, with regard to the content scale scores, the first step included
conducting bivariate analyses to determine which of the scales were significantly
correlated with recidivism. Subsequently, only scales that were significantly correlated
with recidivism at the bivariate level were included in the survival analyses. This was
done considering there are eight content PPI–R scales in total and the sample size used
in the analyses ranged between 76 and 84 (depending on the outcome variable used),
thus, entering all of the content scales into a single survival analysis would have
substantially lowered the power of the analysis.
57
interpreted as the probability that a randomly selected recidivist scored higher on the
PPI–R than a randomly selected non-recidivist.
For all of the variables, forced entry method was used, where all variables (i.e.,
(a) PPI–R total score and recidivism or type of post release offense[s], (b) PPI–R factor
scores and recidivism or type of post release offense[s], and (c) PPI–R content scale
scores and recidivism or type of post release offense[s]) were entered into a single step
to ensure that all variables were included in the regression equation. In survival
analysis, the hazard represents the risk (i.e., the immediate potential) of the event
occurring (which in this case was recidivism), and time to the event is modeled as the
outcome. The hazard ratio (HR) estimates a relative risk that compares the rate of event
occurrence for those who have recidivated versus those who have not.
In addition, given that the majority of the individuals who recidivated did so during
the first 6 months following their release from custody, a series of logistic regression
analyses were conducted to examine the predictive validity of the PPI–R for a set follow-
up period (namely 6 months), where the PPI–R total, factor, and select scale scores
were used as predictors. First, a series of univariate analyses were conducted to
determine which of the PPI–R scores demonstrated significant associations with
recidivism within 6 months. The next step was to enter those scores as predictors into
multivariable logistic regression models. Implementing two different types of analyses to
evaluate the predictive validity of the PPI–R was done as a way of evaluating the
statistical conclusion validity of the survival analyses with regard to this research
question. Additionally, utilizing logistic regression in addition to survival analyses to
address this research question was deemed worthwhile as having a set follow-up period
can be policy relevant and of practical importance in the fields of law enforcement and
criminal justice.
In the present study, the observed value of odds ratios (in the case of logistic
regression) or hazard ratios (in the case of survival analysis) can be interpreted as
indicating an association between PPI–R scores and the increased relative risk of
recidivism. While odds/hazard ratios of 1.0 indicate no association between PPI–R
scores and recidivism, those greater than 1.0 indicate a positive association (i.e., high
scores on the PPI–R indicate an increased risk for recidivism), and those less than 1.0
58
indicate a negative association (i.e., decreased risk for recidivism associated with high
PPI–R scores). As an interpretive guideline, Fleiss, Williams, & Dubro (1986) suggested
that when dealing with dichotomous predictors, odds ratios of 2.5 and higher are
considered to represent clinically or practically meaningful associations, whereas
according to Haddock, Rindskopf, & Shadish (1998) odds ratios of 3.0 or greater are
considered to be large. It is important to note that when dealing with continuous
predictor variables that have large potential score ranges, odds or hazard ratios will have
small values as they reflect the effect of a one-unit increase in the predictor variable.
Research Question 3:
Incremental Validity of the PPI–R with Respect to Recidivism
Two data analytic approaches were used to evaluate the incremental validity of
the PPI–R with respect to recidivism relative to general self-report measures of
personality traits. First, point biserial correlations were computed between the criterion
measure (i.e., recidivism) and the scale/factor scores of the PPI–R, the AGG and ANT
scales of the PAI, as well as the Conscientiousness (C), Neuroticism (N), and
Agreeableness (A) scales of the NEO–FFI. Next, a series of hierarchical regression
analyses were computed, where the PAI and NEO–FFI scales listed above, if significant
at the bivariate level, were entered together in the first block of predictors, while the PPI–
R total or factor scores were entered in the second block. A separate regression
analysis was carried out for each of the PPI–R scores. Of note, these analyses were
carried out for the entire follow-up period where survival analyses using a Cox model
were employed, as well as for the 6-month follow-up period where logistic regression
59
analyses were conducted. Entry and removal criteria were set at p = .05 to enter, and p
= .10 to remove. The decision to choose a hierarchical over a stepwise approach to test
the incremental validity of the PPI–R was influenced by the fact that a hierarchical
approach tends to produce results that are tested in a theory-driven manner and are
more generalizable in comparison to those from a stepwise approach (Hunsley & Meyer,
2003).
Research Question 4:
Moderating Effects of Gender on PPI–R-Assessed Psychopathy
60
inflation. In addition, effect sizes for any detected gender differences were estimated
using Cohen’s d.
Does Gender Moderate the Predictive Utility of the PPI–R with Respect to
Recidivism?
The same steps were taken to examine whether gender moderated the
relationship between PPI–R-assessed psychopathy and the type of offense(s). In
addition, hierarchical regression analyses were used to determine whether gender
moderated the utility of the PPI–R in differentiating between the type of post release
offense(s) committed by the individuals who recidivated. These analyses were carried
out with a subsample that included only those individuals who recidivated and the same
steps as described above were implemented.
61
3. Results
The information presented in this section is based on data from participants who
provided valid responses on the PPI–R (i.e., n = 84; data from participants who provided
invalid responses, responded to an insufficient number of items, or provided no
responses at all, are not included here). Of note, for the analyses that focused on the
type of offense(s) outcome variable and included only participants who recidivated, the
sample size was reduced to 59. Participants in the study were almost equally split by
gender (i.e., 43 females and 41 males, or in the case of the type of offense(s) outcome,
where only those who recidivated were included, 32 females and 27 males), and the
mean age at the time of recruitment was 33.04 years (SD = 8.04). A majority of the
sample was Caucasian (n = 60, or 71%), followed by Aboriginal (n = 14, or 17%), East
Indian (n = 4, or 5%), Black (n = 3, or 4%), and Other (n = 3, or 4%). A majority of the
participants were never married (n = 61, or 73%), or at the time of recruitment were in a
common law relationship (n = 10, or 12%). English was a first language for a majority of
the sample (n = 66, or 79%), and a majority had not completed high school (n = 51, or
61%).
With regard to the participants’ index offense(s), 29 (35%; 42% males vs. 28%
females) engaged in acts that were classified as violent, 29 (35%; 32% males vs. 38%
females) engaged in non-violent acts, and 26 (31%; 27% males vs. 35% females)
engaged in other criminal acts (i.e., breaches or a combination of non-violent offenses
and breaches). As for the type of post release offense (i.e., recidivism), 23 (27%; 52%
males vs. 28% females) engaged in violent offenses, 12 (14%; 11% males vs. 28%
females) committed non-violent offenses, and 24 (29%; 37% males vs. 44% females)
engaged in other offenses. Details about participants’ prior criminal history were also
gathered and included: (a) age of first arrest or conviction, (b) number of prior court
dispositions, (c) failures to comply, or (d) supervision periods, and data was missing for
62
5 participants. Of the 79 (94%) for whom data was available, a majority (n = 29, or 35%;
30% males vs. 44% females) had their first arrest or conviction between the ages of 18
and 24. In addition, a majority had 5 or more prior court dispositions (n = 54, or 64%;
75% males vs. 62% females), 3 or more failures to comply (n = 56, or 67%; 73% males
vs. 69% females), and 3 or more prior supervision periods (n = 53, or 63%; 73% males
vs. 62% females). With regard to the individuals who were imprisoned after a prior
conviction, data were missing for 7 (8%) individuals, and of the remaining 77 (92%; 92%
males vs. 95% females), 72 (86%; 92% males vs. 95% females) had a history of multiple
incarcerations (see Table 1 for more details).
63
Total Sample (N = 84)
Characteristic n %
First language
English 66 78.6
Other 16 19.0
Vietnamese 1 1.2
Filipino 1 1.2
Type of index offense
Violent 29 34.5
Non-violent 29 34.5
Other 26 31.0
Type of recidivism
Violent 23 27.4
Non-violent 12 14.3
Other 24 28.6
Prior incarceration
NO 5 6.0
YES 72 85.7
Missing 7 8.3
Age at first arrest of conviction
25 and over 23 27.4
18 to 24 29 34.5
14 to 17 20 23.8
13 and UNDER 7 8.3
Missing 5 6.0
Number of prior court dispositions
0 4 4.8
1 7 8.3
2 to 4 14 16.7
5 or more 54 64.3
Missing 5 6.0
Number of prior supervision periods
0 8 9.5
1 11 13.1
2 7 8.3
64
Total Sample (N = 84)
Characteristic n %
3 or more 53 63.1
Missing 5 6.0
Number of prior failures to comply
0 7 8.3
1 8 9.5
2 8 9.5
3 or more 56 66.7
Missing 5 6.0
Assumptions for normality and linearity of the data were checked by evaluating
the skewness and kurtosis of each variable, as well as examining scatter plots for each
of the predicted correlations. Based on the skewness and kurtosis, as well as the
produced scatter plots of each variable, it was determined that the assumptions of
normality and linearity were not violated. The bivariate correlations between the
predictors were examined to assess for multicollinearity in the regression analyses, and
there was no indication of any issues in that regard. The assumption of homogeneity of
error variance was tested using Levene’s test of equality of variances, which revealed
non-significant results, suggesting that group variance could be treated as equal. The
proportional hazards (PH) assumption was checked by examining the residual plots of
the PPI–R scores and it was determined that the PH assumption was not violated.
65
Table 2. Internal Consistency Reliability of the PPI–R and Comparative
Measures
66
3.3. Descriptive Information
Measures of central tendency, error, range, and variance were calculated for the
PPI–R total and factor t-scores. Mean scores for each of the scales were as follows:
63.37 (SD = 13.25) for the total score, 63.54 (SD = 10.77) for the SCI factor score, 53.89
(SD = 12.18) for the FD factor score, and 48.99 (SD = 9.58) for the C factor score (see
Table 3 for more details). Of note, there was no restriction in the range of the PPI–R
total, factor, and scale scores that could have had a negative impact on power.
67
3.4. Research Question 1: Reliability of the PPI–R
68
.837, or α = .832 respectively. For the C scale, removal of item 142 would have resulted
in an α = .840. Removal of items 50 and 54 on the STI scale would have yielded α =
.826 or α = .829, respectively. Removal of item 51 on the CN scale would have led to an
α = .812. Finally, removal of items 16, 19, and 82, on the BE scale would have resulted
in the following α coefficients: α = .859, α = .857, or α = .853, respectively. Even so, in
all instances the improvement would have been negligible, enhancing alpha only by .001
to .007.
69
Table 4. Bivariate Correlations between PPI–R Total, Factor, and Scale
Scores, and Overall Recidivism
Scale Recidivism
PPI–R total score .339**
SCI factor score .339**
FD factor score .162
C factor score .136
ME content scale score .315**
RN content scale score .219*
BE content scale score .220*
CN content scale score .283**
SOI content scale score .111
F content scale score .111
STI content scale score .113
*p ≤ .05, two-tailed. **p ≤ .01, two-tailed.
70
Variable AUC SE CI (95%) p-value
Other offenses
PPI–R total score 0.622 0.068 [.488, .756] .081
SCI factor score 0.590 0.066 [.461, .718] .200
FD factor score 0.562 0.072 [.422, .702] .376
C factor score 0.574 0.064 [.448, .700] .292
Note. AUC = area under the curve; SE = standard error; CI = confidence interval.
Detailed information regarding the survival analyses, including full model details,
is presented in Table 6 (i.e., recidivism) and Table 7 (i.e., the type of post release
offense[s]). The survival curves for the entire sample, as well as for each gender, are
presented in Figures 1 and 2 respectively. The log-rank test showed no significant
difference between the survival curves for males versus females (p = .846). As a
reminder, while the recidivism outcome variable captures whether or not an individual
engaged in criminal activities subsequent to their release from custody for the index
offense, the type of offense(s) variable groups any criminal activity that might have
occurred into three categories based on the severity or gravity of the offense(s) (i.e., 1.
violent; 2. non-violent; and 3. other = breaches, or a combination of non-violent offenses
and breaches). While all of the final models with regard to recidivism were significant,
the only significant final model with respect to the type of offense(s) was that of the PPI–
R total score vis-a-vis violent offenses. With regard to the total score, the survival
analyses revealed that the PPI–R total score was significantly and independently
predictive of recidivism (i.e., HR = 1.029, p =.001, Wald = 10.368). The PPI–R total
score was also significant in predicting violent offenses (i.e., HR = 1.036, p = .012, Wald
= 6.349), but was not significant in predicting non-violent or other offenses. In terms of
the factor score analyses, only the SCI factor score was significantly predictive of
recidivism (i.e., HR = 1.035, p = .011, Wald = 6.407). None of the factor scores were
significant in predicting the type of post release offenses. It is worth pointing out,
however, that when correlations between the PPI–R scores and type of offense(s) were
computed separately for males and females, there was a significant correlation between
the SCI factor and violent offenses for females (i.e., r = .308, p = .045), but not for males
(i.e., r = .079, p = .623). The content scales scores that were significantly correlated with
recidivism at the bivariate level (i.e., ME, RN, BE, and CN) were not significant in
71
predicting recidivism or the type of post release offense(s) based on the survival
analyses. Finally, none of the PPI–R scores were significant in terms of differentiating
between the type of post release offenses (i.e., violent vs. non-violent vs. other)
committed by the individuals who recidivated (see Table 8 for details).
72
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
Non-Violent Offenses
PPI–R Total d 0.021 0.022 0.939 1.021 [0.979, 1.065] .332 0.11
SCI factor score e 0.051 0.032 2.561 1.053 [0.989, 1.121] .110 0.19
FD factor score e -0.022 0.029 0.560 0.979 [0.925, 1.036] .454 0.19
C factor score e 0.004 0.027 0.023 1.004 [0.953, 1.058] .879 0.19
ME content scale score f 0.029 0.050 0.331 1.029 [0.933, 1.134] .565 0.25
RN content scale score f -0.026 0.032 0.700 0.974 [0.915, 1.036] .403 0.25
BE content scale score f 0.015 0.037 0.174 1.015 [0.945, 1.091] .677 0.25
CN content scale score f 0.066 0.038 3.037 1.069 [0.992, 1.151] .081 0.25
Other offenses
PPI–R Total g 0.026 0.014 3.464 1.026 [0.999--1.055] .063 0.21
SCI factor score h 0.026 0.021 1.599 1.027 [0.986, 1.069] .206 0.21
FD factor score h 0.009 0.019 0.245 1.009 [0.973, 1.047] .620 0.21
C factor score h 0.006 0.020 0.100 1.006 [0.967, 1.048] .751 0.21
ME content scale score i 0.029 0.030 0.905 1.029 [0.970, 1.091] .341 0.22
RN content scale score i -0.001 0.023 0.003 0.999 [0.954, 1.045] .956 0.22
BE content scale score i -0.006 0.027 0.056 0.994 [0.942, 1.048] .812 0.22
CN content scale score i 0.026 0.026 1.011 1.026 [0.976, 1.080] .315 0.22
Note. SE = standard error; eB = univariate transformation of the estimated Beta coefficient, which
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991).
a Model -2LL = 178.695, χ2 (df = 1, n = 84) = 6.557, p = .010.
b Model -2LL = 178.655, χ2 (df = 3, n = 84) = 6.584, p = .086.
c Model -2LL = 179.536, χ2 (df = 4, n = 84) = 5.358, p = .252.
d Model -2LL = 82.290, χ2 (df = 1, n = 76) = .948, p = .330.
e Model -2LL = 80.289, χ2 (df = 3, n = 76) = 2.750, p = .432.
f Model -2LL = 78.371, χ2 (df = 4, n = 76) = 4.849, p = .303.
g Model -2LL = 192.338, χ2 (df = 1, n = 83) = 3.524, p = .060.
h Model -2LL = 192.303, χ2 (df = 3, n = 83) =3.532, p = .317.
i Model -2LL = 191.774, χ2 (df = 4, n = 83) = 4.068, p = .397.
