You are on page 1of 10

Scandinavian Journal of Psychology, 2019, 60, 447–455 DOI: 10.1111/sjop.

12566

Health and Disability


Psychopathy in a forensic sample – the factor structure of the PCL:SV
in a Danish forensic sample
SUNE BO,1 LISELOTTE PEDERSEN,1 KARL BANG CHRISTENSEN2 and KIRSTEN RASMUSSEN3
1
Psychiatric Research Unit, Region Zealand, Slagelse, Denmark
2
Department of Biostatistics, University of Copenhagen, Copenhagen, Denmark
3
Brøset Centre for Research and Education in Forensic Psychiatry, Trondheim & Department of Psychology, NTNU, Trondheim, Norway

Bo, S., Pedersen, L., Christensen, K. B. & Rasmussen, K. (2019). Psychopathy in a forensic sample – the factor structure of the PCL:SV in a Danish
forensic sample. Scandinavian Journal of Psychology, 60, 447–455.

The research literature reveals an ongoing debate regarding the most appropriate conceptualization of psychopathic personality disorder. Specifically, it is
discussed to what degree antisocial behavior is part of the conceptualization of the psychopathy construct and what constitutes the best factor model of the
Psychopathy Checklist scales. The aim of the present study is to consider the underlying factor structure of the PCL:SV (Psychopathy Checklist Screening
Version) in a Danish sample as well as considering the role of antisocial behavior in the psychopathy construct. Data from a Danish forensic patient
sample (N = 225) was used and item response theory (IRT), aonfirmatory factor analyses (CFA) and structural equations model (SEM) analyses were
carried out. Overall, the findings suggest appropriate item and model fit for the PCL:SV as well as superiority of the three-factor model over the four-factor
model. The results are discussed in relation to the broader concept of personality disorder as well as clinical practice in regards to violence risk assessments
and treatment of psychopathy.
Key words: Antisocial behavior, PCL, personality disorder, personality traits, psychopathy.
Sune Bo, Psychiatric Research Unit, Region Zealand Slagelse, Denmark. E-mail: subh@regionsjaelland.dk

INTRODUCTION with criminal behavioral patterns, and others claim that


The research literature reveals an ongoing debate regarding the psychopathy is better defined without the criminal behavioral
most appropriate conceptualization of personality disorders (PDs) components, and that the latter is a consequence of the former.
in general and psychopathic personality disorder in particular This debate has primarily originated from a dispute about what
(e.g., Blackburn, 2007; Skeem & Cooke, 2010a). The criteria constitutes the best factor model of the best validated and most
defining the specific PDs described in Section II of the Diagnostic applied psychopathy scale in research and clinical practice; the
and statistical manual of mental disorders, fifth edition (DSM-5; Psychopathy Checklist Revised (PCL-R; Hare, 2003) and its
American Psychiatric Association, 2013) are a conglomerate of derivates PCL:SV (Psychopathy Checklist Screening Version;
traits, feelings, thoughts and concrete behavior. However, due to Hart, Cox & Hare, 1995) and PCL:YV (Hare PCL Youth
extensive critic of the prevailing system an alternative model for Version; Forth, Kosson & Hare, 2003). The Psychopathy
classifying PDs has been introduced in Section III in the DSM-5. Checklists follow the expert rater format based on a semi-
This hybrid-model entails an assessment of levels of personality structured interview and file material. Good internal consistency,
functioning and pathological personality traits, and meets some of inter-rater reliability and validity of the PCL have been
the concerns raised that PDs are a mixture of both behavior and demonstrated in a range of studies. (see the manuals for extensive
traits. reviews: Forth et al., 2003; Hare, 2003; Hart et al., 1995).
For psychopathic personality, the debate more specifically Despite the controversy surrounding the description of the
concerns to what degree antisocial and criminal behavior should cardinal features of psychopathy and whether criminal behavior
be included in the conceptualization of the construct (Boduszekab pertains to or is a related feature, the PCL-R and its derivates,
& Debowskac, 2016; Storey, Hart, Cooke & Michie, 2016). There including the PCL:SV are acknowledged as gold standard
seems to be a general agreement that antisocial behavior is a part instruments for measuring psychopathy (Acheson, 2005; Hare
of psychopathic personality (Hare, Neumann & Mokros, 2018; et al., 2018). Most research into the PCL instruments have
Neumann, Hare & Pardini, 2015), however, the disagreement focused on the properties of the PCL-R instrument and the
specifically concerns if criminal behavior is a core feature of majority of the available meta-analytic reviews have also focused
psychopathy or should be considered a secondary or related on the PCL-R (Edens, Campbell & Weir, 2007; Guy & Douglas,
feature (CAPP, 2008; Skeem & Cooke, 2010a, 2010b). Antisocial 2006; Leistico, Salekin, DeCoster & Rogers, 2008; Salekin,
behavior could be defined as the type of behavior that clashes Rogers & Sewell, 1996). In some of the reviews data from the
with the social order, and criminal behavior is the specific type of PCL:SV have been included, but only one study specifically
antisocial behavior that is sanctioned by the legal system (Skeem explores the psychometric properties of the PCL:SV in forensic
& Cooke, 2010a). settings (Higgs et al., 2018). In this review the authors compared
Thus, some argue that psychopathy is best defined as a the PCL:SV with the full PCL-R and found comparable validity
combination of interpersonal and affective dysfunctions along and reliability as established for the PCL-R in forensic settings.

