You are on page 1of 15

1015312

research-article2021
ASMXXX10.1177/10731911211015312AssessmentMcCallum et al.

Article
Assessment

Psychopathy Profiles and Personality


1­–15
© The Author(s) 2021
Article reuse guidelines:
Assessment Inventory Scores in a Sex sagepub.com/journals-permissions
DOI: 10.1177/10731911211015312
https://doi.org/10.1177/10731911211015312

Offender Risk Assessment Field Setting journals.sagepub.com/home/asm

Katherine E. McCallum1 , Marcus T. Boccaccini2 ,


Jorge G. Varela2, and Darrel B. Turner3

Abstract
A growing body of research suggests there are identifiable psychopathy subtypes among offenders scored on Hare’s
Psychopathy Checklist–Revised (PCL-R; Hare, 2003). We used latent profile analysis to examine the generalizability
of these subtype findings to PCL-R scores (N = 615) assigned in a sex offender risk assessment field setting and to
examine how offender subtypes differ on measures of comorbid psychopathology, risk, and treatment amenability from
the Personality Assessment Inventory. Consistent with prior research, we identified four subtypes when using PCL-R
scores from all offenders: Prototypic psychopathy (n = 239, 38.9%), callous-conning (n = 154, 25.0%), sociopathic (n =
96, 15.6%), and general offenders (n = 126, 20.5%). Prototypic and sociopathic subtypes exhibited the highest levels of
comorbid psychopathology and risk for potential violence. We identified classes consistent with primary (n = 66, 36.7%)
and secondary (n = 114, 63.3%) psychopathy among offenders with PCL-R total scores ≥ 25, and found higher levels of
comorbid psychopathology and potential for violence among those in the secondary psychopathy class. Findings provide
support the generalizability of existing PCL-R subtype findings to field scores and show how those with similar PCL-R total
scores may differ on scores from commonly used multiscale inventories.

Keywords
psychopathy, PCL-R, PAI, field validity, sex offender

Hare’s PCL-R (Hare, 1991, 2003) is a commonly used mea- the practical implications from subtype study findings have
sure for assessing psychopathic traits in forensic evalua- been key issues in recent disagreements about the use of the
tions (Neal & Grisso, 2014; Viljoen, et al., 2010), including PCL-R for predicting institutional violence in real-world
sex offender risk assessments (Boccaccini et al., 2017). The cases, especially those involving the death penalty (see
clinician-scored PCL-R includes 20 items that assess affec- DeMatteo et al., 2020; Olver et al., 2020).
tive, interpersonal, lifestyle, and behavioral traits associated Because many subtyping studies have relied on scores
with psychopathy. Due to the number and variety of traits assigned for research purposes, they often provide limited
measured by the PCL-R, there can be notable differences in information about how the subtype groups may differ on
trait profiles among individuals who receive similar total scores from the types of multiscale personality and psycho-
scores. Researchers have used large data sets of PCL-R pathology measures commonly used in routine clinical/
scores to identify replicable psychopathy subtypes (see forensic practice to inform diagnosis and treatment plan-
Klein Haneveld et al., 2018; Lehmann et al., 2019; Neumann ning. In other words, there is limited information available
et al., 2016). to practitioners who may want to apply PCL-R subtyping
Although these PCL-R subtype findings provide impor- studies to clinical practice, to help them better understand
tant information for psychopathology research, their gener- how evaluees with different PCL-R profiles may differ with
alizability to and implications for routine forensic practice
are less clear. There is growing recognition that the psycho- 1
Private Practice, Denver, CO, USA
metric properties of instrument scores assigned for forensic 2
Sam Houston State University, Huntsville, TX, USA
evaluations can be weaker than the psychometric properties 3
Private Practice, Lake Charles, LA, USA
of scores assigned for research purposes (see Edens &
Corresponding Author:
Boccaccini, 2017), and replication of subtype findings with Marcus T. Boccaccini, Department of Psychology and Philosophy, Sam
field scores assigned for real-world decision making is Houston State University, Box 2447, Huntsville, TX 77341, USA.
needed. Indeed, PCL-R field reliability, field validity, and Email: Boccaccini@shsu.edu
2 Assessment 00(0)

respect to comorbid psychopathology and treatment needs. researchers as: (a) prototypic, (b) callous-conning, (c)
In the present study, we use latent profile analysis (LPA) sociopathic, and (d) nonpsychopathic or general offenders
with PCL-R field scores to examine the field validity of (see Hare et al., 2018; Klein Haneveld et al., 2018; Neumann
existing PCL-R subtype findings and then compare psy- et al., 2016).
chopathy subgroups on scores from the PAI (Morey, 1991), The prototypic subtype is characterized by high average
a commonly used multiscale inventory for assessing a wide scores across all four PCL-R facets: Interpersonal (Facet 1),
range of diagnostic and treatment characteristics relevant Affective (Facet 2), Lifestyle (Facet 3), and Antisocial
for clinical decision-making. Behavior (Facet 4). The callous-conning subtype is charac-
terized by elevated interpersonal and affective traits with
comparatively lower lifestyle and antisocial traits. The
Empirical Support for Psychopathy Subtypes
sociopathic subtype is characterized by elevated lifestyle
Several lines of research have investigated whether there is and antisocial traits with comparatively lower interpersonal
empirical support for specific combinations of psycho- and affective traits. Finally, the nonpsychopathic subtype
pathic traits clustering together among identifiable sub- includes offenders who exhibit average or lower psycho-
groups of patients or offenders. Subtype studies focusing on pathic traits across all four facets.
offenders with high psychopathy scores (e.g., PCL-R total Recent psychopathy subtype studies using self-report
score ≥ 25) have typically found support for two subtypes, psychopathy measures have identified somewhat different
often described as primary and secondary psychopathy subtype groups. For example, a recent study of psycho-
(Hicks et al., 2010; Olver et al., 2015; Skeem et al., 2007). pathic traits in the general population identified five sub-
In these studies, primary psychopathy is characterized by types, which the authors termed psychopathic personality,
elevated interpersonal and affective traits with compara- impulsive-irresponsible, callous-unemotional, grandiose-
tively lower lifestyle and antisocial traits, whereas second- manipulative, and nonpsychopathic (Colins et al., 2017).
ary psychopathy is characterized by elevated lifestyle and Among juvenile offenders, LPA analyses with scores from
antisocial traits, with comparatively lower interpersonal self-report psychopathy and anxiety measures revealed four
and affective traits (Klein Haneveld et al., 2018; Neumann classes: control, high anxiety, moderate psychopathy, and
et al., 2016). Similar findings have emerged from subtyping high psychopathy (Colins et al., 2018).
studies using other psychopathy measures (see Hicks &
Drislane, 2018; Lee & Salekin, 2010).
Others have concluded that there may be three subtypes
External Correlates of Psychopathy Subtypes
among high scorers. In a large (N = 1,451) sample of Implications for practice from psychopathy subtype studies
offenders with PCL-R total scores of 27 and higher, come from follow-up validity analyses comparing offend-
researchers identified three subtypes, which they labeled ers from different psychopathy subtype groups on external
manipulative, aggressive, and sociopathic (Mokros et al., correlates relevant to clinical practice. In the primary versus
2015). The manipulative subtype exhibited higher interper- secondary psychopathy literature, researchers have found
sonal traits and lower antisocial traits. The aggressive sub- that primary psychopathy is characterized by lower stress
type exhibited the highest antisocial traits, higher levels of reaction, low harm avoidance, higher social dominance,
aggressiveness, and a larger number of self-reported crimi- relatively stable emotional functioning and a lower rate of
nal tendencies. The sociopathic subtype exhibited very low internalizing disorders, whereas secondary psychopathy is
affective traits but high levels of social deviance. The characterized by higher levels of emotional instability, neu-
authors proposed that the manipulative and aggressive sub- roticism, anxiety, substance use disorders, and a more fre-
types reflected subtypes of the traditional construct of psy- quent rate of comorbid psychopathology and co-occurring
chopathy, while the sociopathic subtype was more consistent disorders (Hicks & Drislane, 2018; Skeem et al., 2007;
with antisocial personality disorder. Vassileva et al., 2005).
Focusing on only those with high psychopathy scores The most consistent finding from studies using all
inevitably omits many individuals from the analyses, and PCL-R scores is that prototypic offenders differ from those
some researchers have argued for the inclusion of the entire in the other subtype groups. Prototypic offenders receive
sample in psychopathy subtype research (Krstic et al., 2018; higher scores on measures specifically designed to assess
Neumann et al., 2016). Using the entire sample provides risk for violence, have a history of committing more violent
researchers the opportunity to examine the possibility of offenses, are more likely to be diagnosed with a personality
subtypes among those with high or low levels of psycho- disorder, and drop out of treatment at a significantly higher
pathic traits, arguably offering a richer and more complete rate than other offenders (Klein Haneveld et al., 2018;
discussion of the psychopathy construct. Recent PCL-R Krstic et al., 2018; Lehmann et al., 2019). Fewer consistent
studies using scores from all available participants have differences between callous-conning, sociopathic, and gen-
found support for four subgroups, described by the eral offender subtypes have emerged from these studies.
McCallum et al. 3