73
Figure 1. Kaplan-Meier Survival Estimate
74
Table 8. Utility of the PPI–R to Differentiate between the Type of Post Release
Offense(s) among Those Who Recidivated
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
Violent offenses
PPI–R Total a 0.021 0.015 1.969 1.021 [0.992, 1.052] .161 0.18
SCI factor score b 0.009 0.023 0.162 1.009 [0.965, 1.056] .688 0.21
FD factor score b 0.026 0.021 1.507 1.026 [0.985, 1.069] .220 0.21
C factor score b -0.006 0.026 0.053 0.994 [0.944, 1.047] .817 0.21
ME content scale score c 0.005 0.030 0.032 1.005 [0.948, 1.066] .858 0.16
RN content scale score c 0.016 0.023 0.483 1.016 [0.972, 1.062] .487 0.16
BE content scale score c -0.001 0.031 0.002 0.999 [0.940, 1.061] .967 0.16
CN content scale score c -0.009 0.027 0.100 0.991 [0.940, 1.045] .751 0.16
Non-Violent Offenses
PPI–R Total d 0.018 0.027 0.479 1.019 [0.967, 1.073] .489 0.10
SCI factor score e 0.029 0.038 0.588 1.030 [0.956, 1.109] .443 0.11
FD factor score e -0.001 0.040 0.001 0.999 [0.924, 1.080] .982 0.11
C factor score e 0.000 0.051 0.000 1.000 [0.904, 1.106] .993 0.11
ME content scale score f -0.035 0.050 0.479 0.966 [0.875, 1.066] .489 0.21
RN content scale score f 0.053 0.040 1.726 1.054 [0.974, 1.140] .189 0.21
BE content scale score f -0.008 0.054 0.022 0.992 [0.893, 1.102] .883 0.21
CN content scale score f 0.002 0.047 0.002 1.002 [0.914, 1.099] .961 0.21
Other offenses
PPI–R Total g 0.010 0.015 0.473 1.010 [0.981, 1.041] .492 0.09
SCI factor score h 0.001 0.023 0.003 1.001 [0.958, 1.047] .955 0.11
FD factor score h 0.014 0.021 0.447 1.014 [0.974, 1.056] .504 0.11
C factor score h -0.001 0.026 0.000 0.999 [0.949, 1.052] .982 0.11
ME content scale score i 0.007 0.029 0.061 1.007 [0.951, 1.067] .805 0.14
RN content scale score i 0.013 0.022 0.336 1.013 [0.970, 1.058] .562 0.14
BE content scale score i -0.025 0.030 0.686 0.975 [0.919, 1.035] .407 0.14
CN content scale score i 0.008 0.026 0.092 1.008 [0.957, 1.062] .762 0.14
Note. SE = standard error; = univariate transformation of the estimated Beta coefficient, which
eB
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991).
a Model -2LL = 151.978, χ2 (df = 1, n = 59) = 1.997, p = .158.
75
b Model -2LL = 151.408, χ2 (df = 3, n = 59) = 2.557, p = .465.
c Model -2LL = 152.408, χ2 (df = 4, n = 59) = 1.537, p = .820.
d Model -2LL = 49.550, χ2 (df = 1, n = 51) = .485, p = .486.
e Model -2LL = 49.356, χ2 (df = 3, n = 51) = .661, p = .882.
f Model -2LL = 47.724, χ2 (df = 4, n = 51) = 2.242, p = .691.
g Model -2LL = 165.327, χ2 (df = 1, n = 58) = .475, p = .491.
h Model -2LL = 165.127, χ2 (df = 3, n = 58) = .679, p = .878.
i Model -2LL = 164.633, χ2 (df = 4, n = 58) = 1.119, p = .891.
Results from the univariate analyses are presented in Table 9, where odds ratios
reflect the magnitude of the associations between the PPI–R scores (i.e., total, factor,
and select scale scores that showed significant correlations) and recidivism over the
specified follow-up period (i.e., 6 months). Specifically, the PPI–R total score, two of the
factor scores (i.e., SCI and FD), as well as two of the scale scores (i.e., ME and RN)
were determined to be significant predictors of recidivism during the 6-month follow-up
period.
76
Scale 0 to 6 months (OR)
SOI content scale score 1.036
F content scale score 1.030
STI content scale score 1.021
*p ≤ .05. **p ≤ .01.
77
Similarly, when the SCI and FD factor scores were entered as predictors, a test
of the full model against the constant only model was statistically significant, and the
variance explained by the model was moderate, at 15% (i.e., Nagelkerke R-squared =
.146). Therefore, it could be concluded that those two factor scores could also be used
to distinguish between individuals who recidivated versus those who did not (χ2 = 9.962,
p = .008, df = 2). The overall prediction accuracy of this model was 71% (81% for those
who recidivated and 58% for those who did not). Based on the Wald criterion only the
SCI factor was a significant predictor in the model (p = .035). The odds ratio value
(Exp(B) = 1.051) indicates that for every one point increase in the SCI score, the
likelihood of recidivism increased 1.051 times.
With regard to the scale scores, when the ME and RN scales were entered as
predictors, a test of the full model against the constant only model was statistically
significant, and based on the Nagelkerke R-squared, the variance explained by the
model was modest, at 11%. This indicated that those two scale scores could also be
used to distinguish between individuals who recidivated versus those who did not (χ2 =
7.413, p = .025, df = 2). The overall prediction accuracy of this model was 63% (77% for
those who recidivated and 44% for those who did not). The scale scores themselves
were no longer significant predictors.
Finally, based on the test of differences between correlations, the SCI factor
evidenced a coefficient of 0.131, which represented 13% of the maximum possible
change in the probability of recidivism, as opposed to a coefficient of 0.098 for the FD
factor, which represented 9% of the total change possible in the probability of recidivism.
78
Antisocial features (ANT) scales of the PAI, as well as the Conscientiousness (C),
Neuroticism (N), and Agreeableness (A) scales of the NEO–FFI. Results revealed that
the PPI–R total and SCI factor scores (r = .342, p = .004 and r = .326, p = .006), as well
as the ANT scale of the PAI (r = .295, p = .014), were correlated positively with
recidivism over the entire follow-up period (see Table 11). Results were somewhat
different with regard to recidivism over the 6-month follow-up period. In particular, the
PPI–R total, SCI and FD factor scores (r = .328, p = .006, r = .265, p = .028, and r =
.268, p = .026 respectively), as well as the ANT and AGG scale of the PAI (r = .302, p
=.012 and r = .289, p =.016 respectively), were correlated positively with recidivism
(details provided in Table 11).
Table 11. Bivariate Correlations between the PPI–R, PAI, NEO–FFM Scores
and Recidivism
Recidivism
Scale Entire Follow-up Period 6-month Follow-up Period
PPI–R total score .342** .328**
SCI factor score .326** .265*
FD factor score .210 .268*
C factor score .139 .127
PAI ANT scale .295* .302*
PAI AGG scale .191 .289*
NEO N scale -.008 -.011
NEO A scale -.173 -.159
NEO C scale -.030 .066
*p ≤ .05, two-tailed. **p ≤ .01, two-tailed.
Next, two survival analyses were carried out (i.e., [1] PPI–R total and ANT scale,
and [2] SCI factor and ANT scale) to examine the incremental validity of the PPI–R
relative to the ANT scale over the entire follow-up period. Results revealed that while
overall models were significant, and the ANT scale when entered alone in the model was
predictive of recidivism (i.e., HR = 1.040, p = .001, Wald = 10.717), adding the PPI–R
scores to the models did not add incrementally to the prediction of recidivism (see Table
12 for detailed information regarding the hazard ratios and the final models).
79
Table 12. Incremental Validity of the PPI–R Relative to the PAI vis-à-vis
Recidivism over the Entire Follow-Up Period
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
PAI ANT scale a 0.032 0.019 3.021 1.033 [.996, 1.071] .082 0.41
PPI–R total a 0.007 0.014 0.278 1.008 [.980, 1.036] .598 0.41
PAI ANT scale b 0.037 0.020 3.441 1.038 [.998, 1.080] .064 0.40
SCI factor score b 0.003 0.021 0.018 1.003 [.963, 1.044] .893 0.40
Note. SE = standard error; eβ = univariate transformation of the estimated Beta coefficient, which
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991).
a Model -2LL = 361.303, χ2 (df = 2, n = 69) = 11.331, p = .003.
b Model -2LL = 361.567, χ2 (df = 2, n = 69) = 10.975, p = .004.
Table 13. Incremental Validity of the PPI–R vis-à-vis Recidivism over the
6-Month Follow-Up Period
80
B SE (B) Wald df p-value OR (95% CI)
SCI factor score 0.010 0.044 0.049 1 .825 1.010 [0.927, 1.100]
Constant -4.648 1.885 6.077 1 .014 0.010
Model 3 (FD factor score and ANT, AGG) c
ANT scale 0.034 0.026 1.724 1 .189 1.034 [0.984, 1.087]
AGG scale 0.025 0.023 1.197 1 .274 1.025 [0.981, 1.072]
FD factor score 0.030 0.024 1.602 1 .206 1.030 [0.984, 1.079]
Constant -5.336 1.959 7.418 1 .006 0.005
Note. SE = standard error; df = degrees of freedom; OR = odds ratio; CI = confidence interval.
a Model -2LL = 83.303, χ2 (df = 3, n = 69) = 9.888, p = .020.
b Model -2LL = 84.496, χ2 (df = 3, n = 69) = 8.694, p = .034.
c Model -2LL = 82.877, χ2 (df = 3, n = 69) = 10.313, p = .016.
Table 14. Bivariate Correlations between the PPI–R Scores, CRNA Ratings,
and Recidivism (Entire Follow-Up/6-Month Follow-Up Period)
1 2 3 4 5 6 7 8 9 10
1. Recidivism -- entire
1
follow-up period
n 84
2. Recidivism -- 6-month
-- 1
follow-up period
n 84
81
1 2 3 4 5 6 7 8 9 10
3. Summary CRNA rating .558** .432** --
n 79 79
4. Supervision Level rating .560** .436** 1.000** --
n 80 80 79
5. Needs rating .309** .294** .664** .665** --
n 79 79 78 79
6. Risk rating .421** .351** .734** .725** .353** --
n 79 79 78 79 79
7. Total score .339** .340** .192 .200 .018 .422** --
n 84 84 79 80 79 79
8. SCI factor score .339** .284* .112 .121 -.041 .346** .852** --
n 84 84 79 80 79 79 84
9. FD factor score .162 .237* .127 .135 .024 .260* .686** .262* --
n 84 84 79 80 79 79 84 84
10. C factor score .136 .153 .265* .255* .156 .258* .436** .143 .274* --
n 84 84 79 80 79 79 84 84 84
*p ≤ .05, two-tailed. **p ≤ .01, two-tailed.
Next, six survival analyses were carried out (i.e., [1] PPI–R total and Risk CRNA
rating, [2] SCI factor and Risk CRNA rating scale, [3] FD factor and Risk CRNA rating,
[4] C factor and Risk CRNA rating, [5] C factor and Summary CRNA rating, and [6] C
factor and Supervision CRNA rating) to examine the potential incremental validity of the
PPI–R relative to the CRNA over the entire follow-up period. Results revealed that all of
the overall models were significant (p ≤ .001 for all), and while the CRNA ratings were
predictive of recidivism (i.e., [1] HR = 2.063, p = .003, Wald = 8.769; [2] HR = 2.089, p =
.002, Wald = 9.272; [3] HR = 2.199, p = .001, Wald = 11.282; [4] HR = 2.223, p = .001,
Wald = 11.592, [5] HR = 4.487, p < .001, Wald = 15.795, and [6] HR = 4.569, p < .001,
Wald = 16.223), adding the PPI–R scores to the model did not add incrementally to the
prediction of recidivism (see Table 15 for detailed information regarding the hazard ratios
and the full models).
82
Table 15. Incremental Validity of the PPI–R vis-à-vis Recidivism over the
Entire Follow-Up Period
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
CRNA Risk ratings a 0.724 0.245 8.769 2.063 [1.277, 3.33] .003 0.44
PPI–R total a 0.014 0.010 1.997 1.014 [0.995, 1.033] .158 0.44
CRNA Risk ratings b 0.737 0.242 9.272 2.089 [1.300, 3.357] .002 0.44
SCI factor score b 0.021 0.013 2.557 1.021 [0.995, 1.048] .110 0.44
CRNA Risk ratings c 0.788 0.235 11.282 2.199 [1.388, 3.482] .001 0.42
FD factor score c 0.007 0.011 0.423 1.007 [0.986, 1.029] .515 0.42
CRNA Risk ratings d 0.799 0.235 11.592 2.223 [1.404, 3.522] .001 0.42
C factor score d 0.005 0.012 0.147 1.005 [0.981, 1.028] .701 0.42
CRNA Supervision Level
1.519 0.377 16.223 4.569 [2.181, 9.570] .000 0.50
ratings e
C factor score e 0.002 0.012 0.030 1.002 [0.978, 1.027] .863 0.50
CRNA Summary ratings f 1.501 0.378 15.795 4.487 [2.140, 9.407] .000 0.50
C factor score f 0.004 0.012 0.084 1.004 [0.980, 1.028] .771 0.50
Note. SE = standard error; eβ = univariate transformation of the estimated Beta coefficient, which
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991).
a Model -2LL = 420.734, χ2 (df = 2, n = 79) = 15.584, p < .001.
b Model -2LL = 420.152, χ2 (df = 2, n = 79) = 15.416, p < .001.
c Model -2LL = 422.287, χ2 (df = 2, n =79) = 14.132, p = .001.
d Model -2LL = 422.562, χ2 (df = 2, n = 79) = 13.794, p = .001.
e Model -2LL = 415.694, χ2 (df = 2, n = 80) = 20.074 p < .001.
f Model -2LL = 407.423, χ2 (df = 2, n = 79) = 19.783, p < .001.
83
Table 16. Incremental Validity of the PPI–R vis-à-vis Recidivism over the
6- Month Follow-Up Period
B SE (B) Wald df p-value OR (95% CI)
Model 1 (PPI–R total score and
CRNA Risk ratings) a
CRNA Risk ratings 0.876 0.402 4.753 1 .029 2.400 [1.092, 5.273]
PPI–R total score 0.035 0.022 2.454 1 .117 1.035 [0.991, 1.081]
Constant -3.994 1.463 7.449 1 .006 0.018
Model 2 (SCI factor score and
CRNA Risk ratings) b
CRNA Risk ratings 0.971 0.392 6.149 1 .013 2.641 [1.226, 5.690]
SCI factor score 0.032 0.025 1.641 1 .200 1.033 [0.983, 1.085]
Constant -4.115 1.729 5.666 1 .017 0.016
Model 3 (FD factor score and CRNA
Risk ratings) c
CRNA Risk ratings 1.011 0.378 7.134 1 .008 2.748 [1.309, 5.768]
FD factor score 0.021 0.022 0.930 1 .335 1.022 [0.978, 1.067]
Constant -3.290 1.374 5.734 1 .017 0.037
Model 4 (C factor score and CRNA
Risk ratings) d
CRNA Risk ratings 1.036 0.379 7.470 1 .006 2.817 [1.340, 5.970]
C factor score 0.018 0.028 0.428 1 .513 1.018 [0.964, 1.076]
Constant -3.091 1.488 4.317 1 .038 0.045
Model 5 (C factor score and CRNA
Supervision Level ratings) e
CRNA Risk ratings 1.902 0.559 11.584 1 .001 6.701 [2.241, 20.037]
C factor score 0.010 0.028 0.119 1 .730 1.010 [0.956, 1.066]
Constant -5.266 1.802 8.543 1 .003 0.005
Model 6 (C factor score and CRNA
Summary ratings) f
CRNA Risk ratings 1.863 0.559 11.090 1 .001 6.444 [2.152, 19.292]
C factor score 0.012 0.028 0.192 1 .662 1.012 [0.958, 1.070]
Constant -5.312 1.808 8.634 1 .003 0.005
Note. SE = standard error; df = degrees of freedom; OR = odds ratio; CI = confidence interval.
a Model -2LL = 94.839, χ2 (df = 2, n = 79) = 12.530, p = .002.
b Model -2LL = 95.730, χ2 (df = 2, n = 79) = 11.639, p = .003.
c Model -2LL = 96.459, χ2 (df = 2, n =79) = 10.909, p = .004.
d Model -2LL = 96.975, χ2 (df = 2, n = 79) = 10.393, p = .006.
84
e Model -2LL = 92.876, χ2 (df = 2, n = 80) = 16.221 p < .001.
f Model -2LL = 92.126, χ2 (df = 2, n = 79) = 15.854, p <.001.
85
Gender n Mean SD
FD factor score male 41 57.98 10.829
female 43 50.00 12.236
C factor score male 41 52.37 10.754
female 43 45.77 7.044
Note. SD = standard deviation.
98.75% CI
Mean SE
t df Sig. (1-tailed) Difference Difference Lower Upper
PPI–R total score 3.134 82 .001 8.618 2.749 1.596 15.639
SCI factor score 1.386 82 .085 3.242 2.338 -2.730 9.213
FD factor score 3.158 82 .001 7.976 2.526 1.525 14.426
C factor score 3.342 82 .001 6.598 1.974 1.556 11.641
Note. t = t statistic; df = degrees of freedom; Sig. = significance value; SE = standard error; CI =
confidence interval.
Type of Offense(s)
Gender Recidivism Violent Non-Violent Other
PPI–R total score .327** .339** .238* -.073 .164
SCI factor score .151 .339** .199 .002 .146
86
Type of Offense(s)
Gender Recidivism Violent Non-Violent Other
PPI–R total score .327** .339** .238* -.073 .164
FD factor score .329** .162 .175 -.142 .101
C factor score .346** .136 .093 -.042 .078
Gender 1 -.094 .148 -.194 -.090
*p ≤ .05, two-tailed. **p ≤ .01, two-tailed.