© 2019 Scandinavian Psychological Associations and John Wiley & Sons Ltd
448 S. Bo et al. Scand J Psychol 60 (2019)

Furthermore they reported diagnostic concordance between the psychopathy factor, and other studies have failed to report
two instruments, and the PCL:SV performed equally well in terms sufficient model fits to determine the most appropriate latent
of predictive validity of violence (Higgs et al., 2018). structure of the psychopathy construct (Boduszekab &
High scores on the PCL scales have repeatedly shown to be Debowskac, 2016). The research literature suggests empirical
associated with antisocial and violent behavior (Guy, Edens, support for superiority of the three-factor (e.g., Cooke & Michie,
Anthony & Douglas, 2005; Hemphill, Hare & Wong, 1998; 2001; Skeem, Mulvey & Grisso, 2003) and the four-factor model,
Leistico et al., 2008). For that reason they are frequently used in respectively (e.g., Bolt, Hare, Vitale & Newman, 2004; Hare &
relation to assessment of violence risk (Guy, Douglas & Hendry, Neumann, 2008; Hare, 2003)(Hare et al., 2018). The preferred
2010). Research has shown that factor 2 and especially the four-factor model suggested by Hare et al. (2018) is the
antisocial factor is important in relation to violence (Leistico correlated four-factor model.
et al., 2008). However, recent meta-analyses show the PCL scales The PCL scales have been used for several years both in
to have the lowest predictive validity compared to a range of research and clinical contexts in Denmark (see e.g., Andersen,
violence risk assessment tools (Singh, Grann & Fazel, 2011) as Sestoft, Lillebaek, Mortensen & Kramp, 1999; Bengtson &
well as factor 1 to be predictive at chance level only (Yang, Pedersen, 2008; Pedersen, Kunz, Rasmussen & Elsass, 2010) and
Wong & Coid, 2010). Using specialized tools for assessing have been available for quite a while in Danish (Hare, 2003).
violence risk is found to be superior to utilizing the PCL scales Despite the widespread use, to date, no study has been carried out
for this purpose. This is, however, apparent, since the PCL scales in relation to the debated factor models in a Danish data sample.
should be better used as intended (i.e., to measure psychopathic In order to contribute to the cross-group/cultural discussion of the
personality traits) and not in order to measure violence risk (Guy dimensionality and underlying factor structure of the PCL scales
et al., 2010; Singh et al., 2011). this study aims to investigate the fit of items and suggested factor
Recently, it has been suggested to leave out the criminal models in PCL:SV data from a Danish forensic sample. The
behavior factor of the PCL-R when utilizing this scale to assess overall aim is first to investigate to what degree the individual
psychopathy in clinical practice (Cooke & Sellbom, 2018; PCL:SV items have appropriate item fits and relations to the
Preszler, Marcus, Edens & McDermott, 2018; Skeem & Cooke, latent trait of psychopathy. Second, the aim is to explore whether
2010a). Excluding criminal behavior may support the the three-factor model without testlets or the four-factor model
understanding of the relationship, and particularly a possible have appropriate model fit and to what extent one is superior to