Existing studies suggest that those in the sociopathic group potential indices. Similarly, when the analysis was limited
tend to show the second highest levels of risk and violence to only those with high PCL-R scores (> 25) we hypothe-
and those in the callous-conning group tend to have the sec- sized those in the secondary psychopathy subgroup would
ond highest rate of personality disorder diagnoses (Klein have higher scores on these PAI measures of comorbid psy-
Haneveld et al., 2018; Krstic et al., 2018; Lehmann et al., chopathology than those in the primary psychopathy
2019). Risk measure scores tend to be lowest for those in subgroup.
the callous-conning and general offender groups (Lehmann
et al., 2019). Thus, in terms of risk assessment, prototypic
Method
and sociopathic offenders tend receive the highest scores,
whereas callous-conning and general offenders tend to We report how we determined our sample size, all data
receive the lowest scores. exclusions, and all measures in the study. Approval for this
study, based on data collected from deidentified offender
records, was obtained by the institutional review board at
Current Study Sam Houston State University.
The current study used a LPA approach to examine the gen-
eralizability of existing PCL-R subtype findings to scores
Offenders and SVP Evaluators
from a sex-offender risk-assessment field setting and pro-
vides a comparison of LPA-identified psychopathy sub- Data for this study come from 615 sexually violent predator
groups on scores across the PAI. Our primary goal was to (SVP) evaluations conducted by state-contracted evaluators
provide findings that would be useful for practitioners, in Texas between 1999 and 2018. The mean age among the
who may be interested in the potential implications of offenders at the time of evaluations was 45.47 (SD = 11.65).
PCL-R subtype findings for clinical practice. Our findings Offenders (100% male) were identified in the records pro-
add to the existing PCL-R subtyping literature by provid- vided to evaluators by the state as White (n = 317, 51.5%),
ing information about how psychopathy subgroups differ Black (n = 166, 27.0%), Latino (n = 114, 18.5%), or from
across the broad range of internalizing, externalizing, vio- another race/ethnic background (n = 7, 1.1%; missing n =
lence potential, and treatment amenability measures pro- 11, 1.8%). Evaluators identified offenders as having at least
vided by the widely available PAI. Existing PCL-R LPA one child victim (n = 402, 65.4%) or no child victims (n =
studies typically provide information about a more limited 184, 29.9%; missing n = 29, 4.7%).
range of external correlates. For example, data about The research team collected deidentified data directly
comorbid psychopathology comes from a forensic hospital from the SVP evaluators for a series of ongoing projects
sample with information about only psychosis, autism, and examining evaluator differences in PCL-R scoring (see
a collapsed category of personality disorder diagnoses Boccaccini et al., 2008; Boccaccini et al., 2014). Evaluators
(Klein Haneveld et al., 2018). entered PCL-R item scores, PAI scale scores, and basic
The availability of PAI scores allows us to provide infor- demographic and victim information about the offenders
mation about psychopathy subgroup differences on external into an Excel worksheet. We contacted all 21 state-con-
correlates using scales and scores familiar to many practi- tracted evaluators known to have conducted contracted SVP
tioners. Although not designed specifically for forensic use, evaluations (at the onset of the study) and 12 provided data.
the PAI is often used to assist with diagnosis and treatment The mean number of evaluations per evaluator was 51.25
planning in forensic contexts (Douglas et al., 2001; Ruiz & (SD = 36.17, range = 9 to 139).
Ochshorn, 2010) and scores from several scales are predic- Although we do not have detailed information related to
tive of institutional misconduct, treatment program viola- the type or severity of prior offenses, each offender must
tions, and postrelease arrests (see Boccaccini et al., 2013; have had a history of two or more qualifying sexual offenses
Gardner et al., 2015; Percosky et al., 2013). to be considered for civil commitment. Sexual offenses that
We focused the PAI analyses on composite scores and may qualify a person as an SVP under Texas statute are the
indices designed to measure internalizing psychopathology, following: Indecency With a Child (sexual contact), Sexual
externalizing psychopathology, violence potential, suicide Assault, Aggravated Sexual Assault, Aggravated
potential, and poor treatment amenability. Based on prior Kidnapping (with intent to sexually abuse or violate),
research, we hypothesized that prototypic and sociopathic Burglary (with intent to commit a sexual offense mentioned
offenders would obtain higher scores across the externaliz- above), a murder that is determined beyond a reasonable
ing, violence potential, and poor treatment amenability doubt to have been based on sexually motivated conduct, or
indices than callous-conning and general offenders. We an attempt, conspiracy, or solicitation to commit any offense
hypothesized that callous-conning offenders and prototypic mentioned above. Qualifying offenses can also include
offenders would display a lower rate of comorbid pathology those under prior state law with elements substantially like
and therefore lower scores on the internalizing and suicide those mentioned above, and offenses under other state law,
4 Assessment 00(0)

federal law, or the Uniform Code of Military Justice with analyses focused on index and composite scores that
elements substantially like those mentioned above (Texas describe performance across the PAI. Composite scores
Health and Safety Code, Title 11, § 841, 2010). include the Mean Clinical Elevation, which is the mean
score across the 11 clinical scales, as well as Internalizing
(INT) and Externalizing (EXT) composite scores developed
Measures
using correctional samples (Ruiz & Edens, 2008). INT is
Psychopathy Checklist–Revised. The Hare (1991, 2003) the mean score across six scales (Anxiety, Depression,
PCL-R is a 20-item, clinician-scored measure of interper- Somatic Complaints, Schizophrenia, Anxiety-Related Dis-
sonal, emotional, and behavioral traits scored on the basis orders, Suicidal Ideation), and EXT is the mean score across
offender records and (if possible) a semistructured inter- seven scales (Antisocial Features, Borderline Features
view. Item scores range from 0 to 2, with higher scores (BOR), Alcohol Problems, Drug Problems, Aggression,
reflecting a higher level of the psychopathic trait. The 20 Mania, Paranoia). We also report effects for Violence Poten-
item scores are summed to yield a total score that reflects tial Index (VPI), Treatment Process Index (TPI), and Sui-
the overall degree of psychopathic characteristics. Eighteen cide Potential Index (SPI). Each of these index scores is
items are used to calculate facet scores that identify indi- based on the presence of theoretically or empirically derived
vidual components of psychopathy. The Interpersonal profile characteristics associated with future violence (VPI),
(Facet 1) and Affective (Facet 2) facet scores are each based a difficult treatment process (TPI), or suicidality (SPI; see
on four items, whereas the Impulsive Lifestyle (Facet 3) Morey, 2007). VPI and SUI scores are based on 20 profile
and Antisocial Behavior (Facet 4) facet scores are each characteristics, whereas TPI scores are based on 12 profile
based on five items. The PCL-R can be scored based on a characteristics.
combination of information obtained from file-review and Finally, we report effects for the Antisocial Features
interview, or file-review alone. Although we did not ask (ANT) and BOR clinical scales due to their associations
evaluators to identify cases in which they scored the PCL-R with institutional misconduct and treatment compliance in
based on file-review alone, our experience with Texas SVP forensic settings (see Gardner et al., 2015). ANT is a mea-
cases is that evaluators rarely score the PCL-R without sure of personality and behavioral traits associated with
interviewing the offender. antisocial personality and psychopathy, whereas BOR is a
We had access to PCL-R item scores for all 615 offend- broad measure of severe personality disorder symptoms,
ers. If item scores were missing, we used prorating proce- including borderline personality disorder (see Morey,
dures described in the manual to assign values to missing 2007). In forensic settings, ANT scores are more strongly
items (i.e., prorated facet score minus the sum of the avail- correlated with PCL-R Factor 2 scores than Factor 1 scores
able item scores = prorated item score). The mean PCL-R and are predictive of institutional misconduct and reoffend-
total score in the sample was 19.99 (SD = 7.33): Facet 1 (M ing after treatment (see Gardner et al., 2015; Ruiz et al.,
= 3.49, SD = 2.36), Facet 2 (M = 4.98, SD = 2.18), Facet 2014). BOR scores are associated with both treatment pro-
3 (M = 4.55, SD = 2.36), Fact 4 (M = 4.77, SD = 2.59). gram violations and noncompliance in forensic settings
The reliability of PCL-R scores is strong within research (Boccaccini et al., 2013; Percosky et al., 2013).
contexts, with the PCL-R manual (Hare, 2003) reporting In Texas SVP cases, the Texas Department of Criminal
interrater reliability values (ICC1) from .86 for male inmates Justice administers the PAI to offenders prior to their SVP
to .88 for male forensic psychiatric patients. Research evaluations and includes computerized reports of the PAI
within the SVP evaluation sample used for the current study results in the records they provide to SVP evaluators. PAI
has revealed much lower levels of rater agreement (e.g., scale scores were available for 484 of the 615 offenders
ICCA,1 ≈ .40 to .50 for PCL-R total scores), with some eval- (78.7%) in this study, but we used scores from only 475
uators assigning scores that are, on average, more than 15 offenders due to concerns about the possibility of random
points higher than other evaluators (see Boccaccini et al., responding (Inconsistency or Infrequency ≥ 80T) for nine
2012; Boccaccini et al., 2014). Researchers have reported offenders. One evaluator did not provide clinical subscale
similarly low levels of rater agreement for PCL-R scores scores for 47 offenders. Because these scores are needed to
assigned in other SVP field settings (e.g., Florida), with calculate PAI index scores, the sample size for SPI, TPI, and
facet level ICCA,1 values of .28 (Facet 1), .39 (Facet 2), .56 VPI analyses was 428. Although we do not know why PAI
(Facet 3), and .75 (Facet 4; Miller et al., 2012). scores were unavailable for any individual offender, prior
research with PAI scores in Texas SVP evaluation cases
Personality Assessment Inventory.  The PAI is a 344-item self- indicated that some offenders refuse the PAI evaluation,
report inventory of personality and psychopathology. The others are considered to be too severely ill to complete the
PAI includes 11 clinical scales, 4 validity scales, 5 treatment measure, and scores are sometimes not provided to evalua-
consideration scales, and 2 interpersonal scales. Although tors when the computerized PAI report suggests that the
we provide data for all 22 scales in the appendices, our scores are invalid (see Boccaccini et al., 2010).
McCallum et al. 5