Table 20. Moderation Analyses for Gender and the PPI–R Total Score vis-à-vis
Future Crime and Violence
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
PPI–R total score a 0.027 0.013 4.488 1.028 [1.002, 1.054] .034 0.37
Gender a 0.222 0.281 0.629 1.249 [0.721, 2.165] .428 0.37
Gender x PPI.R total score a 0.009 0.019 0.242 1.009 [0.973, 1.047] .623 0.37
Note. SE = standard error; eβ = univariate transformation of the estimated Beta coefficient, which
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991).
a Model -2LL = 452.662, χ2 (df = 3, n = 84) = 11.586, p = .009.
87
0.109), and there were no moderating effects (HR = 1.008, p =.754, Wald = 0.098; see
Table 21 for details).
Table 21. Moderation Analyses for Gender and the SCI Factor Score vis-à-vis
Future Crime and Violence
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
SCI Factor score b 0.036 0.017 4.519 1.037 [1.003, 1.072] .034 0.35
Gender b 0.090 0.273 0.109 1.094 [0.641, 1.869] .741 0.35
Gender x SCI Factor score b 0.008 0.025 0.098 1.008 [0.960, 1.058] .754 0.35
Note. SE = standard error; eβ = univariate transformation of the estimated Beta coefficient, which
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991).
b Model -2LL = 453.407, χ2 (df = 3, n = 84) = 10.017, p = .018.
In terms of the hierarchical regression analyses, Cox regression was carried out
over the entire follow-up period and detailed information regarding the final models is
presented in Tables 22 through 33. As in the previous research question, males served
as the reference group and the results are presented from females’ perspective. In
summary, none of the final models with regard to the type of offense(s) were significant.
In addition, there were no significant main effects of the PPI–R scores or gender, or
moderating effects of gender on the predictive utility of the PPI–R scores with regard to
the type of the offense(s).
88
Table 22. Moderation Analyses for Gender and the PPI–R Total Score vis-à-vis
Gravity of the Offense (VIOLENT offenses)
Table 23. Moderation Analyses for Gender and the PPI–R Total Score vis-à-vis
Gravity of the Offense (NON-VIOLENT offenses)
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
PPI–R total score b 0.044 0.038 1.326 1.045 [0.970, 1.126] .249 0.22
Gender b 1.149 0.749 2.350 3.154[ 0.726, 13.697] .125 0.22
Gender x PPI.R total score b -0.013 0.047 0.082 0.987 [0.900, 1.082] .775 0.22
Note. SE = standard error; eβ = univariate transformation of the estimated Beta coefficient, which
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991).
b Model -2LL = 79.431, χ2 (df = 3, n = 76) = 3.580 p = .311.
Table 24. Moderation Analyses for Gender and the PPI–R Total Score vis-à-vis
Gravity of the Offense (OTHER offenses)
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
PPI–R total score c 0.026 0.021 1.489 1.027 [0.984, 1.071] .222 0.24
Gender c 0.404 0.452 0.797 1.497 [0.617, 3.633] .372 0.24
Gender x PPI.R total score c 0.009 0.029 0.100 1.009 [0.953, 1.069] .752 0.24
Note. SE = standard error; eβ = univariate transformation of the estimated Beta coefficient, which
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991).
c Model -2LL = 191.210, χ2 (df = 3, n = 83) = 4.804, p = .187.
89
Table 25. Moderation Analyses for Gender and the SCI Factor Score vis-à-vis
Gravity of the Offense (VIOLENT offenses)
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
SCI factor score d 0.018 0.023 0.591 1.018 [0.973, 1.065] .442 0.29
Gender d -0.671 0.460 2.129 0.511 [0.207, 1.259] .145 0.29
Gender x SCI factor score d 0.066 0.043 2.375 1.068 [0.982, 1.161] .123 0.29
Note. SE = standard error; eβ = univariate transformation of the estimated Beta coefficient, which
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991).
d Model -2LL = 176.153, χ2 (df = 3, n = 84) = 7.115, p = .068.
Table 26. Moderation Analyses for Gender and the SCI Factor Score vis-à-vis
Gravity of the Offense (NON-VIOLENT offenses)
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
SCI factor score e 0.081 0.051 2.527 1.085 [0.981, 1.199] .112 0.23
Gender e 1.004 0.752 1.782 2.730 [0.625, 11.930] .182 0.23
Gender x SCI factor score e -0.049 0.062 0.639 0.952 [0.844, 1.074] .424 0.23
Note. SE = standard error; eβ = univariate transformation of the estimated Beta coefficient, which
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991).
e Model -2LL = 78.294 χ2 (df = 3, n = 76) = 3.968, p = .265.
Table 27. Moderation Analyses for Gender and the SCI Factor Score vis-à-vis
Gravity of the Offense (OTHER offenses)
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
SCI factor score f 0.042 0.029 2.137 1.043 [0.986, 1.103] .144 0.20
Gender f 0.349 0.453 0.595 1.418 [0.584, 3.442] .440 0.20
Gender x SCI factor score f -0.012 0.039 0.097 0.988 [0.914, 1.067] .756 0.20
Note. SE = standard error; eβ = univariate transformation of the estimated Beta coefficient, which
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991).
f Model -2LL = 192.152, χ2 (df = 3, n = 83) = 3.324, p = .344.
90
Table 28. Moderation Analyses for Gender and the FD Factor Score vis-à-vis
Gravity of the Offense (VIOLENT offenses)
Table 29. Moderation Analyses for Gender and the FD Factor Score vis-à-vis
Gravity of the Offense (NON-VIOLENT offenses)
Table 30. Moderation Analyses for Gender and the FD Factor Score vis-à-vis
Gravity of the Offense (OTHER offenses)
91
Table 31. Moderation Analyses for Gender and the C Factor Score vis-à-vis
Gravity of the Offense (VIOLENT offenses)
Table 32. Moderation Analyses for Gender and the C Factor Score vis-à-vis
Gravity of the Offense (NON-VIOLENT offenses)
Table 33. Moderation Analyses for Gender and the C Factor Score vis-à-vis
Gravity of the Offense (OTHER offenses)
92
Further, survival analyses were carried out over the entire follow-up period to
determine whether gender moderated the utility of the PPI–R in differentiating between
the type of post release offense(s) committed by the individuals who recidivated. As a
reminder, the subsample used for these analyses included only individuals who
recidivated. Detailed information regarding the final models is presented in Tables 34
through 36. In summary, all of the final models with regard to violent offenses were
significant. In addition, results revealed significant main effects for gender in each of the
models pertaining to violent offenses (i.e., when gender was entered as a predictor
together with the total and each of the factor scores). With regard to non-violent
offenses, gender also yielded a main effect when entered in a model with the C factor
score. Results with regard to violent offenses were as follows: HR = 0.190, p = .002,
a
Wald = 9.743 for model (i.e., including the total score as predictor), HR = 0.131, p <
b
.001, Wald = 13.148 for model (i.e., including the SCI factor), HR = 0.286, p = .012,
Wald = 6.262 for model c (i.e., including the FD factor), and HR = 0.181, p = .004, Wald
= 8.168 for model d (i.e., including the C factor; see Table 34). The main effect of gender
d
with regard to non-violent offenses in model (i.e., including the C factor score as a
predictor) was HR = 0.136, p = .049, Wald = 3.867 (see Table 35).
Further, there was a moderation effect of gender (HR = 1.153 for females and
HR = 0.867 for males, p = .004, Wald = 8.088) on the predictive utility of the SCI factor
score with regard to violent offenses in this subsample. The interaction is presented in
Figure 3. Specifically, under conditions of high SCI factor scores, males and females
had comparable rates of violent recidivism. Conversely, under conditions of low SCI
factor scores, males had significantly higher frequency of violent recidivism. The SCI
factor score itself and gender (as already reported above) were also found to be
significant predictors of violent offenses in the model (p = .035 and p < .001 respectively;
see Table 34). Of note, when broken down by gender, the bivariate correlations
between the SCI factor score and violent re-offending were non-significant and in the
opposite direction of each other (i.e., r = -.185, p =.356 for men, and r = .262, p = .147
for women). No moderating effects of gender were found on the SCI factor score with
regard to non-violent or other offenses in this subsample. Results were also non-
significant for the moderating effects of gender on the total score and the rest of the
factor scores (i.e., FD and C) and type of offense(s) (see Tables 34 through 36).
93
Figure 3. Moderating Effects of Gender on the Relationship between SCI
Factor Score and Violent Offenses
Table 34. Moderation Analyses for Gender and the PPI–R Scores vis-à-vis
Utility of the PPI–R to Differentiate between the Type of Post-
Release Offense(s) (VIOLENT offenses)
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
PPI–R total score a -0.020 0.021 0.912 0.980 [0.941, 1.021] .339 0.46
Gender a -1.659 0.531 9.743 0.190 [0.067, 0.539] .002 0.46
Gender x PPI.R total score a 0.061 0.036 2.887 1.063 [0.991, 1.141] .089 0.46
SCI factor score b -0.065 0.031 4.432 0.937 [0.882, 0.996] .035 0.58
Gender b -2.035 0.561 13.148 0.131 [0.043, 0.393] .000 0.58
Gender x SCI factor score b 0.143 0.050 8.088 1.153 [1.045, 1.273] .004 0.58
FD factor score c 0.025 0.026 0.962 1.026 [0.975, 1.079] .327 0.45
Gender c -1.250 0.500 6.262 0.286 [0.108, 0.763] .012 0.45
Gender x FD factor score c -0.014 0.039 0.126 0.986 [0.914, 1.065] .723 0.45
C factor scored -0.001 0.027 0.001 0.999 [0.948, 1.054] .979 0.43
Gender d -1.711 0.599 8.168 0.181 [0.056, 0.584] .004 0.43
Gender x C factor score d -0.087 0.068 1.613 0.917 [0.802, 1.048] .204 0.43
Note. SE = standard error; eβ = univariate transformation of the estimated Beta coefficient, which
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
94
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991).
a Model -2LL = 141.630, χ2 (df = 3, n = 59) = 12.351, p = .006.
b Model -2LL = 135.864, χ2 (df = 3, n = 59) = 19.876, p < .001.
c Model -2LL =143.375, χ2 (df = 3, n = 59) = 12.176, p = .007.
d Model -2LL =142.390, χ2 (df = 3, n = 59) = 10.980, p = .012.
Table 35. Moderation Analyses for Gender and the PPI–R Scores vis-à-vis
Utility of the PPI–R to Differentiate between the Type of Post-
Release Offense(s) (NON-VIOLENT offenses)
Table 36. Moderation Analyses for Gender and the PPI–R Scores vis-à-vis
Utility of the PPI–R to Differentiate between the Type of Post-
Release Offense(s) (OTHER offenses)
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
PPI–R total score a -0.007 0.025 0.075 0.993 [0.946, 1.043] .784 0.15
Gender a -0.435 0.493 0.777 0.647 [0.246, 1.702] .378 0.15
Gender x PPI.R total score a 0.024 0.033 0.509 1.024 [0.959, 1.093] .476 0.15
95
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
SCI factor score b -0.011 0.034 0.100 0.989 [0.925, 1.058] .752 0.13
Gender b -0.452 0.499 0.821 0.636 [0.240, 1.691] .365 0.13
Gender x SCI factor score b 0.021 0.045 0.209 1.021 [0.934, 1.116] .647 0.13
FD factor score c -0.001 0.030 0.002 0.999 [0.941, 1.059] .965 0.16
Gender c -0.392 0.454 0.745 0.676 [0.278, 1.645] .388 0.16
Gender x FD factor score c 0.020 0.037 0.296 1.020 [0.949, 1.097] .587 0.16
C factor score d -0.011 0.033 0.113 0.989 [0.928, 1.054] .736 0.14
Gender d -0.398 0.466 0.729 0.672 [0.269, 1.675] .393 0.14
Gender x C factor score d 0.032 0.052 0.397 1.033 [0.934, 1.143] .529 0.14
Note. SE = standard error; eβ = univariate transformation of the estimated Beta coefficient, which
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991).
a Model -2LL = 164.425, χ2 (df = 3, n = 58) = 1.343, p = .719.
b Model -2LL = 164.866, χ2 (df = 3, n = 58) = 0.949, p = .814.
c Model -2LL =164.310, χ2 (df = 3, n = 58) = 1.433, p = .698.
d Model -2LL =164.679, χ2 (df = 3, n = 58) = 1.123, p = .772.
96
as using a cut off score of 45 on IR40. This was done as power for the moderation
analyses was lower than determined by the a-priori power calculations due to
invalid/missing data.
Table 37. Bivariate Correlations between PPI–R Total, Factor and Scale scores
and Overall Recidivism
Recidivism
Entire Sample IR40 < 45
PPI–R total score .284** .317**
SCI factor score .305** .333**
FD factor score .096 .142
C factor score .049 .061
ME content scale score .278** .322**
RN content scale score .190* .245*
BE content scale score .216* .239*
CN content scale score .199* .164
SOI content scale score .042 .067
F content scale score .155 .158
STI content scale score .009 .060
*p ≤ .05, two-tailed. **p ≤ .01, two-tailed.
97
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
SCI factor score e 0.041 0.013 9.845 1.042 [1.015, 1.069] .002 0.38
FD factor score e 0.008 0.011 0.512 1.008 [0.986, 1.030] .474 0.38
C factor score e 0.000 0.012 0.000 1.000 [0.976, 1.025] .987 0.38
ME content scale score f 0.034 0.017 4.260 1.035 [1.002, 1.070] .039 0.39
RN content scale score f 0.004 0.013 0.093 1.004 [0.979, 1.029] .760 0.39
BE content scale score f 0.012 0.016 0.568 1.012 [0.981, 1.043] .451 0.39
CN content scale score f 0.008 0.015 0.262 1.008 [0.978, 1.038] .609 0.39
Note. SE = standard error; eβ = univariate transformation of the estimated Beta coefficient, which
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991)
a Model -2LL = 644.773, χ2 (df = 1, n = 112) = 11.305, p = .001.
b Model -2LL = 641.556, χ2 (df = 3, n = 112) = 13.728, p = .003.
c Model -2LL = 650.759, χ2 (df = 4, n = 113) = 14.083, p = .007.
d Model -2LL = 549.971, χ2 (df = 1, n = 99) = 13.071, p < .001.
e Model -2LL = 548.212, χ2 (df = 3, n = 99) = 14.216, p = .003.
f Model -2LL = 547.554, χ2 (df = 4, n = 99) = 14.906, p = .005.