functional link, between psychopathy and criminal behavior. As the other, thus exploring the relevance of antisocial behavior in
of today there is, however, still an ongoing debate as to leave out the construct of and clinical work around psychopathy.
or include the criminal aspects as a central feature defining
psychopathy.
Some of the initial studies exploring the underlying latent METHODS
structure of the PCL, applied principal component analyses and
found five factors (Hare, 1980). Later, more complex statistics The study sample
were applied and various factor solutions emerged (Hare et al., The study sample consisted of 225 male forensic psychiatric
2018). To date, three models of psychopathy have been proposed patients and prisoners from three treatment institutions in the east
based on the PCL measures; two-, three-, and four-factor models part of Denmark. Samples are derived from a larger study on
(Cooke, Michie & Skeem, 2007). The original two-factor model violence risk and psychopathy.
(Factor 1, Interpersonal/affective and Factor 2, Social deviance) Mean age of the sample was 35.3 years (SD = 9.9, range 19–
was proposed by Harpur, Hakstian & Hare (1988) and later 64). Sixty-four percent were of non-Danish ethnicity and a
revised with the revision of the PCL in 2003 (Hare, 2003). majority had either no education or completed compulsory
Harpur and colleagues’ method using split-half cross-validation schooling only (76%). All participants had at least one conviction,
relying on a measure of factor similarity meet criticism (Floyd & and the vast majority had a previous conviction for violent crime
Widaman, 1995), however, studies supported the two-factor (96.5%). The participants had on average been sentenced nine
model (Cooke, 1995). The three-factor model (Factor 1, Arrogant times (range 1–47).
and Deceitful Interpersonal Style; Factor 2, Deficient Affective
Experience, and Factor 3, Impulsive and Irresponsible Behavioral
Style) leaving out the antisocial factor was suggested as an Measure
alternative to the two-factor model (Cooke & Michie, 2001), and The Psychopathy Checklist: Screening Version (PCL:SV; Hart
the four-factor model (Interpersonal; Affective; Lifestyle, and et al., 1995) is a 12-item instrument designed to measure
Antisocial behavior) including the antisocial factor as an psychopathic traits (see Table 1 for a description of the individual
alternative to the three-factor model (Hare, 2003). Several items). It is a brief form derived from the Psychopathy Checklist
versions of the three and four-factor models have been proposed, – Revised (PCL-R; Hare, 2003). It has good reliability and
for example, hierarchical or correlated models and with or without validity and has been found conceptually and empirically related
testlets or parcels (Cooke et al., 2007). Cooke et al. (2007) found to the PCL-R (Cooke, Michie, Hart & Hare, 1999; Guy &
more than 10 four-factor models, including a four-factor Douglas, 2006). Extensive reviews of the psychometric properties
correlated model, a four-factor hierarchical model and a two- of the scale are reported in the manual. Each item is scored as 0
factor with four-facet hierarchical model. Some models have (not present), 1 (possibly present), or 2 (definitely present). The
parceled a subset of items and used these to indicate an overall maximum score is 24 and the most commonly used score as