Table 1.  Latent Class Enumeration for Psychopathy Checklist Scores.

Number of Bootstrap, N(%) of


Sample/model Log likelihood parameters BIC Entropy L-M-R, LRT (p) LRT (p) smallest class
Full sample (N = 615)
 One-class −5566.72 8 11184.80  
 Two-class −5340.64 13 10764.76 .76 438.50 (<.01) 452.16 (<.01) 247 (40)
 Three-class −5306.07 18 10727.73 .67 67.05 (.03) 69.14 (<.01) 141 (23)
 Four-class −5273.84 23 10695.38 .68 62.51 (.13) 64.45 (<.01) 96 (16)
 Five-class −5256.32 28 10692.44 .70 34.03 (.70) 35.09 (<.01) 79 (13)
PCL-R total score ≥ 25 (N = 180)
 One-class −1307.48 8 2656.49  
 Two-class −1286.67 13 2640.84 .72 40.08 (<.01) 41.62 (<.01) 66 (37)
 Three-class −1276.05 18 2645.57 .75 20.44 (.09) 21.23 (<.01) 24 (13)

Note. BIC = Bayesian Information Criterion; L-M-R LRT = Lo–Mendell–Ruben adjusted Likelihood Ratio Test; PCL-R = Psychopathy Checklist–
Revised. Bold = Model selected for interpretation.

Latent Profile Analysis classes, and the Lo–Mendell–Ruben adjusted Likelihood


Ratio Test (L-M-R LRT). The best model fit can be deter-
We used LPA with PCL-R facet scores to identify offender mined by information criteria (e.g., the BIC) as well as
subtypes, which has been a common analytic approach in LRTs. Overall, a model with lower BIC, higher entropy, and
recent psychopathy subtype research (Colins et al., 2018; significant L-M-R LRT and bootstrap LRT tests indicates
Mokros et al., 2015; Neumann et al., 2016). We conducted better fit. When fit indices are incompatible, many suggest
one set of analyses using all offenders to examine the gen- relying on information criteria as the primary means for
eralizability of the four-class solution identified in previous deciding model fit. According to Nylund-Gibson and Masyn
research (see Neumann et al., 2016), and a separate set of (2016), the BIC is the best indicator of the correct number
analyses using only those offenders with PCL-R total scores of latent classes, when compared with other fit indices. All
≥25 to examine the generalizability of primary and second- models described were estimated using Mplus, Version 7
ary subtype findings among high scorers. PCL-R subtyping (Muthén & Muthén, 1998-2008), which uses the
studies have used somewhat different PCL-R cut scores as Expectation-Maximization (EM) algorithm under the
inclusion criteria for analyses of high-scoring offenders, assumption that data are missing at random (MAR) to com-
including 25 (Olver et al., 2015), 27 (Mokros et al., 2015), pute maximum likelihood estimates of the model parame-
29 (Skeem et al., 2007), and 30 (Blackburn et al., 2008). We ters (Muthén & Shedden, 1999).
used a cut score of ≥ 25 because it resulted in a subsample
that included 29% of the full sample (180 of 615), which is
generally consistent with previous studies (e.g., 27% to Results
33%; Mokros et al., 2015; Skeem et al., 2007).
LPA is a statistical method of sorting individuals into
Full Sample
similar groups (latent classes) based on a set of observed Latent Profile Analysis Results.  Table 1 provides fit indices for
(manifest) continuous variables as measures of an underly- the LPA models using data from all offenders (N = 615).
ing (latent) categorical variable. In an LPA, class probabili- We considered both these fit values and model selection
ties (the probability that a specific individual belongs to the procedures from existing PCL-R LPA studies to identify
emerging latent class) are measured for each individual case the best model for interpretation in this study. The four-
in the data set. For comparative and comprehensive pur- class solution had a significantly lower BIC value when
poses, analyses were structured to begin with a one-class compared with the three-, two-, and one-class solutions.
model, and add additional classes to subsequent models, Although the five-class solution had the lowest BIC value
similar in structure to the procedure followed by Miller in terms of absolute value, the difference in BIC values
et al. (2009). To determine the number of classes in these between the four-class and five-class models was small (see
data, multiple models (one-class, two-class, three-class . . . ) Kass & Raftery, 1995) and the five-class solution did not
were compared using the fit indices of these models such as provide a substantially better model fit. The average prob-
the Bayesian Information Criterion (BIC), entropy of the abilities for the most likely latent class membership were
6 Assessment 00(0)

Figure 1.  Four LPA subtype profiles by PCL-R mean item facet score.
Note. Class 1 = prototypic psychopaths (n = 239), Class 2 = callous-conning offenders (n = 154); Class 3 = sociopathic offenders (n = 96), Class 4
= nonpsychopathic general offenders (n = 126). LPA = latent profile analysis; PCL-R = Psychopathy Checklist–Revised.

also similar for the four-class (range = .73-.87) and five- for identifying small (d = 0.20), medium (d = 0.50) and
class (range = .75-.85) solutions. large (d = 0.80) effects.
Multiple recent PCL-R LPA studies have reported simi- Figure 1 provides PCL-R facet score profiles for the
lar BIC values for four-class and five-class models, but four classes of offenders as identified by the full-sample
based interpretation on only the four-class model for con- LPA model. This facet-score profile is consistent with
ceptual reasons (Klein Haneveld et al., 2018; Krstic et al., previous studies using an LPA approach (Klein Haneveld
2018; Lehmann et al., 2019). We moved forward with the et al., 2018; Krstic et al., 2018; Neumann et al., 2016).
four-class solution because of both model fit values and the Class 1 offenders (n = 239, 38.9%) exhibited an average
opportunity to compare the four-class model with those PCL-R total score of 27.22 (SD = 3.59) and the highest
from other recent studies, but provide PCL-R descriptive elevations across all facets (see Table 2), similar to the
statistics for the five-class solution in Appendix A. prototypic class discussed in prior studies. Similarly,
Class 4 offenders (n = 126, 20.5%) exhibited low eleva-
PCL-R Facet Scores for Offender Subtypes.  We assigned each tions across all four facets and had an average PCL-R
offender to one class based on most-likely class membership total score of only 9.93 (SD = 3.25), similar to the gen-
probabilities from the LPA. We calculated descriptive statis- eral offenders subtype from prior research.
tics for the members of these classes by averaging their Offenders in Class 2 (n = 154, 25.0%) and Class 3 (n =
PCL-R item scores for the items included on the facet. These 96, 15.6%) had nearly identical mean PCL-R total scores
mean item scores are more useful for making comparisons in (M = 18.12 and 18.22, respectively), but notable differ-
this study than summed facet scores because two of the fac- ences in their facet score profiles (see Figure 1 and Table 2).
ets have fewer items (Facets 1 and 2) than the others (Facets Similar to the callous-conning class from prior research,
3 and 4). We do, however, provide descriptive statistics for Class 2 offenders had significantly higher Facet 1 (Cohen’s
summed facet scores in Appendix B to allow for compari- d = 0.98) and Facet 2 mean item scores (d = 2.09) than
sons with the summed facet scores evaluators use in clinical Class 3 offenders. Similar to the sociopathic class from
practice. We compare facet scores for offenders in the same prior research, Class 3 offenders had significantly higher
class using independent-samples t tests and Cohen’s d as a Facet 3 (d = 0.58) and Facet 4 (d = 2.25) mean item scores
measure of effect size. We follow Cohen’s (1992) guidelines than Class 2 offenders.
Table 2.  Psychopathy Checklist–Revised Facet Scores and Personality Assessment Inventory Composite and Index Scores for Offenders Each Psychopathy Class.