Table 39. Prospective Validity of the PPI–R vis-à-vis Type of Post Release
Offense(s)
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
Entire Sample
Violent offenses
PPI–R total a 0.033 0.013 6.341 1.034 [1.007, 1.061] .012 0.24
SCI factor score b 0.046 0.020 5.531 1.047 [1.008, 1.089] .019 0.25
FD factor score b 0.002 0.016 0.020 1.002 [0.972, 1.033] .886 0.25
C factor score b 0.005 0.017 0.099 1.005 [0.972, 1.040] .754 0.25
ME content scale score c 0.045 0.024 3.533 1.046 [0.998, 1.096] .060 0.29
RN content scale score c -0.002 0.018 0.009 0.998 [0.964, 1.034] .925 0.29
BE content scale score c 0.023 0.022 1.032 1.023 [0.979, 1.069] .310 0.29
CN content scale score c -0.002 0.022 0.007 0.998 [0.955, 1.043] .931 0.29
Non-Violent offenses
PPI–R total d 0.013 0.022 0.349 1.013 [0.970, 1.059] .555 0.06
SCI factor score e 0.043 0.031 1.907 1.044 [0.982, 1.111] .167 0.16
FD factor score e -0.026 0.027 0.932 0.974 [0.925, 1.027] .334 0.16
C factor score e 0.004 0.026 0.022 1.004 [0.954, 1.056] .883 0.16
98
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
ME content scale score f 0.012 0.037 0.109 1.012 [0.941, 1.089] .742 0.20
RN content scale score f -0.020 0.026 0.605 0.980 [0.931, 1.031] .437 0.20
BE content scale score f -0.005 0.033 0.021 0.995 [0.933, 1.061] .884 0.20
CN content scale score f 0.063 0.033 3.661 1.065 [0.998, 1.137] .056 0.20
Other offenses
PPI–R total g 0.031 0.013 5.206 1.031 [1.004, 1.058] .023 0.22
SCI factor score h 0.039 0.019 4.040 1.040 [1.001, 1.080] .044 0.23
FD factor score h 0.010 0.016 0.383 1.010 [0.979, 1.041] .536 0.23
C factor score h -0.002 0.018 0.014 0.998 [0.964, 1.033] .906 0.23
ME content scale score i 0.018 0.024 0.566 1.018 [0.972, 1.066] .452 0.22
RN content scale score i 0.009 0.018 0.248 1.009 [0.974, 1.046] .618 0.22
BE content scale score i 0.013 0.022 0.321 1.013 [0.969, 1.058] .571 0.22
CN content scale score i 0.022 0.022 0.987 1.022 [0.979, 1.067] .320 0.22
IR40 < 45
Violent offenses
PPI–R total j 0.038 0.013 7.976 1.038 [1.012, 1.066] .005 0.29
SCI factor score k 0.045 0.020 5.001 1.046 [1.006, 1.088] .025 0.29
FD factor score k 0.011 0.017 0.476 1.012 [0.979, 1.045] .490 0.29
C factor score k 0.002 0.019 0.011 1.002 [0.966, 1.040] .915 0.29
ME content scale score l 0.045 0.026 3.064 1.046 [0.995, 1.099] .080 0.32
RN content scale score l 0.002 0.019 0.014 1.002 [0.965, 1.041] .906 0.32
BE content scale score l 0.028 0.024 1.337 1.028 [0.981, 1.078] .248 0.32
CN content scale score l -0.012 0.024 0.267 0.988 [0.943, 1.035] .606 0.32
Non-Violent offenses
PPI–R total m 0.021 0.023 0.837 1.021 [0.976, 1.068] .360 0.10
SCI factor score n 0.048 0.032 2.241 1.049 [0.985, 1.117] .134 0.16
FD factor score n -0.015 0.028 0.292 0.985 [0.933, 1.040] .589 0.16
C factor score n -0.001 0.029 0.002 0.999 [0.944, 1.056] .967 0.16
ME content scale score o 0.028 0.042 0.440 1.028 [0.947, 1.116] .507 0.20
RN content scale score o -0.013 0.028 0.219 0.987 [0.934, 1.043] .640 0.20
BE content scale score o 0.003 0.036 0.006 1.003 [0.935, 1.076] .940 0.20
CN content scale score o 0.051 0.036 2.004 1.053 [0.980, 1.130] .157 0.20
99
Variable B SE (B) Wald eB (95% CI) p-value Model Φ
Other offenses
PPI–R total p 0.030 0.014 4.537 1.030 [1.002, 1.059] .033 0.22
SCI factor score q 0.033 0.020 2.579 1.033 [0.993, 1.075] .108 0.22
FD factor score q 0.013 0.017 0.564 1.013 [0.979, 1.048] .453 0.22
C factor score q -0.001 0.020 0.002 0.999 [0.960, 1.039] .962 0.22
ME content scale score r 0.025 0.026 0.872 1.025 [0.973, 1.079] .350 0.22
RN content scale score r 0.012 0.021 0.339 1.012 [0.972, 1.054] .561 0.22
BE content scale score r -0.001 0.025 0.003 0.999 [0.950, 1.050] .959 0.22
CN content scale score r 0.010 0.024 0.171 1.010 [0.964, 1.058] .679 0.22
Note. SE = standard error; eB = univariate transformation of the estimated Beta coefficient, which
represents the hazard ratio; CI = confidence interval. The model phi coefficient is provided as an
effect size estimate for the predictive strength of the overall model. It was derived based on the
following formula: Φ = √ (χ2 / N), (Rosenthal, 1991).
a Model -2LL = 267.996, χ2 (df = 1, n = 112) = 6.418, p = .011.
b Model -2LL = 266.837, χ2 (df = 3, n = 112) = 7.169 p = .067.
c Model -2LL = 264.811, χ2 (df = 4, n = 113) = 9.481, p = .050.
d Model -2LL = 103.811, χ2 (df = 1, n = 99) = .349, p = .555.
e Model -2LL = 101.609, χ2 (df = 3, n = 99) = 2.452, p = .484.
f Model -2LL = 109.061, χ2 (df = 4, n = 100) = 4.072, p = .396.
g Model -2LL = 272.330, χ2 (df = 1, n = 112) = 5.262, p = .022.
h Model -2LL = 271.523, χ2 (df = 3, n = 112) = 5.859, p = .119.
I Model -2LL = 271.833, χ2 (df = 4, n = 113) = 5.657, p = .226.
j Model -2LL = 233.219, χ2 (df = 1, n = 99) = 8.164, p = .003.
k Model -2LL = 232.656, χ2 (df = 3, n = 99) = 8.413, p = .038.
l Model -2LL = 230.511, χ2 (df = 4, n = 99) = 10.320, p = .035.
m Model -2LL = 91.615, χ2 (df = 1, n = 90) = .840, p = .359.
n Model -2LL = 89.977, χ2 (df = 3, n = 90) = 2.322, p = .508.
o Model -2LL = 88.946, χ2 (df = 4, n = 90) = 3.535, p = .473.
p Model -2LL = 224.690, χ2 (df = 1, n = 98) = 4.602, p = .032.
q Model -2LL = 224.417, χ2 (df = 3, n = 98) = 4.769, p = .189.
r Model -2LL = 224.486, χ2 (df = 4, n = 98) = 4.767, p = .312.
Results revealed similar predictive patterns for the PPI–R scores as reported in
the main analyses. Specifically, all of the final models were statistically significant. In
addition, the total score was predictive of recidivism in both analyses—with the entire
sample (i.e., HR = 1.029, p = .001, Wald = 11.207), and in a sub-sample where the IR40
score < 45 inclusion criterion was implemented (i.e., HR = 1.032, p < .001, Wald =
12.872). With regard to the factor scores, only the SCI factor was significantly predictive
of recidivism, which was the case in both analyses: HR = 1.044, p = .001, Wald = 11.589
100
(i.e., entire sample), and HR = 1.042, p = .002, Wald = 9.845 (i.e., sub-sample with IR40
scores < 45). Of the content scale scores, only ME was significantly predictive of
recidivism when a cut off of 45 for the IR40 scale was used (i.e., HR = 1.035, p = .039,
Wald = 4.260).
With regard to the type of offense(s), based on the entire sample, the final
models were significant for the PPI–R total and scale scores for violent offenses, and for
the PPI–R total score for other offenses. In addition, the PPI–R total and SCI factor
scores were predictive of violent and other offenses (i.e., [1] for the total score: HR =
1.034, p = .012, Wald = 6.341, and HR = 1.031, p =.023, Wald = 5.206 respectively; and
[2] for the SCI factor score: HR = 1.047, p = .019, Wald = 5.531, and HR = 1.040, p =
.044, Wald = 4.040 respectively). Further, based on a sub-sample where the IR40 score
< 45 inclusion criteria was implemented, the final models were significant for all of the
PPI–R scores with respect to violent offenses, as well as for the PPI–R total score with
regard to other offenses. Also, the PPI–R total score was predictive of violent and other
offenses (i.e., HR = 1.038, p = .005, Wald = 7.976, and HR = 1.030, p = .033, Wald =
4.537 respectively), while the SCI factor score was predictive of violent recidivism (i.e.,
HR = 1.046, p = .025, Wald = 5.001).
Similar to the survival analyses, the logistic regression analyses used to evaluate
the predictive validity of the PPI–R vis-à-vis recidivism over the 6-month follow-up period
were also repeated while less stringent inclusion criteria were used (i.e., by including the
entire sample, as well as profiles that had IR40 scores < 45). The patterns of
association between the PPI–R scores and recidivism within the 6-month follow-up
period were fairly similar to those reported above. Namely, the total score, the SCI
factor score, and three of the content scale score (i.e., ME, RN, and BE) were
significantly correlated with recidivism at the univariate level in both analyses (i.e., the
entire sample, and the sub-sample with IR40 scores < 45). Detailed information is
provided in Table 40.
101
Table 40. Univariate Logistic Regression -- Odds Ratios for the Associations
between PPI–R Scores and Recidivism over the 6-Month Follow-Up
Period
Variable 0 to 6 months (OR)
Entire Sample
PPI–R total score 1.060**
SCI factor score 1.068**
FD factor score 1.019
C factor score 1.027
ME content scale score 1.075**
RN content scale score 1.037*
BE content scale score 1.050*
CN content scale score 1.029
SOI content scale score 1.022
F content scale score 1.027
STI content scale score 0.990
IR40 < 45
PPI–R total score 1.063**
SCI factor score 1.067**
FD factor score 1.033
C factor score 1.030
ME content scale score 1.080**
RN content scale score 1.042*
BE content scale score 1.048*
CN content scale score 1.021
SOI content scale score 1.031
F content scale score 1.030
STI content scale score 1.004
*p ≤ .05, two-tailed. **p ≤ .01, two-tailed.
At the multivariable level, results were significant for the PPI–R total score, SCI
factor score, and ME content scale score in both analyses, indicating that those scores
could be used to distinguish between individuals who recidivated versus those who did
not. Detailed information on the multivariable logistic regression analyses is presented
102
in Table 41. Specifically, when the entire sample was, used all of the final models were
significant, and the Nagelkerke R-squared was modest to moderate across the models,
showing that the variance explained ranged from 12% to 14%. Results for the total
score were: χ2 = 11.466, p = .001, df = 1, and the prediction accuracy of the overall
model was 64% (74% for individuals who recidivated and 52% for those who did not).
For the SCI factor: χ2 = 10.384, p = .001, df = 1, with prediction accuracy for the overall
model at 66% (81% for those who recidivated, and 48% for those who did not). Results
for the ME content scale score were as follows: χ2 = 12.392, p = .006, df = 3, with 63%
prediction accuracy of the overall model (76% for those who recidivated and 46% for
those who did not).
103
Further, when a cut off score of 45 for the IR40 scale was used, all of the final
models were once again significant, with a modest to moderate Nagelkerke R-squared
across the models, showing that the variance explained ranged from 12% to 16%.
Results for the total score were: χ2 = 11.962, p = .001, df = 1, and the prediction
accuracy of the overall model was 67% (76% for individuals who recidivated and 56% for
those who did not). Results for the SCI factor score were as follows: χ2 = 9.564, p =
.002, df = 1, with 66% prediction accuracy for the overall model (i.e., 80% for those who
recidivated and 49% for those who did not). Finally, results for the ME content scale
score were: χ2 = 12.576, p = .006, df = 3, and the prediction accuracy for the overall
model was 65% (74% for those who recidivated and 53% for those who did not).
Next, with regard to the moderating effects of gender, bivariate correlations were
carried out first to evaluate the association between the PPI–R total and factor scores,
as well as gender, and recidivism across the entire sample. The results revealed very
similar patterns of associations as in the main analyses. Namely, the PPI–R total and
SCI factor scores were correlated positively with recidivism (i.e., r = .284, p = .002; and r
= .305, p = .001 respectively). In addition, gender was correlated positively with the
PPI–R total, FD and C factor scores (i.e., r = .282, p = .003; r = .321, p = .001; and r =
.281, p = .003 respectively), but there were no significant correlations between gender
and the SCI factor score (i.e., r = .073, p = .440) or recidivism (i.e., r = -.122, p = .177).
The next step involved hierarchical survival analysis using Cox model to
determine whether gender moderated the predictive utility of the PPI–R total score vis-à-
vis crime and violence over the entire follow-up period. As stated previously, gender
was dummy coded, where males served as the reference group (i.e., male = 0, female =
1); thus, the results are presented from females’ perspective. Results revealed patterns
of associations that were nearly identical to the main analyses when the PPI–R total
score and gender were entered as predictors (i.e., the only difference was the
emergence of a main effect for gender, which was not detected in the main analyses).
Specifically, the overall model was significant (i.e., Model -2LL = 639.772, χ2 (df = 3, n =
112) = 17.206, p = .001), and the PPI–R total score and gender were predictive of future
crime and violence (HR = 1.030, p = .023, Wald = 5.193; HR = 1.617, p = .048, Wald =
0.629 respectively), but no moderating effects were found (HR = 1.012, p = .510, Wald =
0.435. As for the SCI factor score, results revealed patterns of associations that were
104
identical to the main analyses. Namely, while the overall model was significant (i.e.,
Model -2LL = 649.126, χ2 (df = 3, n = 113) = 15.998, p = .001), and the SCI factor score
was a significant predictor of crime and violence (HR = 1.036, p = .037, Wald = 4.353),
gender was not (HR = 1.330, p = .226, Wald = 1.464), and there were no moderating
effects (HR = 1.020, p =.397, Wald = 0.718).
The bivariate patterns of associations were identical to the ones based on the
entire sample when a cut off score of 45 on IR40 was used. Specifically, there were
significant positive correlations between the PPI–R total and SCI factor scores and
recidivism (i.e., r = .317, p = .001; and r = .333, p = .001 respectively). Further, gender
was correlated positively with the PPI–R total, FD and C factor scores (i.e., r = .299, p =
.003; r = .355, p < .001; and r = .288, p = .004 respectively). There were no significant
correlations between gender and the SCI factor score (i.e., r = .102, p = .315) or
recidivism (i.e., r = -.108, p = .288). As for the hierarchical survival analyses, the
patterns of associations were identical to the main analyses. Specifically, when the PPI–
R total score and gender were entered as predictors, the overall model was significant
(i.e., Model -2LL = 548.033, χ2 (df = 3, n = 99) = 15.055, p = .002) and the PPI–R total
score was a significant predictor of crime and violence (HR = 1.032, p = .015, Wald =
5.967), however, gender was not (HR = 1.356, p = .239, Wald = 1.384), and there were
no moderating effects (HR = 1.010, p =.603, Wald = 0.271). Similarly, with the SCI
factor and gender as predictors, the overall model was significant (i.e., Model -2LL =
548.051, χ2 (df = 3, n = 99) = 13.763, p = .003) and the SCI factor score was a
significant predictor of crime and violence (HR = 1.041, p = .018, Wald = 5.641),
however, gender was not (HR = 1.148, p = .585, Wald = 0.298), and there were no
moderating effects (HR = 1.012, p =.626, Wald = 0.238).
105
4. Discussion
This study was designed with a few goals in mind. First, given that the PPI–R is
a relatively new measure, evaluating its reliability in a correctional sample was deemed
important as a way of expanding existing research, especially since it was originally
developed for use with non-criminal populations. Further, examining the predictive
validity of the PPI-R was considered essential in and of itself as there are currently no
studies that have focused on this aspect of its psychometric properties. This in turn
could have practical implications specifically relevant to correctional settings and could
speak to the utility of the PPI–R.
Another goal for this study was to evaluate the performance of the PPI–R vis-à-
vis future offending in comparison to the PAI and the NEO–FFI, which are general
measures of personality. Specifically, prior research has demonstrated that some of the
general measures of personality tend to assess non-specific behavioral deviance, and
are not particularly useful in assessing personality features of psychopathy. As a result,
they cannot be used to make a differential diagnosis of psychopathy in correctional
settings (see Sandoval et al., 2000). The use of a measure such as the PPI–R, which
addresses both, the personality and behavioral aspects of psychopathy, may make the
assessment and diagnosis of psychopathy more reliable. Therefore, it was deemed
important to evaluate whether the PPI–R (as opposed to the PAI or NEO–FFI) captures
unique variance with regard to future offending, hence performs better than a general
measure of personality, as that would indicate that it could be used with confidence to
identify individuals who are at high risk to engage in institutional infractions or recidivate
once released from custody.
In addition, the CITC values were also examined to determine whether all items
included in the PPI–R scales were representative of the construct. It appeared that the
removal of a few items would have led to a slight improvement of the internal
consistency of their respective scale. Nevertheless, the amount of improvement was
considered to be negligible, and therefore it was concluded that based on this sample’s
results, all of the items included in the PPI–R are representative of psychopathy.
107
4.2. Research Question 2:
Predictive Validity of the PPI–R vis-à-vis Crime and
Violence
This study found partial support for the predictive validity of the PPI–R.
Specifically, the results from the survival analyses were in line with the hypothesis and
yielded support for the predictive validity of the PPI–R total and SCI factor scores vis-à-
vis recidivism, where a point increase in the total or SCI scores was associated with a
3% or 4% increased likelihood of recidivism respectively. There was no support,
however, for the predictive validity of the FD and C factor, or any of the content scales
vis-à-vis recidivism. Further, there was support for the predictive utility of the PPI–R
total score with regard to violent offenses, where a point increase in the total score was
associated with a 4% increased likelihood of violent recidivism. Nevertheless, none of
the PPI–R factor or content scales evidenced predictive validity with regard to the type of
the offense(s). It is worth pointing out, however, that even though non-significant, there
appeared to be some interesting trends in terms of the association between the PPI–R
scores and type of the offense(s). Specifically, while the associations between the PPI–
R scores and violent offenses were positive, they were negative with regard to non–
violent offenses, except for the SCI factor score. Given no interpretation of those
patterns of associations can be offered here as they were non-significant, it might be
worth exploring whether gender composition of the sample plays a role, as the literature
has shown that men and women tend to engage in different types of violence or crime
(see Robbins et al., 2003; Skeem et al., 2011). There was no support for the utility of
the PPI–R in differentiating between the type of post release offense(s) among
individuals who recidivated.