© 2019 Scandinavian Psychological Associations and John Wiley & Sons Ltd
Scand J Psychol 60 (2019) PCL:SV in a Danish forensic sample 449

indicative for psychopathy is 18 and higher. The items can be Procedure


grouped into two parts: part 1 (items 1–6) relating to interpersonal All assessments were made by trained clinicians (i.e.,
and affective characteristics and part 2 (items 7–12) relating to psychologists and psychiatrists) as part of a structured assessment
antisocial and unstable lifestyle. Furthermore, the two parts can be of violence risk. Assessments were conducted as a clinical routine
divided into four factors: Interpersonal (item 1, 2, 3), Affective at the participating institutions or retrospectively as on the day of
(item 4, 5, 6), Lifestyle (item 7, 9, 10), and Antisocial (item 8, 11, discharge. All assessments were based on very thorough and
12). The original North American manual was used in all parts of comprehensive information, including forensic psychiatric
this research, since no Danish language version of the PCL:SV evaluations, psychological test results, court records and records
exists. The PCL:SV was implemented at the institutions prior to of clinical observations and evaluations from psychiatrists,
data collection. Two independent inter-rater reliability studies psychologists, nurses and social workers, and patient interview
showed excellent interrater reliability of the PCL:SV when used (the latter were not available for the 2001–2002 retrospective
in the participating institutions (Pedersen & Rasmussen, 2006; sample). For inter-rater reliability purposes two studies were
Pedersen et al., 2010). To evaluate inter-rater reliability, a trained conducted prior to and during this data collection (Pedersen &
rater reviewed the file of every fifth patient. Interrater reliability of Rasmussen, 2006; Pedersen et al., 2010). Furthermore,
PCL:SV total scores was excellent, ICC1 = 0.82. throughout the data collection all participating clinicians from all
The National Crime Register (Det Centrale Kriminalregister, three institutions took part in a series of sessions concerned with
2009) shows all known criminal acts committed by a person discussing the administration and scoring of the PCL:SV. The
above the criminal responsible age of 15 years. Printed PCL:SV was chosen due to the shorter administration time
complete transcripts of criminal cases were received for all compared to the PCL-R. Prior to the data collection the study was
individuals included. All criminal cases were carefully studied reported to the Regional Ethics Committee of the county of
and essential information was extracted, that is, detailed Copenhagen as well as the Danish Data Protection Agency. In
information regarding all sentences. Only solved criminal cases order to include the full patient samples it was decided to apply
were included, that is, charges awaiting sentencing were not the Danish National Board of Health for an extraordinary
included in this study. dispensation to circumvent patient consent. This dispensation was
Crime was grouped into two overall categories; any violent granted, and consequently all material on all patients could be
crime and any non-violent crime, with a range of specific included in the study.
categories following the Danish Penal Code. Classified as violent
crimes were: homicide, attempted homicide, aggravated assault
and violence, violence, arson, other violence against the person, Statistical analyses
rape and attempted rape, other sexual offences, and robbery.
The statistical software used to conduct the descriptive analyses
Classified as non-violent crimes were: malicious damage to
was SPSS for Windows, version 19.0 and SPSS for Macintosh,
property, aggravated drug offences under the penal code, burglary
version 20.0. Confirmatory factor analyses (CFA) were performed
and theft, other misappropriations and offences against property,
using AMOS 20 for windows.
homicide or bodily harm by negligence, other sections of the
CFA were used to quantify the degree of fit of different
Penal code, special laws; Euphoriants Act, Arms Act, Road
factor structures, that is, two-factor, three-factor and four-factor
Traffic Act, and other special laws.
model. The following indexes were used to assess the fitness:
absolute fit (v2), non-normed fit index (NNFI), comparative fit
Table 1. PCL:SV scores for 225 male forensic patients and prisoners
index (CFI), and root means square error of approximation
Description (Max score) M (SD) (RMSEA). Criterion for adequate fit was defined as CFI and
NNFI ≥ 0.90 and RMSEA ≤ 0.08. RMSEA was classified into
PCL:SV (24)a 13.64 5.43 four categories: close fit (0.00–0.05), fair fit (0.05–0.08),
Part 1 (12) 6.21 3.09 mediocre fit (0.08–0.10), and poor fit (>0.10) (Kim & Hagtvet,
Superficial 0.79 0.72
2003). The Akaike information criteria (AIC) was used to
Grandiose 0.68 0.75
Deceitful 0.93 0.78 determine the best model fit, where the lowest AIC value
Lacks remorse 1.3 0.68 indicate best model fit (Anderson, 2008; Burnham, 2004;
Lacks empathy 1.25 0.66 Burnham & Anderson, 2002).
Doesn’t accept responsibility 1.25 0.68 We fitted an item response theory (IRT) model, the generalized
Part 2 (12) 7.44 2.9
partial credit model (GPCM) (Muraki, 1992) to each of the four
Impulsive 1.28 0.74
Poor behavioral controls 1.19 0.70 factors separately and evaluated the fit of the model by comparing
Lacks goals 1.31 0.66 observed and expected counts. We also tested whether a more
Irresponsible 1.23 0.71 parsimonious model, the partial credit model (PCM; Masters,
Adolescent antisocial behavior 0.89 0.81 1982) fitted the data equally well. This was done using a
Adult antisocial behavior 1.54 0.61
likelihood ratio test. The generalized partial credit model assumes
Note: aFactor 1/Interpersonal (Superficial, Grandiose, Deceitful), Factor for item i that:
2/Affective (Lacks remorse, Lacks empathy, Doesn’t accept responsibility),
Factor 3/Lifestyle (Impulsive, Lack goals, Irresponsible) Factor 4/
1
Antisocial (Poor behavioral controls, Adolescent antisocial behavior, Adult PðRESPONSE ¼ 0Þ ¼ ;
antisocial behavior). 1 þ expðaðh  b1 ÞÞ þ expðað2h  b1  b2 ÞÞ