Class 1: Class 2: Callous- Class 3: Class 4:


Prototypic Conning Sociopathic General C1 vs. C2 C1 vs. C3 C1 vs. C4 C2 vs. C3 C2 vs. C4 C3 vs. C4

PCL-R/PAI Score M (SD) M (SD) M (SD) M (SD) d d d d d d


PCL-R: Mean item score n = 239 n = 154 n = 96 n = 126  
  Facet 1: Interpersonal 1.26 (0.50) 0.94 (0.48) 0.50 (0.39) 0.35 (0.41) 0.65 1.61 1.93 0.98 1.31 0.37
  Facet 2: Affective 1.64 (0.28) 1.47 (0.29) 0.83 (0.33) 0.54 (0.35) 0.60 2.74 3.60 2.09 2.92 0.85
  Facet 3: Lifestyle 1.31 (0.30) 0.74 (0.32) 0.93 (0.34) 0.38 (0.28) 1.85 1.22 3.17 −0.58 1.19 1.79
  Facet 4: Antisocial 1.31 (0.39) 0.53 (0.32) 1.24 (0.31) 0.55 (0.38) 2.14 0.19 1.97 −2.25 -0.06 1.96
PCL-R Total (mean item) 1.36 (0.18) 0.91 (0.17) 0.91 (0.14) 0.50 (0.16) 2.55 2.65 4.96 <0.01 2.48 2.70
PCL-R Total (sum) 27.22 (3.59) 18.11 (3.35) 18.22 (2.82) 9.93 (3.27) — — — — — —
PAI Composite/Index n = 175 n = 115 n = 74 n = 111  
  Mean Clinical Elevation 55.50 (7.36) 51.43 (7.08) 53.58 (7.71) 50.34 (7.04) 0.56 0.26 0.71 −0.29 0.15 0.44
 Internalizing 53.63 (9.11) 50.36 (8.66) 51.16 (9.26) 48.55 (7.36) 0.59 0.36 0.60 −0.09 0.22 0.32
 Externalizing 57.06 (7.67) 52.33 (7.04) 55.60 (7.53) 51.83 (7.88) 0.64 0.19 0.67 −0.45 0.07 0.49
  Violence Potential Index 59.22 (13.92) 51.48 (11.31) 57.44 (13.43) 51.84 (11.20) 0.60 0.13 0.57 −0.43 −0.01 0.46
  Treatment Process Index 59.95 (14.89) 52.50 (10.84) 57.94 (11.63) 54.26 (10.84) 0.56 0.14 0.42 −0.49 −0.16 0.33
  Suicide Potential Index 57.23 (13.13) 51.11 (11.00) 54.00 (12.73) 49.73 (10.56) 0.50 0.25 0.61 −0.25 0.13 0.37
PAI Scale
  Antisocial Features 59.53 (8.86) 54.37 (7.51) 58.47 (7.92) 54.24 (8.55) 0.62 0.12 0.61 −0.53 0.01 0.51
  Borderline Features 56.42 (10.81) 51.13 (9.22) 53.85 (9.77) 50.16 (9.75) 0.52 0.25 0.60 −0.29 0.10 0.38

Note. Cohen’s d values in bold are statistically significant at p < .05. d values in bold and italics are statistically significant at p ≤ .01. n = 160, 107, 66, 95 for the Violence Potential Index, Treatment
Process Index, and Suicide Potential Index. PCL-R = Psychopathy Checklist-Revised; PAI = Personality Assessment Inventory.

7
8 Assessment 00(0)

Given the generally large differences between the classes offenders (i.e., 179 of 180) had been classified in the proto-
across the PCL-R, it is notable that the only instances of typic subgroup in the full-sample analyses (one offender had
nonsignificant differences between the classes at the PCL-R been classified by the LPA into the callous-conning sub-
facet level were for Facet 4 (Antisocial) scores (Table 2). group). We used LPA analysis with scores from these 180
Mean item scores for Facet 4 were similarly high for proto- offenders to examine whether there was evidence for pri-
typic (M = 1.31) and sociopathic (M = 1.24) offenders (d mary and secondary psychopathy subtypes among high
= 0.19), and similarly low for callous-conning (M = 0.53) scorers. BIC values from the LPA analysis indicated that the
and general offenders (M = 0.55; d = −0.06). two-class model best fit the data (Table 1).

PAI Scores for Offender Subtypes.  Table 2 provides PAI com- PCL-R Facet Scores for Offender Subtypes.  The PCL-R facet
posite and index scores for offenders in the four psychopa- score profiles of the two emerging latent profiles are
thy subtype groups (Appendix C provides descriptive depicted in Figure 2. This facet score pattern is generally
statistics for individual PAI scales). There were a number of consistent with previous research regarding primary and
moderate-sized and statistically significant differences secondary psychopathy subtypes, but is also consistent with
between the subtype groups, especially between the proto- the manipulative (instead of primary) and aggressive
typic and general offender groups (d = 0.42-0.71). Proto- (instead of secondary) subtypes identified in some studies
typic offenders reported more distress and impairment than (Mokros et al., 2015). Those in the primary class (n = 66,
general offenders across the PAI, including both external- 36.7%) had higher interpersonal (d = 0.84) and affective (d
izing psychopathology (d = 0.67) and—contrary to our = 0.48) facet scores than those in the secondary class (n =
expectations—internalizing (d = 0.60) psychopathology. 114, 63.3%). Those in the secondary class had higher life-
They also reported higher levels of antisocial (d = 0.62) style facet (d = 0.41) scores and much higher antisocial
and borderline (d = 0.52) traits. facet (d = 3.00) scores than those in the primary class.
Prototypic and sociopathic groups obtained the highest This two-class solution suggests a more nuanced picture
index and composite scores across the PAI (see Table 2). of our sample, beyond the LPA findings using the entire
Both groups also had the highest ANT and BOR scores. In sample. Our full sample analysis evidenced one unitary
terms of absolute value, the composite and index scores class of prototypic offenders with an average PCL-R score
from prototypic offenders were always larger than those of 27.22. However, when considering only offenders with a
from sociopathic offenders, but the differences between PCL-R score at or above 25, two classes emerged, with
these groups were small (d = 0.13 to 0.36). The largest dif- those in the primary class having lower (d = 0.78) PCL-R
ferences between these groups was for internalizing psy- total scores (M = 27.26, SD = 1.90) than those in the sec-
chopathology, with prototypic offenders reporting slightly ondary class (M = 29.93, SD = 3.11).
higher levels of internalizing psychopathology than socio-
pathic offenders (d = 0.36, p = .05). Overall, prototypic PAI Composite and Index Scores for Offender Subtypes. 
and sociopathic offenders received generally similar scores Table 3 provides PAI composite and index scores for
across the PAI indices and composites, despite large differ- offenders in the primary and secondary psychopathy sub-
ences between their PCL-R total scores (d = 2.79). type groups (Appendix D provides descriptive data for indi-
Although sociopathic and callous-conning offenders had vidual PAI scales). Those in the primary class reported
similar PCL-R total scores (d = −0.04), they responded dif- significantly lower levels of externalizing psychopathology
ferently to the PAI. Callous-conning offenders had signifi- (d = −0.54) and had significantly lower scores on the vio-
cantly lower externalizing psychopathology (d = −0.45), lence potential (d = −0.74) and treatment process (d =
violence potential (d = −0.43), and problematic treatment −0.55) indices. They also had somewhat higher Suicide
process (d = −0.49) scores, suggesting lower levels of risk Potential Index scores (d = −0.38), although there was not
for violence and higher levels of treatment amenability. a statistically significant difference between the classes for
Callous-conning offenders also had lower ANT scores (d = internalizing psychopathology (d = −0.24) or BOR scores
−0.53) than sociopathic offenders, which is consistent with (d = −0.29). ANT scores were significantly lower for those
callous-conning offenders having lower Facet 3 and Facet 4 in the primary psychopathy group (d = −0.47), which is
PCL-R scores than sociopathic offenders. consistent with them also having lower scores on PCL-R
Overall, callous-conning offenders reported relatively Facets 3 (Lifestyle) and 4 (Antisocial)
low levels of impairment and distress. As indicated in Table
2, they did not differ significantly from general offenders on
any of the PAI indices or composites, despite large differ- Discussion
ences between their PCL-R total scores (d = 2.47). Our goal was to add to the PCL-R subtyping research litera-
ture by providing findings with a clear connection to clini-
cal practice. We used PCL-R scores from a field setting
High Psychopathy Subsample Results with well-documented concerns related to evaluator differ-
Latent Profile Analysis Results.  There were 180 offenders with ences in scoring tendencies, adversarial allegiance, and
a PCL-R total score of 25 or higher. All but one of these attenuated predictive validity (Boccaccini et al., 2014;
McCallum et al. 9

Figure 2.  Two LPA subtype profiles by PCL-R mean item facet score.
Note. Class 1 = primary psychopathy (n = 66); Class 2 = secondary psychopathy (n = 114). LPA = latent profile analysis; PCL-R = Psychopathy
Checklist-Revised.