Results from the logistic regression analyses, which were used to evaluate the
predictive validity of the PPI–R over a set follow-up period (i.e., 6 months), were similar
to those from the survival analyses. Namely, while the PPI–R total, SCI and FD factor,
and ME and RN scale scores were all significant predictors of recidivism at the
univariate level, only the PPI–R total and the SCI factor scores remained significant at
the multivariable level.
108
Given there are currently no published studies on the predictive validity of the
PPI–R specifically with regard to recidivism, there is no direct basis for comparison of
these results. Nevertheless, findings from this study regarding the relationship between
PPI–R-assessed psychopathy and recidivism in general are in line with prior research
(see Douglas et al., 2006). In addition, as expected (see Hemphill, Hare, et al., 1998;
Wormith, Olver, Stevenson, & Girard, 2007), the SCI (which corresponds to Factor 2 of
the PCL–R and reflects behavioral aspects of psychopathy) was found to be a stronger
predictor of future crime and violence than the FD factor, which was significant only at
the univariate level. Needless to say, these results need to be replicated before any
conclusions about their generalizability could be made.
All in all, these findings are promising in terms of potential practical implications
for the utility of the PPI–R, especially in correctional settings. Specifically, if
implemented, PPI–R scores could be used as a guide to treatment planning, program
participation, and case management of offenders while in custody, as well as following
their release into the community. Such findings are of course crucial as they support the
use of a measure that could reveal information regarding the level of recidivism risk of
offenders. This in turn could have a significant positive impact on the safety and security
of staff and inmates at correctional facilities, which are of utmost importance.
109
recidivism (i.e., it did not provide any unique information that could be useful in
predicting recidivism) relative to risk-relevant indices from general self-report measures
of personality traits and patterns of behavior. While these findings may be
generalizable, alternatively they might be due to specific characteristic of this sample, or
as a result of different operational definitions used in the measures. It is also important
to highlight that medium effect sizes were expected and used as a guide in the a-priori
sample size calculations, as if detected, they would yield meaningful results. Yet, this
may not have been realistic given research has shown that incremental validity effect
sizes are generally small, and therefore require large sample sizes in order to be
detected (see Hunsley & Meyer, 2003). Thus, the sample size of this study alone could
have been the reason for the lack of significant findings with regard to the incremental
validity of the PPI-R. There are currently no other studies that have examined the
incremental validity of the PPI–R, which calls for replication of the analyses carried out
here.
Nevertheless, there was no support for the incremental validity of the PPI–R
relative to the CRNA in relation to recidivism risk. This indicates that using the PPI–R in
addition to the CRNA in an assessment for the risk of recidivism would not add unique
information above and beyond what the CRNA offers. Yet, the PPI–R might be useful to
gain information regarding personality traits and motivations for engaging in criminal
activity. This in turn might be helpful in detecting the challenges in managing the
individual and determining appropriate ways of doing so. Given that the sample size
110
used here was relatively small, no conclusions can be made as to whether these
findings are generalizable to other samples or settings.
111
4.4.2. Does Gender Moderate the Predictive Utility of the PPI–R
with Respect to Recidivism?
Results revealed some significant associations between PPI–R scores, gender,
and recidivism at the bivariate level. Specifically, there were significant positive
correlations between the PPI–R total and SCI factor scores and recidivism. In addition,
gender was correlated positively with the PPI–R total, FD and C factor scores, but there
were no significant correlations between gender and the SCI factor score, recidivism, or
type of offense(s). As mentioned previously, there is currently no direct basis for
comparison for these results as no previous studies validating the PPI–R have
addressed these very questions. Nevertheless, studies that have examined the
association between psychopathy (assessed with measures such as the PCL–R or the
PPI) and recidivism revealed similar results (i.e., there appears to be a well-established
relationship between psychopathy and recidivism). In addition, the current findings are
in line with studies that have shown a stronger relationship between future crime and the
behavioral aspects of psychopathy (which are captured by the SCI factor of the PPI–R),
rather than the interpersonal aspects of the disorder (see Douglas et al., 2006; Yang et
al., 2010). With regard to the association between gender and psychopathy, once again,
there is no direct basis for comparison, as prior studies have examined that question in
either exclusively male or exclusively female samples (rather than mixed samples), and
have yielded inconsistent results.
In terms of the moderation analyses, the PPI–R total score was predictive of
future crime and violence in one of the prediction models, and the SCI factor score was
found to be a significant predictor in another prediction model. These findings indicate
that recidivism rates increase as PPI–R total and SCI factor scores increase, which as
discussed above, is generally consistent with previous literature (see Douglas et al.,
2006). Nevertheless, contrary to what was expected, there were no main or moderating
effects of gender on the relationship between PPI–R-assessed psychopathy and
recidivism. This implies that the PPI–R was not differentially predictive of recidivism
across gender in this sample. It is also important to keep in mind that medium
interaction effect sizes were expected given the associations between psychopathy and
recidivism reported in the literature are low for women and moderate to large for men.
Nevertheless, that may not have been realistic as effect sizes for interactions are
112
generally small, and therefore require a relatively large sample size in order to be
detected (see Frazier, 2004). Thus, potentially existing moderating effects of gender on
the predictive utility of the PPI-R relative to recidivism may not have been detected
simply due to the relatively small sample utilized in the current study. Naturally,
replication of the analyses conducted in this study is needed to determine the
generalizability of these findings.
113
for gender was found in the presence of the C factor score, where males were more
likely to commit non-violent offenses than females. This pattern of associations is
generally consistent with previous research which has shown that even though
psychopathic women tend to be involved in chronic non-violent criminality (see Warren
et al., 2003; Warren & South, 2006), men overall exhibit higher rates of antisocial
behavior and criminality (see Hicks et al., 2011; Stanford & Felthous, 2011).
Further, for the individuals who committed violent post-release offense(s) the SCI
factor score was found to be a significant predictor of violent offenses when entered in a
model with gender, where gender significantly moderated the relationship between the
SCI factor score and violent offenses. The pattern of associations revealed in this model
was interesting. Specifically, males and females who scored high on the SCI factor did
not differ in terms of their commission of violent offenses, which is generally in line with
findings by Edens et al. (2008) who reported strong associations between Factor II of the
PPI (corresponding to the SCI factor of the PPI–R) and aggressive misconduct.
Nevertheless, females who scored low on the SCI factor had low rates of committing
violent offenses, while the opposite was true for males. This finding is consistent with
the conception that while males are more likely to express psychopathy through physical
aggression, psychopathic females tend to engage in relational aggression such as back-
stabbing, manipulation, rumor spreading, etc. (Nicholls & Petrila, 2005; Strand &
Belfrage, 2005; Verona & Vitale, 2006). There were no other significant moderating
effects of gender on the utility of the PPI–R scores to differentiate between the type of
offense(s) committed by individuals who recidivated in this sample. In light of that, it is
possible that the significant moderating effects of gender on the relationship between the
SCI factor score and violent offenses that emerged here were specific to the subsample
used or were due to chance.
All in all, despite the lack of moderating effects of gender on the predictive
validity of the PPI–R with regard to recidivism and type of offense(s), gender differences
were detected in most of the PPI–R scores, and moderating effects of gender emerged
in a subsample that included only those who recidivated. This suggests that
psychopathy likely manifests differently across gender, which is in line with existing
literature. It is important to reiterate, however, that due to invalid/missing data the final
sample included in the analyses was lower than the original sample, which led to lower
114
power for the moderation analyses in particular. This in turn resulted in restricted ability
to fully address the moderating effects of gender on PPI–R-assessed psychopathy with
regard to recidivism and type of the offense(s). Thus, even though supplemental
analyses were ran on the entire sample, as well as after implementing less stringent
exclusion criteria for PPI–R scores in order to compensate for the loss in power, the
results presented here should be considered preliminary at best, and need to be
interpreted with extra caution.
In addition, given the lack of prior research on the questions evaluated in this
study, further exploration is warranted. Looking into the context of recidivism (i.e., time
since release from custody, degree of premeditation, relationship to the victim(s),
motivation for the offense, the presence of supervisory release condition that might have
been violated, etc.), as well as the type of offense(s) (i.e., violent vs. non-violent; or
category of the offense [i.e., against a person, property, drug offenses, traffic infractions,
etc.]), could shed some light on the factors driving such behavior. Finally, personality
and behavioral correlates or co-morbid disorders could also prove useful in terms of
helping us better understand what is driving the differences in the manifestation of
psychopathy across gender.
There were differences from the main analyses with regard to the association
between psychopathy and type of the offense(s) as well. Specifically in addition to the
main findings, there were a couple more significant final models (i.e., [1] when the entire
sample was used, the models of the scale scores with regard to violent offenses, and the
total score with respect to other offenses were significant; and [2] when the IR40 < 45
inclusion criteria was implemented, the final models for the factor and scale scores with
regard to violent offenses, and the total score with regard to other offenses were
significant). Further, the SCI factor score significantly predicted violent and other
offenses based on the entire sample, as well as violent offenses, based on the IR40 <
45 inclusion criteria. Additionally, the total score was found to be a significant predictor
of other offenses in both analyses (i.e., based on the entire sample, as well as when the
IR40 < 45 inclusion criteria was implemented).
Finally, with regard to the gender moderation analyses where recidivism was of
interest, a main effect for gender emerged in the presence of the PPI–R total score as a
predictor in a model based on the entire sample. Such main effect was not detected in
the main analyses. No other differences were detected in terms of the patterns of
associations based on the entire sample or on a subsample where a score of 45 on the
IR40 was used as a cut off.
Overall, even though there was sufficient statistical power, results revealed a few
additional patterns of association between the PPI–R and recidivism or type of the
offense(s), and therefore it is possible that significant associations were not detected
simply due to the relatively small sample size. Thus, replication of this study is highly
recommended in order to determine whether the present findings are generalizable
outside of this sample.
116
4.6. Limitations
While this study answered some very important questions about PPI–R-
assessed psychopathy, it is not free of limitations. First, even though power was
adequate for most of the analyses, the sample size was overall relatively small. This
was particularly relevant in terms of assessing gendered effects, as those analyses were
underpowered by the reduction in sample size due to invalid/missing data, and therefore
caution is warranted when interpreting the results. While an attempt to compensate for
the loss in power was made by running supplemental analyses, the reduction in power
might have stymied the detection of additional gender differences. Therefore, larger
samples are recommended for future research, as they represent the associations
between variables well, allowing the detection of any meaningful relationships that might
exist, and thus should always be preferred.
Further, individuals who were acutely psychotic at the time of recruitment or had
a chart diagnosis of mental retardation were not included in the present study, which
was deemed reasonable, as their ability to provide valid responses to the PPI–R would
have been questionable. Yet, such individuals are often at a heightened risk for violence
(see Robinson, Littrell, & Littrell, 1999; Rueve & Welton, 2008), and therefore it might be
worth exploring whether the present findings would generalize to individuals suffering
from psychotic symptoms or mental retardation.
Finally, recidivism data in this study was based solely on official records.
Although this is a common practice in psychology research and in the risk assessment
field in particular, it ultimately underestimates the true base rates of recidivism. This
issue was further exacerbated by the fact that the official records that were collected
reflected criminal involvement within the Province of British Columbia, rather than nation-
wide criminal activity. The shortcomings associated with such an approach are
intensified in terms of detecting violence perpetrated by females as they typically engage
in violence in familial settings, and as a result, may be less likely to be reported and/or
investigated by police (see Kreis, 2009; Kreis & Cooke, 2011; Logan, 2009; Nicholls &
Petrila, 2005). Nevertheless, as previously stated, this is one of the first studies to
examine the predictive validity of the PPI–R, as well as the moderating effects of gender
on PPI–R-assessed psychopathy, and therefore this feature of the design was deemed
justifiable. It is also worth pointing out that despite this limitation, several significant
patterns of association were still detected in the present study. Even so, future research
may benefit from improving the recidivism data collection procedures by incorporating
nation-wide official criminal records, as well as unofficial sources such as self-report and
collateral information.
118
4.7. Future Research
It is important to reiterate that given this is one of the first studies to evaluate the
predictive validity of PPI–R, and particularly so in a mixed sample, none of the questions
addressed should be considered settled. For most of the research questions that were
addressed, there were no pre-existing studies to serve as a direct basis of comparison in
order to form conclusions regarding the stability and generalizablity of the current
findings. Therefore, to determine whether these findings would apply to different
samples or settings, direct replications of the current study, as well as further research
with improved study designs to address the current limitations are strongly encouraged.
Specifically, using a larger sample, breaking down offenses by category (i.e., against a
person, property, drug offenses, traffic infractions, etc.), including various personality or
behavioral correlates, and/or co-morbid diagnosis, incorporating additional psychopathy
measures, expanding the recidivism data collection sources, including various contextual
factors (i.e., motivation, relationship to the victim, etc.), among others, could be useful in
terms of further exploring the psychometric properties of the PPI–R, as well as shedding
some light on recidivism patterns.
The design and findings of this study raised a number of additional research
questions that are worth examining. For instance, participants’ confidentiality was
assured while they were filling out the measures utilized in this study. This poses the
question whether participants would have provided different responses if they were
completing the measures as part of a routine intake screen at a secure facility, which is
typically non-confidential in nature. Exploring this question would help determine
whether there is a negative impact on the percent of valid profiles provided by
participants as a function of limiting their confidentiality, which in turn might have
practical implications for the implementation of assessment instruments such as the
PPI–R. Namely, if lack of confidentiality leads to a high number of invalid profiles, it
might be impractical to implement psychopathy measures such as the PPI-R in
consideration of the time and financial resources they would require.
In addition, data on institutional infractions was not available, and therefore, not
incorporated as an outcome in the present study. Yet, exploring the predictive utility of
the PPI–R with regard to institutional infractions would be of utmost practical importance
119
for secure facilities. Specifically, given the safety and security of the institution in such
settings comes first, it may be beneficial to implement a measure such as the PPI–R,
which in addition to assessing psychopathy (which as discussed is of relevance to
treatment planning and management), could be useful in estimating whether an
individual is at a high risk of engaging in institutional infractions.
Further, the incremental validity of a measure can have multiple dimensions and
therefore can be assessed in many ways. This study examined the incremental
predictive validity of the PPI–R relative to risk-relevant indices incorporated in general
measures of personality traits, and relative to a risk/needs assessment instrument. As
this approach is useful in addressing questions particularly pertinent to violence risk
assessment, it seemed to fit well within the milieu of the current study. It is essential,
however, to extend this line of research and evaluate the incremental validity of the PPI–
R relative to theoretically, clinically, and legally relevant risk factors such as anxiety or
nervousness (or lack thereof), child abuse history, treatment amenability, malingering, or
feigning psychopathology and cognitive impairment.
120
4.8. Summary and Implications
In summary, this study has some important implications. First, it yielded support
for the internal consistency of the PPI–R, thus furthering the existing literature on its
reliability when used in a correctional sample. It addition, it is among the first to evaluate
and to show support for the predictive validity of at least some of the PPI–R scores vis-à-
vis recidivism. Nevertheless, this study found no support for the predictive utility of the
PPI–R scores with regard to the type of offense(s), for the incremental validity of the
PPI–R relative to general measures of personality in terms of predicting future violence
or crimes, or for the moderating effects of gender on the PPI–R scores vis-à-vis
recidivism. Even so, it is among the first studies to examine these questions, which in
and of itself is an important contribution in terms of establishing the psychometric
properties of the instrument. It also expanded the existing literature on psychopathy in
women, and yielded support for the PPI–R score differences across gender, as well as
the moderating effects of gender on the predictive utility of the SCI factor score with
regard to violent offenses. Finally, this study has the potential to be of particular
importance when it comes to treatment and management of offenders following release
from custody, as it yielded support for the convergent validity of the PPI–R with the
CRNA, as well as the predictive validity of the CRNA itself. Providing correctional
officers with an additional tool that could help manage recidivism in general, would be
highly beneficial, and constitutes perhaps the most important contribution of this study.
All in all, it is my hope that it inspires further research into the issues addressed here, as
they have empirical as well as clinical importance.
121
References
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting
interactions. Newbury Park, CA: Sage.
Anderson, J. L., Sellbom, M., Wygant, D. B., & Edens, J. F. (2012). Examining the
necessity for and utility of the Psychopathic Personality Inventory–Revised (PPI–
R) validity scales. Law and Human Behavior. Advance online publication.
doi:10.1037/lhb0000018
Andrews, D. A., & Bonta, J. L. (1995). The Level of Service Inventory–Revised. Toronto,
ON, Canada: Multi-Health Systems.