© 2019 Scandinavian Psychological Associations and John Wiley & Sons Ltd
450 S. Bo et al. Scand J Psychol 60 (2019)

analysis model where factor loadings are equal. Further the model
expðaðh  b1 ÞÞ
PðRESPONSE ¼ 1Þ ¼ ; belongs to the class of IRT models called Rasch models (Fischer
1 þ expðaðh  b1 ÞÞ þ expðað2h  b1  b2 ÞÞ
& Molenaar, 1995) where the sum of the items is sufficient for h
and thus contains all information about the construct. For these
expðað2h  b1  b2 ÞÞ
PðRESPONSE ¼ 2Þ ¼ ; models item characteristic curves showing the three response
1 þ expðaðh  b1 ÞÞ þ expðað2h  b1  b2 ÞÞ
probabilities for each item as a function of the underlying variable
and is similar to the IRT models previously used for analysis of can be plotted. The item parameters b1 and b2 can be interpreted
the Hare Psychopathy Checklist – Revised (Cooke & Michie, as the intersection point on the latent continuum of adjacent
1997). According to a taxonomy of item response theory models categories: b1 is the location on the underlying h axis where the
(Thissen & Steinberg, 1986) this model is a “divide-by-total” probabilities P(RESPONSE = 0) and P(RESPONSE = 1) are
model whereas the Samejima model used earlier is a “difference” equal, and b2 is the location on the underlying h axis where the
model. However, the basic structure is the same with a probabilities P(RESPONSE = 1) and P(RESPONSE = 2) are
unidimensional trait h underlying the responses and item equal. For this reason the parameters b1 and b2 are called
characteristic curves describing the probability of 0, 1, and 2, threshold parameters. Furthermore, observed and expected item
respectively across values of h. A simpler model, the partial credit response among those with the same total score on the remaining
model (Masters, 1982), assumes for item i that: items can be compared yielding an evaluation of model fit
(Kreiner, 2011).
1 We further analyzed the data using a structural equations model
PðRESPONSE ¼ 0Þ ¼ ;
1 þ expðh  b1 Þ þ expð2h  b1  b2 Þ (SEM) including an additional exogenous variables in order to
distinguish between the three-factor model of psychopathy and
expðh  b1 Þ
PðRESPONSE ¼ 1Þ ¼ ; the four-factor model of psychopathy. In both models we
1 þ expðh  b1 Þ þ expð2h  b1  b2 Þ
specified the factors “Interpersonal” (items P1, P2, P3),
expð2h  b1  b2 Þ “Affective” (items P4, P5, P6), “Lifestyle” (items P7, P9, P10),
PðRESPONSE ¼ 2Þ ¼ ;
1 þ expðh  b1 Þ þ expð2h  b1  b2 Þ and “Antisocial” (P8, P11, P12). Figure 1 shows a four-factor
hierarchical model of psychopathy (left panel) and three-factor
or equivalently that the discrimination parameters a are equal model (right panel) where the factor “Antisocial” is a separate
across items. This simpler model can be interpreted as a factor latent variable that is not a part of the psychopathy construct. In