Table 3.  Psychopathy Checklist Revised Mean Item Scores and Personality Assessment Inventory Composite and Index Scores for
Offenders Each Psychopathy Class in the High Psychopathy Sample: PCL-R Total Score ≥ 25.

Class 1: Primary Class 2: Secondary

PAI Composite/Index M (SD) M (SD) Comparison Cohen’s d


PCL-R: Mean facet score n = 66 n = 114  
  Facet 1: Interpersonal 1.62 (0.32) 1.27 (0.46) t(178) = 5.46, p < .001 0.84
  Facet 2: Affective 1.76 (0.23) 1.63 (0.29) t(178) = 3.12, p = .002 0.48
  Facet 3: Lifestyle 1.27 (0.28) 1.39 (0.30) t(178) = 2.65, p = .009 −0.41
  Facet 4:Antisocial 0.97 (0.21) 1.62 (0.22) t(178) = 19.42, p < .001 −3.00
PCL-R Total (mean item) 1.36 (0.10) 1.47 (0.16) t(178) = 5.04, p < .001 −0.78
PCL-R Total (sum) 27.26 (1.90) 29.93 (3.11)  
PAI Composite/Index n = 51 n = 82  
  Mean Clinical Elevation 53.41 (5.63) 56.30 (7.07) t(131) = 2.48, p = .02 −0.44
 Internalizing 52.18 (7.72) 54.29 (9.45) t(131) = 1.34, p = .18 −0.24
 Externalizing 54.43 (5.91) 57.99 (7.04) t(131) = 3.00, p = .003 −0.54
  Violence Potential Index 54.73 (10.89) 61.16 (12.71) t(120) = 2.83, p = .005 −0.74
  Treatment Process 55.09 (11.55) 62.78 (15.14) t(120) = 2.94, p = .004 −0.55
Index
  Suicide Potential Index 53.49 (10.29) 57.78 (11.87) t(120) = 2.02, p = .05 −0.38
PAI Scale
  Antisocial Features 57.55 (8.58) 61.49 (8.38) t(131) = 2.61, p = .01 −0.47
  Borderline Features 54.33 (9.24) 57.24 (10.74) t(131) = 1.60, p = .12 −0.29

Note. Cohen’s d values in bold are statistically significant at p < .05. d values in bold and italics are statistically significant at p ≤ .01. n
= 45 and 77 for the Violence Potential Index, Treatment Process Index, and Suicide Potential Index. PAI = Personality Assessment
Inventory; PCL-R = Psychopathy Checklist-Revised.
10 Assessment 00(0)

Murrie et al., 2009, 2012). Our identification of prototypic, Those with midrange PCL-R total scores and profiles
callous-conning, sociopathic, and general offender subtypes marked by elevated Facet 4 scores (sociopathic subtype)
in the overall sample points to the robustness of the PCL-R- will likely be at a higher risk for violence, misconduct,
based typology from existing subtype studies (see suicidality, and poor treatment outcomes. Those with mid-
Falkenbach et al., 2014; Neumann et al., 2016) and is range PCL-R total scores marked by elevated Facet 1 and
exactly the type of cross-validation research clinicians 2 scores (callous-conning subtype) will likely be at a lower
should look for to support real-world assessment practices risk for these outcomes.
(Edens & Boccaccini, 2017). PAI findings with higher scorers (PCL-R ≥ 25) provide
The availability of PAI scores for many of the offenders more nuanced information for the highest scoring offend-
in this study allows for a detailed examination of how these ers. Although nearly all of these offenders (179 of 180)
PCL-R subtype findings may be potentially useful for prac- were classified into the prototypic subgroup in the full-sam-
tice. The PAI is one of the more commonly used multiscale ple analyses, the two subgroups that emerged from the
inventories in forensic practice (Neal & Grisso, 2014) with higher scorer LPA still differed somewhat on the PAI. Those
a large and established forensic-sample research literature in the primary and secondary subgroups that emerged from
(see e.g., Boccaccini & Hart, 2018; Gardner et al., 2015; the LPA all had relatively high scores across the PAI, but
Hawes & Boccaccini, 2009; Ruiz & Ochshorn, 2010) and those with elevated PCL-R Facet 4 scores (secondary)
interpretation resources for practitioners (see, e.g., Blais appeared to be most at risk for violence, misconduct, suicid-
et al., 2010; Morey, 2007; Morey & Hopwood, 2007). ality, and poor treatment outcomes as measured by the PAI.
Moreover, the PAI scores in this study are also field scores. Although prior psychopathy subtyping research has doc-
Offenders completed the PAI knowing that the results umented clear differences between classes on measures of
would be available to decision makers, and, thus, represent internalizing and externalizing symptoms (Hicks et al.,
the types of scores that evaluators can expect to see in real- 2004; Skeem et al., 2007; Swogger & Kosson, 2007), we
world sex offender risk assessment contexts. observed smaller differences in internalizing symptoms
than externalizing symptoms on the PAI. For example, the
most notable differences (i.e., d > .35) for scores on mea-
Score Differences and Implications for Practice sures of internalizing symptoms were between the proto-
Several aspects of our findings are consistent with using the typic offenders—who scored higher across the PAI—and
PCL-R as a dimensional measure. Those with the highest other offenders. These findings are consistent with recent
PCL-R scores reported the most pathology across the PAI subtyping research suggesting that higher psychopathy
and those with the lowest PCL-R scores reported the least scores are associated with lower emotion regulation
pathology across the PAI. Overall, these findings suggest (Garofalo et al., 2020). The callous-conning and sociopathic
that prototypic offenders are at the highest risk for a broad subgroups did not differ significantly on the internalizing
range of negative outcomes, including violence, miscon- composite, nor did the primary and secondary subgroups.
duct, suicidality, and poor treatment outcomes. In terms of In prior research, the largest differences for internalizing
absolute value, there was no instance of the prototypic behaviors are between primary and secondary subtypes (see
offenders obtaining a lower PAI composite or index score Hicks & Drislane, 2018). For example, researchers have
than offenders from any other group (see Table 3). found secondary variants to display higher levels of anxiety
At the facet level, those in subtypes with higher Facet 4 and internalizing features than primary variants in both men
scores (i.e., prototypic, sociopathic subtypes) reported more and women (Falkenbach et al., 2017). Others have found
pathology than those with lower scores. Facets 1 and 2, and that although primary and secondary subtypes both display
to some extent Facet 3, were only associated with higher high levels of externalizing behavior, the secondary subtype
pathology among those who also had elevated Facet 4 exhibits comparatively higher rates of internalizing symp-
scores. toms, such as depression, anxiety, and somatization
Perhaps the most clinically useful PAI findings relate to (Drislane et al., 2014). These types of differences were not
differences between the callous-conning and sociopathic readily apparent on the PAI in our study.
subgroups. Although these two subgroups had nearly iden- The subtypes in the high-scoring subsample were also
tical PCL-R total scores (M = 18.11 vs. 18.22), they similar to the aggressive and manipulative subtypes iden-
scored differently on the PAI. Those in the sociopathic tified in some subtyping studies using only those with
subgroup had PAI scores that were similar to those in the high scores, although we did not identify a class consis-
prototypic psychopathic subgroup, whereas those in the tent with the third class (sociopaths) from those studies
callous-conning subgroup had scores that were similar to (Mokros et al., 2015). It could be that the high-scoring
those in the general offender subgroup. The clear implica- sociopath class, which is characterized by especially low
tion for practice from these findings is that practitioners Facet 2 scores, requires especially large samples to detect.
need to go beyond the PCL-R total score when thinking Indeed, the sociopathic class was the smallest in size in
about comorbid psychopathology, risk, and treatment prior studies, including fewer than 15% of offenders
needs for an offender with a mid-range PCL-R score. (Mokros et al., 2015).
McCallum et al. 11