Anestis, J. C., Caron, K. M., & Carbonell, J. L. (2011). Examining the impact of gender
on the factor structure of the Psychopathic Personality Inventory–Revised.
Assessment, 18(3), 340–349.
Babyak, M. A. (2004). What you see may not be what you get: A brief, nontechnical
introduction to overfitting in regression-type models. Psychosomatic Medicine,
66(3), 411–421.
Beltran, Y. T. P., Segura, N. A., & Bello, F. J. (2012). Synanthropy of Calliphoridae and
Sarcophagidae (Diptera) in Bogotá, Colombia. Neurological Entomology, 41(3),
237–242.
Benning, S. D., Patrick, C. J., Hicks, B. M., Blonigen, D. M., & Krueger, R. F. (2003).
Factor structure of the Psychopathic Personality Inventory: Validity and
implications for clinical assessment. Psychological Assessment, 15(3), 340–350.
Berardino, S. D., Meloy, J. R., Sherman, M., & Jacobs, D. (2005). Validation of the
Psychopathic Personality Inventory on a female inmate sample. Behavioral
Sciences & the Law, 23(6), 819–836.
122
Blonigen, D. M. (2013). Is fearless dominance relevant to the construct of psychopathy?
Reconciling the dual roles of theory and clinical utility. Personality Disorders 4(1).
87–88.
Blonigen, D. M., Carlson, S. R., Krueger, R. F., & Patrick, C. J. (2003). A twin study of
self-reported psychopathic personality traits. Personality and Individual
Differences, 35(1), 179–197.
Bobadilla, L., Wampler, M., & Taylor, J. (2012). Proactive and reactive aggression are
associated with different physiological and personality profiles. Journal of Social
and Clinical Psychology, 31(5), 458–487.
Boccaccini, M. T., Murrie, D. C., Hawes, S. W., Simpler, A., & Johnson, J. (2010).
Predicting recidivism with the Personality Assessment Inventory in a sample of
sex offenders screened for civil commitment as sexually violent predators.
Psychological Assessment, 22(1), 142–148.
Bolt, D. M., Hare, R. D., Vitale, J. E., & Newman, J. P. (2004). A multigroup item
response theory analysis of the Psychopathy Checklist–Revised. Psychological
Assessment, 16(2), 155–168.
Bottos, S. (2007). Women and violence: Theory, risk, and treatment implications.
Retrieved from Correctional Services of Canada website: http://www.csc-
scc.gc.ca/text/rsrch/reports/r198/r198-eng.shtml
Botwin, M. D., & Juni, S. (1995). [Review of the NEO Personality Inventory]. In J. C.
Conoley & J. C. Impara (Eds.), The twelfth mental measurements yearbook
[Electronic version]. Retrieved from EBSCO Mental Measurements Yearbook
database.
Boyle, G., & Kavan, M. (1995). [Review of the Personality Assessment Inventory]. In J.
C. Conoley & J. C. Impara (Eds.), The twelfth mental measurements yearbook
[Electronic version]. Retrieved from EBSCO Mental Measurements Yearbook
database.
Brand, G. L., & Fox, M. S. (2008). Child abuse: A painful secret. Retrieved from
University of Nebraska – Lincoln Extension, Institute of Agriculture and Natural
Resources Web site:
http://www.ianrpubs.unl.edu/live/g1309/build/g1309.pdf
Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self-report measurement of adult
attachment: An integrative overview. In J. A., Simpson & W. S. Rholes (Eds.),
Attachment Theory and Close Relationships (pp. 46–76). New York: Guilford
Press.
Brinkley, C. A., Diamond, P. M., Magaletta, P. R., & Heigel, C. P. (2008). Cross-
validation of Levenson’s psychopathy scale in a sample of federal female
inmates. Assessment, 15(4), 464–482. doi:10.1177/1073191108319043
123
Brinkley, C. A., Schmitt, W. A., Smith, S. S., & Newman, J. P. (2001). Construct
validation of self-report psychopathy scale: Does Levenson’s Self-Report
Psychopathy scale measure the same constructs as Hare’s Psychopathy
Checklist-Revised? Personality and Individual Differences, 31(7), 1021–1038.
Butcher, J. N., Dahlstrom, W. G., Graham, J. R., Tellegen, A. M., & Kaemmer, B.
(1989).The Minnesota Multiphasic Personality Inventory–2 (MMPI–2): Manual for
administration and scoring. Minneapolis, MN: University of Minnesota Press.
Campbell, M. A., French, S., & Gendreau, P. (2009). The prediction of violence in adult
offenders: A meta-analytic comparison of instruments and methods of
assessment. Criminal Justice and Behavior, 36(6), 567–590.
Catchpole, R. E. H., & Gretton, H. M. (2003). The predictive validity of risk assessment
with violent young offenders: A 1-year examination of criminal outcome. Criminal
Justice and Behavior, 30(6), 688–708.
Chapman, A. L., Gremore, T. M., & Farmer, R. F. (2003). Psychometric analysis of the
Psychopathic Personality Inventory (PPI) with female inmates. Journal of
Personality Assessment, 80(2), 164–172.
Clark, L. A., & Watson, D. (1995). Constructing validity: Basic issues in objective scale
development. Psychological Assessment, 7(3), 309–319.
Cleckley, H. M. (1976). The mask of sanity (5th ed.). St. Louis, MO: Mosby.
Cloninger, C. R. (1978). The link between hysteria and sociopathy: An integrative model
of pathogenesis based on clinical, genetic, and neurophysiological observations.
In H. S. Akiskal & W. L. Webb (Eds.), Psychiatric diagnosis: Explorations of
biological predictors (pp. 189–218). New York, NY: Spectrum.
Cloninger, C. R., Christiansen, K. O., Reich, T., & Gottesman, I. I. (1978). Implications of
sex differences in the prevalences of antisocial personality, alcoholism, and
criminality for familial transmission. Archives of General Psychiatry, 35(8), 941–
951. Retrieved from http://archpsyc.ama-assn.org/
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.).
Hillsdale, NJ: Lawrence Earlbaum Associates.
124
Cohen, J. (1992). A power primer. Psychological Bulletin, 112(1), 155–159.
Cooney, N. L., Kadden, R. M., & Litt, M. D. (1990). A comparison of methods for
assessing sociopathy in male and female alcoholics. Journal of Studies on
Alcohol and Drugs, 51(1), 42–48.
Cooke, D., Hart, S., & Logan, C. (2005). Comprehensive Assessment of Psychopathic
Personality Disorder: Institutional Rating Scale, Version 1.1. [Unpublished
manual].
Cooke, D. J., Hart, S. D., Logan, C., & Kreis, M. K. F. (2009, June). Recent
developments in the use of the Comprehensive Assessment of Psychopathic
Personality (CAPP). Symposium conducted at the 9th annual conference of the
International Association of Forensic Mental Health Services (IAFMHS),
Edinburgh, Scotland.
Cooke, D. J., Logan, C., Kreis, M. K. F., & Hoff, H. A. (2009, September). New
developments and new findings with the Comprehensive Assessment of
Psychopathic Personality. Symposium conducted at the 19th conference of the
European Association of Psychology and Law (EAPL), Sorrento, Italy.
Cooke, D. J., & Michie, C. (2001). Refining the construct of psychopathy: Towards a
hierarchical model. Psychological Assessment, 13(2), 171–188.
Cooke, D. J, Michie, C., & Hart, S. D. (2006). Facets of clinical psychopathy: Toward
clearer measurement. In C. J. Patrick (Ed.), Handbook of psychopathy (pp. 91–
106). New York, NY: Guilford.
Cooke, D. J., Michie, C., Hart, S. D., & Clark, D. A. (2004). Reconstructing psychopathy:
Clarifying the significance of antisocial and socially deviant behavior in the
diagnosis of psychopathic personality disorder. Journal of Personality Disorders,
18(4), 337–357.
Coolidge, F. L., Marle, P. D., Van Horn, S. A., & Segal, D. L. (2011). Clinical syndromes,
personality disorders, and neurocognitive differences in male and female
inmates. Behavioral Sciences & the Law, 29(5), 741–751.
Corrado, R. R., Watkinson, A. V., Hart, S. D., & Cohen, I. M. (2006, June).
Comprehensive Assessment of Psychopathic Personality Disorder: Initial results
with a serious and violent youth sample. Paper presented at the 6th annual
conference of the International Association of Forensic Mental Health Services
(IAFMHS), Amsterdam, the Netherlands.
125
Corrado, R. R., Watkinson, A. V., Hart, S. D., Lussier, P., & Cohen, I. M. (2007, July).
The interrater reliability of the Comprehensive Assessment of Psychopathic
Personality. Paper presented at the 3rd International Congress of Psychology and
Law, Adelaide, Australia.
Costa, P. T., Jr., & McCrae, R. R. (1992a). NEO PI–R professional manual. Odessa, FL:
Psychological Assessment Resources, Inc.
Costa, P. T., Jr., & McCrae, R. R. (1992b). Revised NEO Personality Inventory (NEO–
PI–R) and NEO Five–Factor Inventory (NEO–FFI) professional manual. Odessa,
FL: Psychological Assessment Resources.
Crime in America.net. (2010, March 15). Female involvement in crime growing: Crime
news. Retrieved from http://crimeinamerica.net/2010/03/15/female-involvement-
in-crime-growing-crime-news/
Crick, N. R., & Grotpeter, J. K. (1995). Relational aggression, gender, and social-
psychological adjustment. Child Development, 66(3), 710–722.
Cronbach, L. J., Schönemann, P., & McKie, D. (1965). Alpha coefficients for stratified-
parallel tests. Educational and Psychological Measurement, 25(2), 291–312.
Crowne, D. P., & Marlowe, D. (1960). A new scale of social desirability independent of
psychopathology. Journal of Consulting Psychology, 24(4) 349-354. doi:
10.1037/h0047358
Czar, K. A., Dahlen, E. R., Bullock, E. E., & Nicholson, B. C. (2011). Psychopathic
personality traits in relational aggression among young adults. Aggressive
Behavior, 37(2), 207–214.
Dattalo, P. (2009). A review of software for sample size determination. Evaluation & the
Health Professions, 32(3), 229–248.
Derefinko, K. J., & Lynam, D. R. (2006). Convergence and divergence among self-report
psychopathy measures: A personality-based approach. Journal of Personality
Disorders, 20(3), 261–280.
126
Diseth, Å., Danielsen, A. G., & Samdal, O. (2012). A path analysis of basic need support,
self-efficacy, achievement goals, life satisfaction and academic achievement
level among secondary school students. Educational Psychology, 32(3), 335–
354.
Dolan, M., & Doyle, M. (2000). Violence risk prediction: Clinical and actuarial measures
and the role of the Psychopathy Checklist. The British Journal of Psychiatry,
177(4), 303–311.
Dolan, M., & Völlm, B. (2009). Antisocial personality disorder and psychopathy in
women: A literature review on the reliability and validity of assessment
instruments. International Journal of Law and Psychiatry, 32(1), 2–9.
Douglas, K. S., Guy, L. S., Edens, J. F., Boer, D. P., & Hamilton, J. (2007). The
Personality Assessment Inventory as a proxy for the Psychopathy Checklist–
Revised: Testing the incremental validity and cross-sample robustness of the
Antisocial Features Scale. Assessment, 14(3), 255–269.
Douglas, K. S., Hart, S. D., & Kropp, P. R. (2001). Validity of the Personality Assessment
Inventory for forensic assessments. International Journal of Offender Therapy
and Comparative Criminology, 45(2), 183–197.
Douglas, K. S., Vincent, G. M., & Edens, J. F. (2006). Risk for criminal recidivism: The
role of psychopathy. In C. J. Patrick (Ed.) Handbook of psychopathy (pp. 533–
554). New York, NY: The Guilford Press.
Dutton, D. G. (2006). Domestic abuse assessment in child custody disputes: Beware the
domestic violence research paradigm. Journal of Child Custody, 2(4), 23–42.
Edens, J. F., Boccaccini, M. T., & Johnson, D. W. (2010). Inter-rater reliability of the
PCL–R total and factor scores among psychopathic sex offenders: Are
personality features more prone to disagreement than behavioral features?
Behavioral Sciences & the Law, 28(1), 106–119.
Edens, J. F., Campbell, J. S., & Weir, J. M. (2007). Youth psychopathy and criminal
recidivism: A meta-analysis of the Psychopathy Checklist Measures. Law and
Human Behavior, 31(1), 53–75.
Edens, J. F., Hart, S. D., Johnson, D. W., Johnson, J. K., & Olver, M. E. (2000). Use of
the Personality Assessment Inventory to assess psychopathy in offender
populations. Psychological Assessment, 12(2), 132–139.
Edens, J. F., & McDermott, B. E. (2010). Examining the construct validity of the
Psychopathic Personality Inventory–Revised: Preferential correlates of Fearless
Dominance and Self-Centered Impulsivity. Psychological Assessment, 22(1), 32–
42.
Edens, J. F., Poythress, N. G., & Watkins, M. M. (2001). Further validation of the
Psychopathic Personality Inventory among offenders: Personality and behavioral
correlates. Journal of Personality Disorders, 15(5), 403–415.
127
Edens, J. F., Poythress, N. G., Jr., Lilienfeld, S. O., & Patrick, C. J. (2008). A prospective
comparison of two measures of psychopathy in the prediction of institutional
misconduct. Behavioral Sciences & the Law, 26(5), 529–541.
Edens, J. F., & Ruiz, M. A. (2005). Personality Assessment Inventory interpretive report
for correctional settings (PAI–CS) professional manual. Odessa, FL: PAR.
Edens, J. F., Skeem, J. L., & Douglas, K. S. (2006). Incremental validity analyses of the
Violence Risk Appraisal Guide and the Psychopathy Checklist: Screening
Version in a civil psychiatric sample. Assessment, 13, 368–374.
Eisenbarth, H., Osterheider, M., Nedopil, N., & Stadtland, C. (2012). Recidivism in
female offenders: PCL–R lifestyle factor and VRAG show predictive validity in a
German sample. Behavioral Sciences & the Law, 30(5), 575–584.
Erdfelder, E., Faul, F., & Buchner, A. (1996). G*Power (Version 3.1) [Software].
Available from http://www.psycho.uni-duesseldorf.de/abteilungen/aap/gpower3
Falkenbach, D., Poythress, N., Falki, M., & Manchak, S. (2007). Reliability and validity of
two self-report measures of psychopathy. Assessment, 14(4), 341–350.
Fleiss, J. L., Williams, J. B. W., & Dubro, A. F. (1986). The logistic regression analysis of
psychiatric data. Journal of Psychiatric Research, 20(3), 195–209.
Forouzan, E., & Cooke, D. J. (2005). Figuring out la femme fatale: Conceptual and
assessment issues concerning psychopathy in females. Behavioral Sciences &
the Law, 23(6), 765–778.
Forth, A. E., Brown, S. L., Hart, S. D., & Hare, R. D. (1996). The assessment of
psychopathy in male and female noncriminals: Reliability and validity. Personality
and Individual Differences, 20(5), 531–543.
Forth, A. E., Kosson, D. S., & Hare, R. D. (2003). The Psychopathy Checklist: Youth
Version manual. Toronto, ON, Canada: Multi-Health Systems.
Frazier, P. A., Tix, A. P., & Barron, K. E. (2004). Testing moderator and mediator effects
in counseling psychology research. Journal of Counseling Psychology, 51(1),
115–134.
Gendreau, P., Goggin, C., & Smith, P. (2002). Is the PCL–R really the 'unparalleled'
measure of offender risk? A lesson in knowledge cumulation. Criminal Justice
and Behavior, 29(4), 397–426.
128
Glackman, W. (n.d.). Community Risk/Needs Assessment. Unpublished measure.
Glackman, W., Ratel, P., Swaab, S., Trytten, T., & Watts, R. (2002, November).
Predictive efficacy of the Community Risk/Needs Assessment instrument. Paper
presented at the annual meeting of the American Society of Criminology,
Chicago, IL.
Goldstein, R. B., Powers, S. I., McCusker, J., Mundt, K. A., Lewis, B. F., & Bigelow, C.
(1996). Gender differences in manifestations of antisocial personality disorder
among residential drug abuse treatment clients. Drug and Alcohol Dependence,
41(1), 35–45.
Gonsalves, V. M., Scalora, M. J., & Huss, M. T. (2009). Prediction of recidivism using the
Psychopathy Checklist—Revised and the Psychological Inventory of Criminal
Thinking Styles within a forensic sample. Criminal Justice and Behavior, 36(7),
741–756.
Gough, H. G., & Bradley, P. (1996). CPI Manual. (3rd ed.). Palo Alto, CA: Consulting
Psychologists Press.