Fig. 1. Two models of psychopathy (F1: Factor 1/Interpersonal (items P1, P2, P3), F2: Factor 2/Affective (items P4, P5, P6), F3: Factor 3/Lifestyle (items
P7, P9, P10) F4: Factor 4/Antisocial (items P8, P11, P12). P: the underlying construct). Four-factor hierarchical model (left panel) and three-factor
hierarchical model (right panel). It is impossible to distinguish between these models based on the observed item responses. [Colour figure can be viewed
at wileyonlinelibrary.com]

© 2019 Scandinavian Psychological Associations and John Wiley & Sons Ltd
Scand J Psychol 60 (2019) PCL:SV in a Danish forensic sample 451

the three-factor model a single underlying construct influences the RESULTS


first three of these while the fourth factor “Antisocial” is an
independent latent variable. In the four-factor model the PCL:SV measure
underlying construct influences all factors. It is impossible to The mean score of the PCL:SV was 13.24 (range 0-24;
distinguish between these models based on the observed item SD = 5.43). See Table 1 for individual factor and item scores.
responses. In the SEM shown in Fig. 2 the total effect of the There were no significant differences in PCL:SV scores between
exogenous variable on the factor F4 is split into a direct effect “b” the included subsamples.
and an indirect effect “ac” mediated through the underlying
construct.
In the four-factor hierarchical model depicted in the left panel CFA
of Fig. 1 “Antisocial” is a factor of the underlying construct Fit-indexes varied across the three models. The two-factor model
and exogenous variables will only have an indirect effect generally showed a poor fit v2 (53, n = 225) = 142.38
mediated through the underlying construct on this factor. Thus, (p < 0.001); CFI = 0.90; RMSEA = 0.09; NNFI = 0.88 and will
evidence of a direct effect can be taken as evidence against the not be considered further in the analyses. The three-factor model
four-factor model. We tested whether or not there was a generally shows a good fit, v2 (24, n = 225) = 40.25 (p = 0.02);
significant direct effect by likelihood ratio tests comparing CFI = 0.97; RMSEA = 0.06; NNFI = 0.96, and the four-factor
nested models. This was done for the three exogenous model generally shows a fair fit, v2 (50, n = 225) = 117.50
variables: age, education, and ethnicity (being Danish or not) on (p < 0.001); CFI = 0.92; RMSEA = 0.08; NNFI = 0.90.
each of the factors. The three and the four-factor model were compared using AIC;
according to this criteria for evaluating three-factor model
(AIC = 82.25) is superior to the four-factor model
(AIC = 173.50). Unidimensionality was rejected.