Overall, the PCL-R and PAI findings from this study the extent to which our PAI findings generalize to other
show how LPA-identified psychopathy subgroups differ on offender or patient samples is a question that should be
scores from a commonly used measure of personality and answered through future replication research.
psychopathology. What we do not know is whether clini- Because we sought to examine the generalizability of
cians can take the findings in our tables and figures, look at existing PCL-R subtype findings, many of our analysis (e.g.,
an individual offender’s PCL-R scores, and accurately place using LPA as opposed to cluster analysis) and interpretation
the offender into the “correct” PCL-R subtype group. (e.g., subtype labels) decisions were guided by prior
Although there will be some offenders with facet scores that research. Although our findings were consistent with those
come close to matching the profiles plotted in our figures from prior studies, they also point to issues that require more
(or those from other studies), it seems likely that many oth- research. For example, consistent with prior research (Krstic
ers will be difficult to classify. There is a clear need for this et al., 2018; Lehmann et al., 2019), BIC values were slightly
type of clinician-accuracy research, which we view as an smaller for five-class model than the four-class model, rais-
important next step for examining the clinical utility of ing questions about the most appropriate model. Just as a
PCL-R LPA subtype findings. three-class model has only emerged very large samples of
high scorers, it may be that larger samples may eventually
reveal support for a five-class model when using all scores.
Limitations
Data for this study were provided by psychologists from one
state, who reviewed their records and reported data from
their completed SVP evaluations. Because each offender Conclusion
was evaluated by only one evaluator, there was no way to Consistent with a growing body of research, the present
evaluate interrater reliability or check for scoring or data study suggests there are four identifiable PCL-R score sub-
entry errors. Because PAI data came from records sent to types among sex offenders evaluated in the field, distin-
evaluators, as opposed to scores obtained at the evaluation guishable by profiles of facet scores. These subtypes were
site (i.e., department of corrections), we do not know why further differentiated by levels of comorbid psychopathol-
PAI scores were missing for some offenders. Furthermore, ogy as measured by the PAI. These findings should help
we have incomplete information about offender characteris- assuage concerns about the generalizability of PCL-R sub-
tics, including race and ethnicity, and only limited informa- typing findings to scores assigned for routine clinical/foren-
tion about their offending history. The focus on scores from sic practice and help translate these findings to practice by
sex offender evaluations in one state necessarily limits the showing how those falling into the subtype groups differ on
generalizability of study findings. Although our LPA find- measures of comorbid psychopathology, treatment amena-
ings were consistent with a larger body of PCL-R research, bility, and risk for violence from the widely used PAI.

Appendix A.  Psychopathy Checklist–Revised Mean Item and Summed Scores for Offenders in Each Class: Alternate Five-Class
Solution.
Class 1 Class 2 Class 3 Class 4 Class 5
PCL-R/PAI Score M (SD) M (SD) M (SD) M (SD) M (SD)
Mean facet scores
PCL-R: Mean facet score n = 79 n = 112 n = 168 n = 164 n = 92
  Facet 1: Interpersonal 1.31 (1.67) 2.12 (1.58) 4.57 (1.99) 4.98 (2.01) 2.57 (1.76)
  Facet 2: Affective 1.38 (1.04) 3.54 (1.15) 6.52 (1.05) 6.49 (1.12) 4.44 (1.10)
  Facet 3: Lifestyle 2.05 (1.54) 4.35 (1.83) 5.12 (1.69) 6.62 (1.53) 2.39 (1.47)
  Facet 4: Antisocial 3.13 (1.80) 5.94 (1.50) 3.63 (1.43) 7.37 (1.33) 2.01 (1.53)
  PCL-R Total 9.29 (3.48) 17.68 (3.05) 21.57 (3.49) 27.65 (3.49) 12.27 (3.13)
Mean item score
PCL-R: Mean item score n = 79 n = 112 n = 168 n = 164 n = 92
  Facet 1: Interpersonal 0.33 (0.42) 0.40 (0.40) 1.14 (0.51) 1.25 (0.53) 0.64 (0.44)
  Facet 2: Affective 0.35 (0.26) 0.88 (0.29) 1.63 (0.26) 1.62 (0.28) 1.11 (0.28)
  Facet 3: Lifestyle 0.41 (0.31) 0.87 (0.29) 1.02 (0.37) 1.32 (0.37) 0.48 (0.29)
  Facet 4: Antisocial 0.63 (0.36) 1.19 (0.30) 0.73 (0.29) 1.47 (0.31) 0.40 (0.31)
  PCL-R Total 0.46 (0.17) 0.88 (0.15) 1.08 (0.17) 1.38 (0.17) 0.64 (0.16)

Note. PCL-R = Psychopathy Checklist-Revised; PAI = Personality Assessment Inventory.


12 Assessment 00(0)

Appendix B.  Psychopathy Checklist–Revised Summed Facet Scores for Offenders in each Class: Full Sample and High Psychopathy
Subsample.
Class 1: Prototypic Class 2: Callous-Conning Class 3: Sociopathic Class 4: General

PCL-R M (SD) M (SD) M (SD) M (SD)


Full sample (N = 615)
PCL-R n = 239 n = 154 n = 96 n = 126
  Facet 1: Interpersonal 5.04 (2.00) 3.76 (1.91) 2.01 (1.55) 1.40 (1.62)
  Facet 2: Affective 6.56 (1.12) 5.89 (1.16) 3.30 (1.32) 2.16 (1.39)
  Facet 3: Lifestyle 6.54 (1.51) 3.69 (1.62) 4.64 (1.71) 1.88 (1.42)
  Facet 4: Antisocial 6.54 (1.95) 2.67 (1.60) 6.22 (1.60) 2.76 (1.91)
  Class 1: Primary Class 2: Secondary  
  M (SD) M (SD)  
PCL-R total ≥ 25 subsample (N = 180)
PCL-R: Mean item score n = 66 n = 114  
  Facet 1: Interpersonal 6.48 (1.28) 5.10 (1.85)  
  Facet 2: Affective 7.01 (0.93) 6.48 (1.15)  
  Facet 3: Lifestyle 6.33 (1.40) 6.94 (1.50)  
  Facet 4:Antisocial 4.87 (1.07) 8.12 (1.08)  

Note. PCL-R = Psychopathy Checklist-Revised.

Appendix C.  PAI Scale Scores for Psychopathy Subtypes: Full Sample.

Class 1: Prototypic Class 2: Callous-conning Class 3: Sociopathic Class 4: General


(n = 175) (n = 115) (n = 74) (n = 111)

PAI Scale T-Scores M SD M SD M SD M SD


Inconsistency (INC) 55.43 9.40 51.50 8.76 53.44 8.31 50.48 8.98
Infrequency (INF) 56.47 9.58 55.83 8.56 56.76 9.71 53.60 8.70
Negative Impression (NIM) 54.44 13.32 50.61 9.89 50.85 11.05 48.52 7.12
Positive Impression (PIM) 48.45 10.19 53.25 9.20 51.32 10.09 53.23 9.81
Somatic Complaints (SOM) 53.93 10.91 51.50 9.65 50.57 10.26 49.41 8.96
Anxiety (ANX) 52.31 10.62 49.35 9.97 50.35 10.60 47.52 8.84
Anxiety-Related Disorders (ARD) 55.33 11.41 52.36 11.18 53.66 11.50 50.71 10.66
Depression (DEP) 54.98 10.86 51.00 9.84 52.43 9.66 49.94 8.74
Mania (MAN) 49.01 10.98 47.56 10.02 48.00 10.18 45.57 8.98
Paranoia (PAR) 56.94 11.87 51.52 9.06 52.15 8.94 49.70 10.23
Schizophrenia (SCZ) 51.61 11.85 47.58 10.00 48.78 11.16 45.17 8.66
Borderline Features (BOR) 56.42 10.81 51.13 9.22 53.85 9.77 50.16 9.75
Antisocial Features (ANT) 59.53 8.86 54.37 7.51 58.47 7.92 54.24 8.55
Alcohol Problems (ALC) 58.91 15.64 53.93 11.88 59.59 15.19 55.40 14.52
Drug Problems (DRG) 61.57 15.26 55.46 12.72 61.53 13.90 55.93 13.82
Aggression (AGG) 50.39 10.37 45.56 8.46 50.53 8.48 45.65 9.28
Suicide Ideation (SUI) 52.69 12.05 49.71 10.69 51.39 11.03 48.59 10.99
Stress (STR) 55.55 9.85 53.25 9.94 53.19 9.29 51.58 9.87
Nonsupport (NON) 55.45 12.93 49.39 10.13 51.96 11.06 47.37 10.10
Treatment Rejection (RXR) 41.53 9.03 45.16 10.14 42.16 9.10 44.52 9.31
Dominance (DOM) 49.50 9.17 50.67 9.53 49.27 9.68 50.19 8.08
Warmth (WRM) 48.27 10.21 50.32 9.86 48.35 10.71 52.51 8.63

Note. PAI = Personality Assessment Inventory.