Grimes, R. D., Lee, Z., & Salekin, R. T. (2011). Psychopathy in forensic psychology:
Assessment and methodology. In B. Rosenfeld, & S. D. Penrod (Eds.), Research
methods in forensic psychology (pp. 347–371). Hoboken, NJ US: John Wiley &
Sons Inc.
Guy, L. S., Douglas, K. S., & Hendry, M. C. (2010). The role of psychopathic personality
disorder in violence risk assessments using the HCR–20. Journal of Personality
Disorders, 24(5), 551–580.
Guy, L. S., Edens, J. F., Anthony, C., & Douglas, K. S. (2005). Does psychopathy predict
institutional misconduct among adults? A meta-analytic investigation. Journal of
Consulting and Clinical Psychology, 73(6), 1056–1064.
Haddock, C. K., Rindskopf, D., & Shadish, W. R. (1998). Using odds ratios as effect
sizes for meta-analysis of dichotomous data: A primer on methods and issues.
Psychological Methods, 3(3), 339–353.
Hamburger, M. E., Lilienfeld, S. O., & Hogben, M. (1996). Psychopathy, gender, and
gender roles: Implications for antisocial and histrionic personality disorders.
Journal of Personality Disorders, 10(1), 41–55.
129
Hare, R. D. (1985). Comparison of procedures for the assessment of psychopathy.
Journal of Consulting and Clinical Psychology, 53(1), 7–16.
Hare, R. D. (2003). Manual for the Revised Psychopathy Checklist (2nd ed.). Toronto,
ON: Multi-Health Systems.
Hare, R. D., Clark, D., Grann, M., & Thornton, D. (2000). Psychopathy and the predictive
validity of the PCL–R: An international perspective. Behavioral Sciences & the
Law, 18(5), 623–645.
Hare, R. D., Harpur, T. J., & Hemphill, J. F. (1989). Scoring pamphlet for the Self–Report
Psychopathy scale: SRP–II. Unpublished manuscript, Simon Fraser University,
Burnaby, BC, Canada.
Hare, R. D., & Neumann, C. S. (2010). The role of antisociality in the psychopathy
construct: Comment on Skeem and Cooke (2010). Psychological
Assessment, 22(2), 446–454.
Harrell, F. E. (2012, August 3). Statistics problems to document and avoid. Retrieved
from http://biostat.mc.vanderbilt.edu/wiki/Main/ManuscriptChecklist
Harrell, F. E., Lee, K. L., Jr., Califf, R. M., Pryor, D. B., & Rosati, R. A. (1984).
Regression modelling strategies for improved prognostic prediction. Statistics in
Medicine, 3(2), 143–152.
Harris, G. T., Rice, M. E., & Quinsey, V. L. (1993). Violent recidivism of mentally
disordered offenders: The development of a statistical prediction instrument.
Criminal Justice and Behavior, 20(4), 315–335.
Hart, S. D. (1998a). Psychopathy and risk for violence. In D. J. Cooke, A. E. Forth, & R.
D. Hare (Eds.), Psychopathy: Theory, Research and Implications for Society (pp.
355–374). Dordrecht, The Netherlands: Kluwer Academic Publishers.
130
Hart, S. D. (1998b). The role of psychopathy in assessing risk for violence: Conceptual
and methodological issues. Legal and Criminological Psychology, 3(1), 121–137.
Hart, S. D., Cox, D. N., & Hare, R. D. (1995). Manual for the Psychopathy Checklist:
Screening Version (PCL:SV). Toronto, ON, Canada: Multi-Health Systems.
Hart, S. D., & Hare, R. D. (1997). Psychopathy: Assessment and association with
criminal conduct. In D. M. Stoff, J. Breiling, & J. Maser (Eds.), Handbook of
Antisocial Behavior (pp. 22–35). Hoboken, NJ: Wiley & Sons Inc.
Hart, S. D., Kropp, P. R., & Hare, R. D. (1988). Performance of male psychopaths
following conditional release from prison. Journal of Consulting and Clinical
Psychology, 56(2), 227–232.
Hartung, C. M., & Widiger, T. A. (1998). Gender differences in the diagnosis of mental
disorders: Conclusions and controversies of the DSM–IV. Psychological Bulletin,
123(3), 260–278.
Hemphill, J. F., Hare, R. D., & Wong, S. (1998). Psychopathy and recidivism: A review.
Legal and Criminological Psychology, 3(1), 139–170.
Hemphill, J. F., Templeman, R., Wong, S., & Hare, R. D. (1998). Psychopathy and
crime: Recidivism and criminal careers. In D. J. Cooke, A. E. Forth, & R. D. Hare
(Eds.), Psychopathy: Theory, research, and implications for society (pp. 375–
399). Boston, MA: Kluwer Academic Publishers.
Hicks, B. M., Carlson, M. D., Blonigen, D. M., Patrick, C. J., Iacono, W. G., & MGue, M.
(2011). Psychopathic personality traits and environmental contexts: Differential
correlates, gender differences, and genetic mediation. Personality Disorders:
Theory, Research, and Treatment. Advance online publication.
doi:10.1037/a0025084
Hunsley, J., & Meyer, G. J. (2003). The incremental validity of psychological testing and
assessment: Conceptual, methodological, and statistical issues. Psychological
Assessment, 15(4), 446–455.
Jackson, R., & Richards, H. (2007). Psychopathy in women: A valid construct with clear
implications. In H. Herve & J. C. Yuille (Eds.), The psychopath: Theory, research,
and practice (pp. 389–410). Mahwah, NJ: Lawrence Erlbaum Associates
Publishers.
131
Jackson, R. L., Rogers, R., Neumann, C. S., & Lambert, P. L. (2002). Psychopathy in
female offenders: An investigation of its underlying dimensions. Criminal Justice
and Behavior, 29(6), 692–704.
Jensen, P. M., Fraser, F., Shankardass, K., Epstein, R., & Khera, J. (2009). Are long-
term care residents referred appropriately to hospital emergency departments?
Canadian Family Physician, 55(5), 500–505.
Jones, S., & Miller, J. D. (2012). Psychopathic traits and externalizing behaviors: A
comparison of self- and informant reports in the statistical prediction of
externalizing behaviors. Psychological Assessment, 24(1), 255–260.
Kamata, A., Turhan, A., & Darandari, E. (2003, April). Estimating reliability for
multidimensional composite scale scores. Paper presented at the annual meeting
of American Educational Research Association, Chicago, IL. Retrieved from:
http://mailer.fsu.edu/~akamata/papers/MD_rel_paper.pdf
Karpman, B. (1948). The myth of the psychopathic personality. The American Journal of
Psychiatry, 104(9), 523–534.
Kennealy, P. J., Skeem, J. L., Walters, G. D., & Camp, J. (2010). Do core interpersonal
and affective traits of PCL–R psychopathy interact with antisocial behavior and
disinhibition to predict violence? Psychological Assessment, 22(3), 569–580.
Kreis, M. K. F, & Cooke, D. J. (2008, July). Capturing the psychopathic female: Piloting
the CAPP–IRS with women offenders. Paper presented at the 8th annual
conference of the International Association of Forensic Mental Health Services
(IAFMHS), Vienna, Austria.
132
Kruh, I. P., Whittemore, K., Arnaut, G. L.Y, Manley, J., Gage, B., & Gagliardi, G. J.
(2005). The concurrent validity of the Psychopathic Personality Inventory and its
relative association with past violence in a sample of insanity acquittees.
International Journal of Forensic Mental Health, 4(2), 135–145.
Leistico, A.-M. R., Salekin, R. T., DeCoster, J., & Rogers, R. (2008). A large-scale meta-
analysis relating the Hare measures of psychopathy to antisocial conduct. Law
and Human Behavior, 32(1), 28–45.
Lester, W. S., Salekin, R. T., & Sellbom, M. (2013). The SRP–II as a rich source of data
on the psychopathic personality. Psychological Assessment 25(1), 32–46.
Lilienfeld, S. O., & Hess, T. H. (2001). Psychopathic personality traits and somatization:
Sex differences and the mediating role of negative emotionality. Journal of
Psychopathology and Behavioral Assessment, 23(1), 11–24.
Lilienfeld, S. O., Patrick, C. J., Benning, S. D., Berg, J., Sellbom, M., & Edens, J. F.
(2012). The role of fearless dominance in psychopathy: Confusions,
controversies, and clarifications. Personality Disorders, 3(3), 327–340.
133
Lilienfeld, S. O., Van Valkenburg, C., Larntz, K., & Akiskal, H. S. (1986). The relationship
of histrionic personality disorder to antisocial personality and somatization
disorders. The American Journal of Psychiatry, 143(6), 718–722. Retrieved from
http://www .ncbi.nlm.nih.gov/
Lilienfeld, S. O., & Widows, M. R. (2005). Professional manual for the Psychopathic
Personality Inventory–Revised (PPI–R). Lutz, FL: Psychological Assessment
Resources.
Loucks, A. D., & Zamble, E. (2000). Predictors of criminal behavior and prison
misconduct in serious female offenders. Empirical and Applied Criminal Justice
Review, 1, 1–47.
Louth, S. M., Hare, R. D., & Linden, W. (1998). Psychopathy and alexithymia in female
offenders. Canadian Journal of Behavioural Science, 30(2), 91–98.
Luke, D. A., & Homan, S. M. (1998). Time and change: Using survival analysis in clinical
assessment and treatment evaluation. Psychological Assessment, 10(4), 360–
378.
Lykken, D. T., Tellegen, A., & Katzenmeyer, C. G. (1973). Manual for the Activity
Preference Questionnaire. Unpublished manuscript, University of Minnesota,
MN.
Lynam, D. R., Whiteside, S., & Jones, S. (1999). Self-reported psychopathy: A validation
study. Journal of Personality Assessment, 73(1), 110–132.
Malterer, M. B., Lilienfeld, S. O., Neumann, C. S., & Newman, J. P. (2010). Concurrent
validity of the Psychopathic Personality Inventory with offender and community
samples. Assessment, 17(1), 3–15.
134
Mahmut, M. K., Menictas, C., Stevenson, R. J., & Homewood, J. (2011). Validating the
factor structure of the Self-Report Psychopathy scale in a community sample.
Psychological Assessment, 23(3), 670–678.
Marcus, D. K., Fulton, J. J., & Edens, J. F. (2012). The two-factor model of psychopathic
personality: Evidence from the Psychopathic Personality Inventory. Personality
Disorders: Theory, Research, and Treatment, 3(2), 140–154.
Marsee, M. A., Silverthorn, P., & Frick, P. J. (2005). The association of psychopathic
traits with aggression and delinquency in non-referred boys and girls. Behavioral
Sciences & the Law, 23(6), 803–817.
McCormick, A. V., Corrado, R. R., Hart, S. D., & Cohen, I. M. (2008, July). Interrater
reliability and internal consistency reliability of the CAPP–IRS among
incarcerated young offenders. Paper presented at the 8th annual conference of
the International Association of Forensic Mental Health Services (IAFMHS),
Vienna, Austria.
Miller, J. D., Jones, S. E., & Lynam, D. R. (2011). Psychopathic traits from the
perspective of self and informant reports: Is there evidence for a lack of insight?
Journal of Abnormal Psychology, 120(3), 758–764.
Miller, J. D., & Lynam, D. R. (2003). Psychopathy and the five-factor model of
personality: A replication and extension. Journal of Personality Assessment,
81(2), 168–178.
Miller, J. D., Lynam, D. R., & Leukefeld, C. (2003). Examining antisocial behavior
through the lens of the Five Factor Model of personality. Aggressive Behavior,
29(6), 497–514.
135
Miller, J. D., Lynam, D. R., Widiger, T. A., & Leukefeld, C. (2001). Personality disorders
as extreme variants of common personality dimensions: Can the five-factor
model adequately represent psychopathy? Journal of Personality, 69(2), 253-
276.
Miller, J. D., Watts, A., & Jones, S. E. (2011). Does psychopathy manifest divergent
relations with components of its nomological network depending on gender?
Personality and Individual Differences, 50(5), 564–569.
Millon, T. (1987). Millon Clinical Multiaxial Inventor–II: Manual for the MCMI–II.
Minneapolis, MN: National Computer Systems.
Morey, L.C. (2001, September). The challenge of discriminant validity in the assessment
of psychopathology. Invited keynote address to the Australia Psychological
Society, Adelaide, Australia.
Motiuk, L., & Porporino, F. (1989). Field test of the Community Risk/Needs Management
Scale: A Study of Offenders on Caseload. Research Report No. R-06. Retrieved
from the Correctional Service of Canada website: http://www.csc-
scc.gc.ca/text/pblct/forum/e093/e093ind-eng.shtml
Murray, G., Rawlings, D., Allen, N. B., & Trinder, J. (2003). NEO Five-Factor Inventory
scores: Psychometric properties in a community sample. Measurement and
Evaluation in Counseling and Development, 36(3), 140–149.
Neal, T. M. S., & Sellbom, M. (2012). Examining the factor structure of the Hare Self-
Report Psychopathy Scale. Journal of Personality Assessment, 94(3), 244–253.
Neumann, C. S., Malterer, M. B., & Newmann, J. P. (2008). Factor structure of the
Psychopathic Personality Inventory (PPI): Findings from a large incarcerated
sample. Psychological Assessment, 20(2), 169–174.
136
Neumann, C. S., Uzieblo, K., Crombez, G., & Hare, R. D. (2013). Understanding the
Psychopathic Personality Inventory (PPI) in terms of the unidimensionality,
orthogonality, and construct validity of PPI-I and PPI-II. Personality Disorders,
4(1), 77–79.
Nicholls, T. L., Odgers, C. L., & Cooke, D. J. (2007). Women and girls with psychopathic
characteristics. In A. R. Felthous and H. Sass (Eds.), International Handbook on
Psychopathic Disorders and the Law (Vol.1, pp.347–366). New York, NY: John
Wiley & Sons, Ltd.
Nicholls, T. L., Ogloff, J. R., Brink, J., & Spidel, A. (2005). Psychopathy in women: A
review of its clinical usefulness for assessing risk for aggression and criminality.
Behavioral Sciences & the Law, 23(6), 779–802.
Nicholls, T. L., Ogloff, J. R., & Douglas, K. S. (2004). Assessing risk for violence among
male and female civil psychiatric patients: The HCR–20, PCL:SV, and VSC.
Behavioral Sciences & the Law, 22(1), 127–158.
Nicholls, T. L., & Petrila, J. (2005). Gender and psychopathy: An overview of important
issues and introduction to the special issue. Behavioral Sciences & the Law,
23(6), 729–741.
Nikolova, N. L., Hendry, M. C., Douglas, K. S., Edens, J. F., & Lilienfeld, S. O. (2012).
The inconsistency of inconsistency scales: A comparison of two widely used
measures. Behavioral Sciences & the Law, 30(1), 16–27.
Novaco, R. W. (1994). Anger as a risk factor for violence among the mentally
disordered. In J. Monahan & H. J. Steadman (Eds.), Violence and mental
disorder (pp. 21–59). Chicago: University of Chicago Press.
Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory (3rd ed.). New York, NY:
McGraw-Hill, Inc.
O’Connor, D. A. (2003). The female psychopath: Validity and factor structure of the
revised Psychopathy Checklist (PCL–R) in women inmates. Dissertation
Abstracts International: Section B. The Sciences and Engineering, 63(12), 6101.
Odgers, C. L., Reppucci, N. D., & Moretti, M. M. (2005). Nipping psychopathy in the bud:
An examination of the convergent, predictive, and theoretical utility of the
PCL:YV among adolescent girls. Behavioral Sciences & the Law, 23(6), 743–
763.
137
Olver, M. E., Neumann, C. S., Wong, S. C. P., & Hare, R. D. (2012). The structural and
predictive properties of the Psychopathy Checklist–Revised in Canadian
Aboriginal and non-Aboriginal offenders. Psychological Assessment. Advance
online publication. doi:10.1037/a0029840
Olver, M. E., Stockdale, K. C., & Wormith, J. S. (2009). Risk assessment with young
offenders: A meta-analysis of three assessment measures. Criminal Justice and
Behavior, 36(4), 329–353.
Paris, J. (2007). An overview on gender, personality and mental health. Personality and
Mental Health, 1(1), 14–20.
Patrick, C. J., Curtin, J. J., & Tellegen, A. (2002). Development and validation of a brief
form of the Multidimensional Personality Questionnaire. Psychological
Assessment, 14(2), 150–163.
Patrick, C. J., Edens, J. F., Poythress, N. G., Lilienfeld, S. O., & Benning, S. D. (2006).
Construct validity of the Psychopathic Personality Inventory two-factor model with
offenders. Psychological Assessment, 18(2), 204–208.