IRT
We studied the four factors separately. For all factors the partial
credit model was tested against the generalized partial credit
model and the two models were found to fit the data equally
well. Item parameters and item fit statistics are presented in
Table 2. No evidence against the model was disclosed by the
item fit statistics. The item characteristic curves are shown in
Fig. 3. The item characteristic curves are similar, but differences
appear with respect to the location and spread of the threshold
parameters. For the two items “Lacks empathy” and “Doesn’t
accept responsibility” the threshold parameters are far apart
indicating that both of these items can provide information
across the whole range of values of the underlying construct.
Conversely, the two items “Adolescent Antisocial Behavior” and
“Adult Antisocial Behavior” are seen to provide information at
different ends of the latent continuum. The first of these items
provides information at the higher end of the continuum while
the latter provides information at the lower end of the
continuum.
In an additional analysis a unidimensional IRT model including
all 12 items was clearly rejected.

SEM
We analyzed the data using the SEM shown in Fig. 2 testing for a
direct effect of the exogenous variables age, ethnicity and
education on each of the four factors. The results are shown in
Table 3. If Factor 4/Antisocial was a factor of the underlying
Fig. 2. Four-factor hierarchical model of psychopathy (F1: Factor 1/ construct we would expect no direct effect of exogenous variables
Interpersonal (items P1, P2, P3), F2: Factor 2/Affective (items P4, P5, P6), on Factor 4, but only an indirect effect mediated through the
F3: Factor 3/Lifestyle (items P7, P9, P10) F4: Factor 4/Antisocial (items
underlying construct. The likelihood ratio tests reported in
P8, P11, P12). P: the underlying construct. Exogenous variable EXO). In
this structural equation model a direct effect of “Ethnicity” of the factor F4 Table 3 showed highly significant direct effects of age and
constitutes evidence against the four-factor model. [Colour figure can be education on Factor 4/Antisocial. Similar analyses for the three
viewed at wileyonlinelibrary.com] other factors showed only borderline significant evidence of direct

© 2019 Scandinavian Psychological Associations and John Wiley & Sons Ltd
452 S. Bo et al. Scand J Psychol 60 (2019)

Fig. 3. Item characteristic curves for the 12 PCL:SV items.

effects. The direct effects constitute evidence against the four- exclusion of the criminal and antisocial behavior in the construct
factor model. of psychopathy.
Item response theory analyses showed good structural and
metric properties in this Danish sample. For each of the four
DISCUSSION factors, the items fitted the data equally well and item fit statistics
The aim of the present study was to investigate to what degree showed no evidence against either model. This means that each of
the individual PCL:SV items have appropriate item fits and the four factors Interpersonal, Affective, Lifestyle, and Antisocial
relations to the latent trait of psychopathy in a Danish forensic are measured accurately. Item characteristic curves were found to
sample. Furthermore, an exploration of model fit and superiority be similar, however, differences were found with respect to item
was carried out for the suggested three and four-factor models. representation across the underlying construct. P5 and P6 were
The purpose of the study was to add to the cross-group/cultural identified as the items covering the full continuum of the
discussion of the dimensionality and underlying factor structure of underlying construct, whereas P11 and P12 were identified as
the PCL scales as well as to the debate concerning in- or providing information at ends of the continuum only. The

© 2019 Scandinavian Psychological Associations and John Wiley & Sons Ltd
Scand J Psychol 60 (2019) PCL:SV in a Danish forensic sample 453