McCallum et al. 13

Appendix D.  PAI Scale Score Descriptive Statistics for Primary and Secondary Psychopathy Subtypes Among Offenders With PCL-R
Total Scores ≥ 25.
Class 1: Primary (n = 51) Class 2: Secondary (n = 82)

PAI Scale T-Scores M SD M SD


Inconsistency (INC) 54.82 8.75 55.99 9.081
Infrequency (INF) 54.76 9.14 56.82 9.19
Negative Impression (NIM) 51.92 10.50 54.94 12.34
Positive Impression (PIM) 49.67 9.75 47.56 10.30
Somatic Complaints (SOM) 54.75 9.92 54.32 11.32
Anxiety (ANX) 50.59 8.98 53.09 11.36
Anxiety-Related Disorders (ARD) 52.94 10.05 56.00 11.90
Depression (DEP) 52.69 10.04 55.85 11.14
Mania (MAN) 48.12 9.69 49.11 10.38
Paranoia (PAR) 54.78 11.24 57.82 11.65
Schizophrenia (SCZ) 49.92 9.78 52.18 11.54
Borderline Features (BOR) 54.33 9.24 57.24 10.74
Antisocial Features (ANT) 57.55 8.58 61.49 8.38
Alcohol Problems (ALC) 55.27 10.48 59.23 15.87
Drug Problems (DRG) 56.55 13.03 63.02 15.74
Aggression (AGG) 47.73 9.11 51.50 9.77
Suicide Ideation (SUI) 50.12 8.50 54.51 14.23
Stress (STR) 53.69 8.84 56.30 9.98
Nonsupport (NON) 51.27 11.80 57.34 12.44
Treatment Rejection (RXR) 45.27 9.34 40.05 8.18
Dominance (DOM) 49.29 8.43 50.26 9.84
Warmth (WRM) 49.20 9.69 48.88 10.39

Note. PAI = Personality Assessment Inventory.

Declaration of Conflicting Interests References


The author(s) declared no potential conflicts of interest with Blackburn, R., Logan, C., Donnelly, J. P., & Renwick, S. D. (2008).
respect to the research, authorship, and/or publication of this Identifying psychopathic subtypes: Combining an empirical
article. personality classification of offenders with the Psychopathy
Checklist-Revised. Journal of Personality Disorders, 22(6),
Funding 604-622. https://doi.org/10.1521/pedi.2008.22.6.604
Blais, M. A., Baity, M. R., & Hopwood, C. J. (2010).
The author(s) received no financial support for the research,
Clinical applications of the Personality Assessment
authorship, and/or publication of this article.
Inventory. Routledge/Taylor & Francis Group. https://doi.
org/10.4324/9780203856604
ORCID iDs Boccaccini, M. T., Chevalier, C., Murrie, D. C., & Varela, J. G.
Katherine E. McCallum https://orcid.org/0000-0002-2011- (2017). Psychopathy Checklist–Revised use and reporting
2371 practices in sexually violent predator evaluations. Sexual
Marcus T. Boccaccini https://orcid.org/0000-0002-1590-4905 Abuse: A Journal of Research and Treatment, 29(6), 592-614.
https://doi.org/10.1177/1079063215612443
Boccaccini, M. T., & Hart, J. R. (2018). Response style on the per-
Public Significance Statement sonality assessment inventory and other multiscale invento-
We replicated existing Psychopathy Checklist–Revised (PCL-R) ries. In R. Rogers & S. D. Bender (Eds.), Clinical assessment
subtype findings using field scores, which often have weaker psy- of malingering and deception (4th ed., pp. 280-300). Guilford
chometric properties than scores assigned for research. Differences Press.
between the subtype groups on measures of psychopathology, Boccaccini, M. T., Murrie, D. C., Hawes, S. W., Simpler,
risk, and treatment amenability from the Personality Assessment A., & Johnson, J. (2010). Predicting recidivism with the
Inventory (PAI) highlight the utility of psychopathy subtype find- Personality Assessment Inventory in a sample of sex offenders
ings for broader clinical decision making. screened for civil commitment as sexually violent predators.
14 Assessment 00(0)

Psychological Assessment, 22(1), 142-148. https://doi.org/ Assessment, 29(6), 599-610. https://doi.org/10.1037/pas0


10.1037/a0017818 000475
Boccaccini, M. T., Murrie, D. C., Rufino, K. A., & Gardner, B. Falkenbach, D. M., Reinhard, E. E., & Larson, F. R. R. (2017).
O. (2014). Evaluator differences in Psychopathy Checklist- Theory based gender differences in psychopathy subtypes.
Revised factor and facet scores. Law and Human Behavior, Personality and Individual Differences, 105(15), 1-6. https://
38(4), 337-345. https://doi.org/10.1037/lhb0000069 doi.org/10.1016/j.paid.2016.09.023
Boccaccini, M. T., Rufino, K. A., Jackson, R. L., & Murrie, D. Falkenbach, D. M., Stern, S. B., & Creevy, C. (2014).
C. (2013). Personality Assessment Inventory scores as pre- Psychopathy variants: Empirical evidence supporting a sub-
dictors of misconduct among civilly committed sex offend- typing model in a community sample. Personality Disorders:
ers. Psychological Assessment, 25(4), 1390-1395. https://doi. Theory, Research, and Treatment, 5(1), 10-19. https://doi.
org/10.1037/a0034048 org/10.1037/per0000021
Boccaccini, M. T., Turner, D. B., & Murrie, D. C. (2008). Do Gardner, B. O., Boccaccini, M. T., Bitting, B. S., & Edens, J. F.
some evaluators report consistently higher or lower PCL-R (2015). Personality Assessment Inventory scores as predic-
scores than others? Findings from a statewide sample of sexu- tors of misconduct, recidivism, and violence: A meta-analytic
ally violent predator evaluations. Psychology, Public Policy, review. Psychological Assessment, 27(2), 534-544. https://
and Law, 14(4), 262-283. https://doi.org/10.1037/a0014523 doi.org/10.1037/pas0000065
Boccaccini, M. T., Turner, D., Murrie, D. C., & Rufino, K. A. Garofalo, C., Neumann, C. S., & Daniel, M. (2020). Associations
(2012). Do PCL-R scores from state or defense experts best between psychopathy and the trait meta-mood scale in incar-
predict future misconduct among civilly committed sex cerated males. Criminal Justice and Behavior, 47(3), 331-
offenders? Law and Human Behavior, 36(3), 159-169. https:// 351. https://doi.org/10.1177/0093854819891460
doi.org/10.1037/h0093949 Hare, R. D. (1991). Manual for the Revised Psychopathy Checklist.
Cohen, J. (1992). A power primer. Psychological Bulletin, 112(1), Multi-Health Systems.
155-159. https://doi.org/10.1037/0033-2909.112.1.155 Hare, R. (2003). Hare Psychopathy Checklist–Revised (PCL-R):
Colins, O. F., Fanti, K. A., Salekin, R. T., & Andershed, H. 2nd edition. Multi-Health Systems.
(2017). Psychopathic personality in the general popula- Hare, R. D., Neumann, C. S., & Mokros, A. (2018). The PCL-R
tion: Differences and similarities across gender. Journal of assessment of psychopathy: Development, structural proper-
Personality Disorders, 31(1), 49-74. https://doi.org/10.1521/ ties, and new directions. In C. Patrick (Ed.), Handbook of psy-
pedi_2016_30_237 chopathy (2nd ed., pp. 39-79). Guilford Press.
Colins, O. F., Fanti, K. A., Salekin, R. T., Mulder, E., & Andershed, Hawes, S. W., & Boccaccini, M. T. (2009). Detection of over-
H. (2018). Psychopathy in detained boys: The search for pri- reporting of psychopathology on the Personality Assessment
mary and secondary variances in a clinical setting. Personality Inventory: A meta-analytic review. Psychological Assessment,
Disorders: Theory, Research, and Treatment, 9(5), 408-419. 21(1), 112-124. https://doi.org/10.1037/a0015036
https://doi.org/10.1037/pas0000277 Hicks, B. M., Vaidyanathan, U., & Patrick, C. J. (2010). Validating
DeMatteo, D., Hart, S. D., Heilbrun, K., Boccaccini, M. T., female psychopathy subtypes: Differences in personality,
Cunningham, M. D., Douglas, K. S., Dvoskin, J. A., Edens, antisocial and violent behavior, substance abuse, trauma, and
J. F., Guy, L. S., Murrie, D. C., Otto, R. K., Packer, I. K., mental health. Personality Disorders: Theory, Research, and
& Reidy, T. J. (2020). Statement of concerned experts on Treatment, 1(1), 38-57. https://doi.org/10.1037/a0018135
the use of the Hare Psychopathy Checklist—Revised in Hicks, B. M., & Drislane, L. E. (2018). Variants (“subtypes”) of
capital sentencing to assess risk for institutional violence. psychopathy. In C. J. Patrick (Ed.), Handbook of Psychopathy
Psychology, Public Policy, and Law, 26(2), 133-144. https:// (2nd ed., pp. 297-332). Guilford Press.
doi.org/10.1037/law0000223 Hicks, B. M., Markon, K. E., Patrick, C. J., Krueger, R. F., &
Douglas, K. S., Hart, S. D., & Kropp, P. R. (2001). Validity of Newman, J. P. (2004). Identifying psychopathy subtypes on
the Personality Assessment Inventory for forensic assess- the basis of personality structure. Psychological Assessment,
ments. International Journal of Offender Therapy and 16(3), 276-288. https://doi.org/10.1037/1040-3590.16.3.276
Comparative Criminology, 45(2), 183-197. https://doi. Kass, R., & Raftery, A. E. (1995) Bayes factors. Journal of the
org/10.1177/0306624X01452005 American Statistical Association, 90(430), 773-795. https://
Drislane, L. E., Patrick, C. J., Sourander, A., Sillanmäki, L., doi.org/10.1080/01621459.1995.10476572
Aggen, S. H., Elonheimo, H., Parkkola, K., Multimäki, P., & Klein Haneveld, E., Neumann, C. S., Smid, W., Wever, E., &
Kendler, K. S. (2014). Distinct variants of extreme psycho- Kamphuis, J. H. (2018). Treatment responsiveness of repli-
pathic individuals in society at large: Evidence from a popula- cated psychopathy profiles. Law and Human Behavior, 42(5),
tion-based sample. Personality Disorders: Theory, Research, 484-495. https://doi.org/10.1037/lhb0000305
and Treatment, 5(2), 154-163. https://doi.org/10.1037/ Krstic, S., Neumann, C. S., Roy, S., Robertson, C. A., Knight, R.
per0000060 A., & Hare, R. D. (2018). Using latent variable- and person-
Edens, J. F., & Boccaccini, M. T. (2017). Taking forensic mental centered approaches to examine the role of psychopathic traits
health assessment “out of the lab” and into “the real world”: in sex offenders. Personality Disorders: Theory, Research,
Introduction to the special issue on the field utility of foren- and Treatment, 9(3), 207-216. https://doi.org/10.1037/
sic assessment instruments and procedures. Psychological per0000249
McCallum et al. 15