Patrick, C. J., Venables, N. C., & Drislane, L. E. (2013). The role of fearless dominance
in differentiating psychopathy from antisocial personality disorder: Comment on
Marcus, Fulton, and Edens. Personality Disorders, 4(1), 80–82.
Paulhus, D. L., Neumann, C. S., & Hare, R. D. (in press). Manual for the Self-Report
Psychopathy Scale (SRP–III). Toronto, ON, Canada: Multi-Health Systems.
Paulhus, D. L., & Williams, K. M. (2002). The Dark Triad of personality: Narcissism,
Machiavellianism, and psychopathy. Journal of Research in Personality, 36(6),
556–563.
Pedersen, L., Kunz, C., Rasmussen, K., & Elsass, P. (2010). Psychopathy as a risk
factor for violent recidivism: Investigating the Psychopathy Checklist Screening
Version (PCL:SV) and the Comprehensive Assessment of Psychopathic
Personality (CAPP) in a forensic psychiatric setting. International Journal of
Forensic Mental Health, 9(4), 308–315.
138
Peduzzi, P. N., Concato, J., Feinstein, A. R., & Holford, T. R. (1995). Importance of
events per independent variable in proportional hazards regression analysis. II.
Accuracy and precision of regression estimates. Journal of Clinical Epidemiology,
48(12), 1503–1510.
Peduzzi, P. N., Concato, J., Kemper, E., Holford, T. R., & Feinstein, A. R. (1996). A
simulation study of the number of events per variable in logistic regression
analysis. Journal of Clinical Epidemiology, 49(12), 1373–1379.
Perri, F. S., & Lichtenwald, T. G. (2010, Summer). The last frontier: Myths and the
female psychopathic killer. The Forensic Examiner. Retrieved from
http://www.all-about-forensic-psychology.com/support-files/female-psychopathic-
killers.pdf
Poythress, N. G., Edens, J. F., & Lilienfeld, S. O. (1998). Criterion-related validity of the
Psychopathic Personality Inventory in a prison sample. Psychological
Assessment, 10(4), 426–430.
Poythress, N. G., Lilienfeld, S. O., Skeem, J. L., Douglas, K. S., Edens, J. F., Epstein,
M., & Patrick, C. J. (2010). Using the PCL–R to help estimate the validity of two
self-report measures of psychopathy with offenders. Assessment, 17(2), 206–
219.
Pratto, F., Sidanius, J., & Levin, S. (2006). Social dominance theory and the dynamics of
intergroup relations: Taking stock and looking forward. European Review of
Social Psychology, 17(1), 271–320.
Ray, J. V., Hall, J., Rivera-Hudson, N., Poythress, N. G., Lilienfeld, S. O., & Morano, M.
(2012). The relation between self-reported psychopathic traits and distorted
response styles: A meta-analytic review. Personality Disorders: Theory,
Research, and Treatment. Advance online publication. doi:10.1037/a0026482
Ray, J. V., Poythress, N. G., Weir, J. M., & Rickelm, A. (2009). Relationships between
psychopathy and impulsivity in the domain of self-reported personality features.
Personality and Individual Differences, 46(2), 83–87.
Ray, J. V., Weir, J. W., Poythress, N. G., & Rickelm, A. (2011). Correspondence
between the Psychopathic Personality Inventory and the Psychopathic
Personality Inventory-Revised: A look at self-reported personality traits. Criminal
Justice and Behavior, 38(4), 375–385.
139
Robbins, P. C., Monahan, J., & Silver, E. (2003). Mental disorder, violence, and gender.
Law and Human Behavior, 27(6), 561–571.
Richards, H. J., Casey, J. O., & Lucente, S. W. (2003). Psychopathy and treatment
response in incarcerated female substance abusers. Criminal Justice and
Behavior, 30(2), 251–276.
Rosenthal, R. (1991). Meta-analytic procedures for social research (rev. ed.) Newbury
Park, CA: Sage Publications Inc.
Ross, S. R., Benning, S. D., Patrick, C. J., Thompson, A., & Thurston, A. (2009). Factors
of the Psychopathic Personality Inventory: Criterion-related validity and
relationship to the BIS/BAS and five-factor models of
personality. Assessment, 16(1), 71–87.
Rueve, M. E., & Welton, R. S. (2008). Violence and mental illness. Psychiatry
(Edgmont), 5(5), 34–48.
Rutherford, M. J., Cacciola, J. S., Alterman, A. I., & McKay, J. R. (1996). Reliability and
validity of the Revised Psychopathy Checklist in women methadone patients.
Assessment, 3(2), 145–156.
Salekin, R. T., Rogers, R., & Sewell, K. W. (1997). Construct validity of psychopathy in a
female offender sample: A multitrait-multimethod evaluation. Journal of Abnormal
Psychology, 106(4), 576–585.
Salekin, R. T., Rogers, R., Ustad, K. L., & Sewell, K. W. (1998). Psychopathy and
recidivism among female inmates. Law and Human Behavior, 22(1), 109–128.
140
Salekin, R. T., Trobst, K. K., & Krioukova, M. (2001). Construct validity of psychopathy in
a community sample: A nomological net approach. Journal of Personality
Disorders, 15(5), 425–441.
Sandoval, A.-M. R., Hancock, D., Poythress, N., Edens, J. F., & Lilienfeld, S. (2000).
Construct validity of the Psychopathic Personality Inventory in a correctional
sample. Journal of Personality Assessment, 74(2), 262–281.
Schmeelk, K. M., Sylvers, P., & Lilienfeld, S. O. (2008). Trait correlates of relational
aggression in a nonclinical sample: DSM-IV personality disorders and
psychopathy. Journal of Personality Disorders, 22(3), 269–283.
doi:10.1521/pedi.2008.22.3.269
Seibert, L. A., Miller, J. D., Few, L. R., Zeichner, A., & Lynam, D. R. (2011). An
examination of the structure of self-report psychopathy measures and their
relations with general traits and externalizing behaviors. Personality Disorders:
Theory, Research, and Treatment, 2(3), 193–208.
Sevecke, K., Lehmkuhl, G., & Krischer, M. K. (2009). Examining relations between
psychopathology and psychopathy dimensions among adolescent female and
male offenders. European Child & Adolescent Psychiatry, 18(2), 85–95.
doi:10.1007/s00787-008-0707-7
Sevecke, K., Pukrop, R., Kosson, D. S., & Krischer, M. K. (2009). Factor structure of the
Hare Psychopathy Checklist: Youth Version in German female and male
detainees and community adolescents. Psychological Assessment, 21(1), 45–56.
doi:10.1037/a0015032
Sidanius, J., & Pratto, F. (1999). Social dominance: An intergroup theory of social
hierarchy and oppression. New York, NY: Cambridge University Press.
Silverthorn, P., Frick, P. J., & Reynolds, R. (2001). Timing of onset and correlates of
severe conduct problems in adjudicated girls and boys. Journal of
Psychopathology and Behavioral Assessment, 23(3), 171–181.
141
Skeem, J. L., & Cooke, D. J. (2010a). Is criminal behavior a central component of
psychopathy? Conceptual directions for resolving the debate. Psychological
Assessment, 22(2), 433–445.
Skeem, J. L., & Cooke, D. J. (2010b). One measure does not a construct make:
Directions toward reinvigorating psychopathy research—reply to Hare and
Neumann (2010). Psychological Assessment, 22(2), 455–459.
Skeem, J. L., Miller, J. D., Mulvey, E., Tiemann, J., & Monahan, J. (2005). Using a five-
factor lens to explore the relation between personality traits and violence in
psychiatric patients. Journal of Consulting and Clinical Psychology, 73(3), 454–
465.
Skeem, J. L., Mulvey, E. P., & Grisso, T. (2003). Applicability of traditional and revised
models of psychopathy to the Psychopathy Checklist: Screening Version.
Psychological Assessment, 15(1), 41–55.
Skeem, J. L., Polaschek, D. L. L., Patrick, C. J., & Lilienfeld, S. O. (2011). Psychopathic
personality: Bridging the gap between scientific evidence and public policy.
Psychological Science in the Public Interest, 12(3), 95–162.
Smith, G. T., McCarthy, D. M., & Anderson, K. G. (2000). On the sins of short-form
development. Psychological Assessment, 12(1), 102–111.
Smith, S. T., Edens, J. F., & Vaughn, M. G. (2011). Assessing the external correlates of
alternative factor models of the Psychopathic Personality Inventory–Short Form
across three samples. Journal of Personality Assessment, 93(3), 244–256.
Soper, D. S. (2011). A-priori sample size calculator. Free Statistics Calculators (Online
Software). Retrieved from http://www.danielsoper.com/statcalc3/default.aspx
Soper, D. S., Demirkan, H., Goul, M., & St. Louis, R. (2012). An empirical examination of
the impact of ICT investments on future levels of institutionalized democracy and
foreign direct investment in emerging societies. Journal of the Association for
Information Systems, 13(3), 116–149.
Sprague, J., Javdani, S., Sadeh, N., Newman, J. P., & Verona, E. (2012). Borderline
personality disorder as a female phenotypic expression of psychopathy?
Personality Disorders: Theory, Research, and Treatment, 3(2), 127–139.
Strand, S., & Belfrage, H. (2001). Comparison of HCR–20 scores in violent mentally
disordered men and women: Gender differences and similarities. Psychology,
Crime & Law, 7(1), 71–79.
Stafford, E., & Cornell, D. G. (2003). Psychopathy scores predict adolescent inpatient
aggression. Assessment, 10(1), 102–112.
142
Stanford, M. S., & Felthous, A. R. (2011). Introduction to this issue: Violent and
antisocial behavior in women. Behavioral Sciences & the Law, 29(5), 621–622.
Stickle, T. R., Marini, V. A., & Thomas, J. N. (2012). Gender differences in psychopathic
traits, types, and correlates of aggression among adjudicated youth. Journal of
Abnormal Child Psychology, 40(4), 513–525.
Strub, D. S., Kreis, M. K. F., & Hart, S. D. (2008, July). A new approach to the
assessment of psychopathy: Discriminant and convergent validity of the CAPP
across clusters of personality disorder. Poster presented at the 8th annual
conference of the International Association of Forensic Mental Health Services
(IAFMHS), Vienna, Austria.
Taylor, J., & Lang, A. R. (2006). Psychopathy and substance use disorders. In C. J.
Patrick (Ed.) Handbook of psychopathy (pp. 495–511). New York, NY: The
Guilford Press.
Tien, G., Lamb, D., Bond, L., Gillstrom, B., & Paris, F. (1993, May). Report on the needs
assessment of women at the Burnaby Correctional Centre for Women. Burnaby,
BC: BC Institute of Family Violence.
The Sentencing Project (2007, May). Women in the criminal justice system: Briefing
sheets. Retrieved from
http://www.sentencingproject.org/doc/publications/womenincj_total.pdf
Uzieblo, K., Verschuere, B., Van den Bussche, E., & Crombez, G. (2010). The validity of
the Psychopathic Personality Inventory-Revised in a community sample.
Assessment, 17(3), 334–346.
Verona, E., & Carbonell, J. L. (2000). Female violence and personality: Evidence for a
pattern of overcontrolled hostility among one-time violent female offenders.
Criminal Justice and Behavior, 27(2), 176–195.
Verona, E., Patrick, C. J., & Joiner, T. E. (2001). Psychopathy, antisocial personality,
and suicide risk. Journal of Abnormal Psychology, 110(3), 462–470.
143
Vincent, G. M., Odgers, C. L., McCormick, A. V., & Corrado, R. (2008). The PCL:YV and
recidivism in male and female juveniles: A follow-up into young adulthood.
International Journal of Law and Psychiatry, 31(3), 287–296.
Vitale, J. E., Smith, S. S., Brinkley, C. A., & Newman, J. P. (2002). The reliability and
validity of the Psychopathy Checklist–Revised in a sample of female offenders.
Criminal Justice and Behavior, 29(2), 202–231.
Walters, G. D. (2011). Psychopathy and crime: Testing the incremental validity of PCL–
R-measured psychopathy as a predictor of general and violent recidivism. Law
and Human Behavior, Advance online publication. doi:10.1037/h0093928
Walters, G. D., & Duncan, S. A. (2005). Use of the PCL–R and PAI to predict release
outcome in inmates undergoing forensic evaluation. Journal of Forensic
Psychiatry & Psychology, 16(3), 459–476.
Walters, G. D., & Heilbrun, K. (2010). Violence risk assessment and Facet 4 of the
Psychopathy Checklist: Predicting institutional and community aggression in two
forensic samples. Assessment, 17(2), 259–268. doi:10.1177/1073191109356685
Walters, G. D., Knight, R. A., Grann, M., & Dahle, K.-P. (2008). Incremental validity of
the Psychopathy Checklist facet scores: Predicting release outcome in six
samples. Journal of Abnormal Psychology, 117(2), 396–405. doi:10.1037/0021-
843X.117.2.396
Walters, G. D., Wilson, N. J., & Glover, A. J. J. (2011). Predicting recidivism with the
Psychopathy Checklist: Are factor score composites really necessary?
Psychological Assessment, 23(2), 552–557.
Warren, J. I., Burnette, M. L., South, S. C., Chauhan, P., Bale, R., Friend, R., & Van
Patten, I. (2003). Psychopathy in women: Structural modeling and comorbidity.
International Journal of Law and Psychiatry, 26(3), 223–242.
144
Warren, J. I., & South, S. C. (2006). Comparing the constructs of antisocial personality
disorder and psychopathy in a sample of incarcerated women. Behavioral
Sciences & the Law, 24(1), 1–20.
Watkinson, A. V., Corrado, R. R., Lussier, P., Hart, S. D., & Cohen, I. M. (2007, July).
Comprehensive Assessment of Psychopathic Personality Disorder and the
Psychopathy Checklist: Youth Version. Paper presented at the 7th annual
conference of the International Association of Forensic Mental Health Services
(IAFMHS), Montreal, Canada.
Wattal, S., Schuff, D., Mandviwalla, M., & Williams, C. B. (2010). Web 2.0 and politics:
The 2008 U.S. presidential election and an e-politics research agenda. MIS
Quarterly, 34(4), 669–688.
Webster, C. D. (1999, December). Risk assessment and risk management with women
offenders. Report to the National Parole Board, Ottawa, ON.
Webster, C. D., Eaves, D., Douglas, K. S., & Wintrup, A. (1995). The HCR–20 scheme:
The assessment of dangerousness and risk. Burnaby, BC, Canada: Mental
Health, Law, and Policy Institute, and Forensic Psychiatric Services Commission
of British Columbia.
Weiler, B. L., & Widom, C. S. (1996). Psychopathy and violent behaviour in abused and
neglected young adults. Criminal Behaviour and Mental Health, 6(3), 253–271.
Weizmann-Henelius, G., Putkonen, H., Grönroos, M., Lindberg, N., Eronen, M., &
Häkkänen-Nyholm, H. (2010). Examination of psychopathy in female homicide
offenders—Confirmatory factor analysis of the PCL–R. International Journal of
Law and Psychiatry, 33(3), 177–183.
Weizmann-Henelius, G., Viemerö, V., & Eronen, M. (2004). Psychological risk markers
in violent female behavior. International Journal of Forensic Mental Health, 3(2),
185–196.
Williams, K. M., & Paulhus, D. L. (2004). Factor structure of the Self-Report Psychopathy
scale (SRP–II) in non-forensic samples. Personality and Individual Differences,
37(4), 765–778.
Williams, K. M., Paulhus, D. L., & Hare, R. D. (2007). Capturing the four-factor structure
of psychopathy in college students via self-report. Journal of Personality
Assessment, 88(2), 205–219.
Wilson, D. L., Frick, P. J., & Clements, C. B. (1999). Gender, somatization, and
psychopathic traits in a college sample. Journal of Psychopathology and
Behavioral Assessment, 21(3), 221–235.
145
Witt, E. A., Donnellan, M. B., Blonigen, D. M., Krueger, R. F., & Conger, R. D. (2009).
Assessment of fearless dominance and impulsive antisociality via normal
personality measures: Convergent validity, criterion validity, and developmental
change. Journal of Personality Assessment, 91(3), 265–276.
Wormith, J. S., Olver, M. E., Stevenson, H. E., & Girard, L. (2007). The long-term
prediction of offender recidivism using diagnostic, personality, and risk/need
approaches to offender assessment. Psychological Services, 4(4), 287–305.
Wynn, R., Høiseth, M. H., & Pettersen, G. (2012). Psychopathy in women: Theoretical
and clinical perspectives. International Journal of Women’s Health, 4, 257–263.
Yang, M., Wong, S. C. P., & Coid, J. (2010). The efficacy of violence prediction: A meta-
analytic comparison of nine risk assessment tools. Psychological Bulletin, 136(5),
740–767.
146