Lee, Z., & Salekin, R. T. (2010). Psychopathy in a noninstitutional Nylund-Gibson, K., & Masyn, K. E. (2016). Covariates and
sample: Differences in primary and secondary subtypes. mixture modeling: Results of a simulation study explor-
Personality Disorders: Theory, Research, and treatment, ing the impact of misspecified effects on class enumeration.
1(3), 153-169. https://doi.org/10.1037/a0019269 Structural Equation Modeling, 23(6), 782-797. https://doi.org
Lehmann, R., Neumann, C. S., Hare, R. D., Biedermann, J., Dahle, /10.1080/10705511.2016.1221313
K. P., & Mokros, A. (2019). A latent profile analysis of vio- Olver, M. E., Sewall, L. A., Sarty, G. E., Lewis, K., & Wong, S. C.
lent offenders based on PCL-R factor scores: Criminogenic P. (2015). A cluster analytic examination and external valida-
needs and recidivism risk. Frontiers in Psychiatry, 10, Article tion of psychopathic offender subtypes in a multisite sample of
627. https://doi.org/10.3389/fpsyt.2019.00627 Canadian federal offenders. Journal of Abnormal Psychology,
Miller, H. A., Turner, K., & Henderson, C. E. (2009). 124(2), 355-371. https://doi.org/10.1037/abn0000038
Psychopathology of sex offenders: A comparison of Olver, M. E., Stockdale, K. C., Neumann, C. S., Hare, R. D.,
males and females using latent profile analysis. Criminal Mokros, A., Baskin-Sommers, A., Brand, E., Folino, J.,
Justice and Behavior, 36(8), 778-792. https://doi. Gacono, C., Gray, N. S., Kiehl, K., Knight, R., Leon-Mayer,
org/10.1177/0093854809336400 E., Logan, M., Meloy, J. R., Roy, S., Salekin, R. T., Snowden,
Miller, C. S., Kimmonis, E. R., Otto, R. K., Kline, S. M., & R., Thomson, N., . . . Yoon, D. (2020). Reliability and valid-
Wasserman, A. L. (2012). Reliability of risk assessment ity of the Psychopathy Checklist-Revised in the assessment of
measures used in sexually violent predator proceedings. risk for institutional violence: A cautionary note on DeMatteo
Psychological Assessment, 24(4), 944-953. https://doi. et al. (2020). Psychology, Public Policy, and Law, 26(4), 490-
org/10.1037/a0028411 510. https://doi.org/10.1037/law0000256
Mokros, A., Hare, R. D., Neumann, C. S., Santila, P., Habermeyer, E., & Percosky, A., Boccaccini, M. T., Bitting, B., & Hamilton, P.
Nitschke, J. (2015). Variants of psychopathy in adult male offend- (2013). Personality Assessment Inventory scores as predictors
ers: A latent profile analysis. Journal of Abnormal Psychology, of treatment compliance and misconduct among sex offend-
124(2), 372-386. https://doi.org/10.1037/abn0000042 ers participating in community-based treatment. Journal of
Morey, L. C. (1991). Personality Assessment Inventory: Forensic Psychology Practice, 13(3), 192-203. https://doi.org
Professional manual. Psychological Assessment Resources. /10.1080/15228932.2013.795425
https://doi.org/10.1037/t03903-000 Ruiz, M. A., Cox, J., Magyar, M., & Edens, J. F. (2014). Predictive
Morey, L. C. (2007). Personality Assessment Inventory (PAI): validity of the Personality Assessment Inventory (PAI) for
Professional manual (2nd ed.). Psychological Assessment identifying criminal reoffending following completion of an
Resources. in-jail addiction treatment program. Psychological Assessment,
Morey, L. C., & Hopwood, C. J. (2007). Casebook for the 26(2), 673-678. https://doi.org/10.1037/a0035282
Personality Assessment Inventory (PAI): A structural sum- Ruiz, M. A., & Edens, J. R. (2008). Recovery and replication of inter-
mary approach. Psychological Assessment Resources. nalizing and externalizing dimensions within the Personality
Murrie, D. C., Boccaccini, M. T., Rufino, K., & Caperton, J. Assessment Inventory. Journal of Personality Assessment,
(2012). Field validity of the Psychopathy Checklist–Revised 90(6), 585-592. https://doi.org/10.1080/00223890802388574
in sex offender risk assessment. Psychological Assessment, Ruiz, M. A., & Ochshorn, E. (2010). Clinical applications of the
24(2), 524-229. https://doi.org/10.1037/a0026015 PAI in criminal justice settings. In M. A. Blais, M. R. Baity, &
Murrie, D. C., Boccaccini, M. T., Turner, D., Meeks, M., Woods, C. J. Hopwood (Eds.), Clinical applications of the Personality
C., & Tussey, C. (2009). Rater (dis)agreement on risk assess- Assessment Inventory (pp. 113-134). Routledge/Taylor &
ment measures in sexually violent predator proceedings: Francis Group.
Evidence of adversarial allegiance in forensic evaluation? Skeem, J., Johansson, P., Andershed, H., Kerr, M., & Louden, J. E.
Psychology, Public Policy, and Law, 15(1), 19-53. https://doi. (2007). Two subtypes of psychopathic violent offenders that
org/10.1037/a0014897 parallel primary and secondary variants. Journal of Abnormal
Muthén, L. K., & Muthén, B. O. (1998-2008). Mplus user’s guide Psychology, 116(2), 395-409. https://doi.org/10.1037/0021-
(5th ed.). Muthén & Muthén. 843X.116.2.395
Muthén, B., & Shedden, K. (1999). Finite mixture model- Swogger, M. T., & Kosson, D. S. (2007). Identifying sub-
ing with mixture outcomes using the EM algorithm. types of criminal psychopaths: A replication and extension.
Biometrics, 55(2), 463-469. https://doi.org/10.1111/j.0006- Criminal Justice and Behavior, 34(8), 953-970. https://doi.
341X.1999.00463.x org/10.1177/0093807300758
Neal, T. M. S., & Grisso, T. (2014). Assessment practices and expert Texas Health and Safety Code, Title 11, § 841. (2010).
judgment methods in forensic psychology and psychiatry: An Vassileva, J., Kosson, D. S., Abramowitz, C., & Conrod, P. (2005).
international snapshot. Criminal Justice and Behavior, 41(12), Psychopathy versus psychopathies in classifying criminal
1406-1421. https://doi.org/10.1177/0093854814548449 offenders. Legal and Criminological Psychology, 10(1), 27-
Neumann, C. S., Vitacco, M. J., & Mokros, A. (2016). Using both 43. https://doi.org/10.1348/135532504X15376
variable-centered and person-centered approaches to under- Viljoen, J. L., McLachlan, K., & Vincent, G. M. (2010). Assessing
standing psychopathic personality. In C. B. Gacono (Ed.), The violence risk and psychopathy in juvenile and adult offenders:
clinical and forensic assessment of psychopathy: A practitio- A survey of clinical practices. Assessment, 17(3), 377-395.
ners guide (2nd ed., pp. 14-31). Routledge. https://doi.org/10.1177/1073191109359587

You might also like