Table 2. Item parameters and item fit statistics for the PCL:SV items

Item fit statistic

Factor Item Beta1 Beta2 obs. exp. SD p

Factor 1Interpersonal Superficial 1.204 1.338 0.642 0.610 0.055 0.5533


Grandiose 0.565 1.305 0.656 0.634 0.055 0.6806
Deceitful 1.373 0.499 0.575 0.615 0.053 0.4518
Factor 2Affective Lacks remorse 1.751 1.503 0.807 0.756 0.050 0.3017
Lacks empathy 1.826 1.903 0.749 0.759 0.049 0.8337
Doesn’t accept respon. 1.646 1.817 0.754 0.756 0.049 0.9586
Factor 3Lifestyle Impulsive 0.894 0.994 0.608 0.605 0.061 0.9670
Lack goals 1.731 1.293 0.595 0.611 0.063 0.8024
Irresponsible 1.050 1.388 0.637 0.605 0.061 0.6059
Factor 4Antisocial Poor behavioral controls 1.263 1.464 0.543 0.549 0.063 0.9234
Adolescent antisocial behav. 0.614 1.585 0.575 0.554 0.062 0.7362
Adult antisocial behav. 2.531 0.132 0.598 0.589 0.068 0.8914

Table 3. SEM analyses estimating direct effect of exogenous variables of against the four-factor model. Hence, the SEM analyses are in
the four factors accordance with the CFA and AIC statistics indicating superiority
of the three-factor model over the four-factor model in our Danish
Factor Exogenous LRT DF p
sample. Based on a Danish sample, this suggests that
Factor 1 Interpersonal Age 0.004 1 0.9496 psychopathic personality disorder may best be conceptualized
Education 4.219 1 0.0400 around the core features of Interpersonal, Affective and Lifestyle
Ethnicity 0.038 1 0.8454 components. These findings are in accordance with studies
Factor 2 Affective Age 2.374 1 0.1234 specifically exploring the PCL:SV applying network analysis,
Education 3.741 1 0.0531
where findings showed that it was specifically the affective facet
Ethnicity 1.288 1 0.2564
Factor 3 Lifestyle Age 3.496 1 0.0615 that were central to psychopathy, and that, interpersonal, lifestyle,
Education 0.113 1 0.7368 and antisocial facets generally showed low centrality (Preszler
Ethnicity 0.000 1 1.0000 et al., 2018). Thus, the results using the PCL:SV lend support to
Factor 4 Antisocial Age 13.203 1 0.0003 the importance of affective deficits as the primary feature of
Education 8.595 1 0.0034
psychopathy. Other studies explicitly applying the PCL:SV have
Ethnicity 0.776 1 0.3784
found little difference between the two-,three-, or four-factor

model (Zukauskien_ e, Laurinavicius & C_ esnien_e, 2010) (Wilson
individual items thus differ with respect to the amount of et al., 2014) suggesting that these studies have been unable to
information they can provide across the range of the underlying decide which models were superior.
construct. This is similar to results found by Cooke and Michie Overall, our findings accentuate less focus on antisocial
(1997). The GPCM and the PCM both fit the data equally well, behavior, contributing to the current debate regarding antisocial
and we opted for the more parsimonious model. behavioral features as part of the psychopathy construct.
Confirmatory factor analyses showed variances in fit-indexes Furthermore, our findings may have practical concerns for
for the different factor models. As expected, a poor fit was found clinicians with respect to understanding and working with
for the two-factor model. psychopathy in forensic settings. The PCL:SV has been available
This is in line with a range of studies showing poor fit for this in Danish for a while, but with no studies examining the
model and consistently finding the two-factor model to be inferior to competing Factor models in a Danish sample.
the three and four-factor models (Cooke & Michie, 2001; Hare, Though many psychopathic individuals have no history of
2003; Pedersen et al., 2010; Vitacco, Rogers, Neumann, Harrison & violence, meta-analyses report the predictive validity of the PCL
Vincent, 2005). Both the three and four-factor models showed scales to be moderate in regards to violence and similar outcomes.
acceptable fit, however, the three-factor model was found to be Empirically it is quite clear that psychopathic traits increase the
within the good fit range and the four-factor model to be within the risk of violence, especially traits like impulsivity, irresponsibility
fair fit range only. This finding was in line with AIC statistics finding and antisocial behavior (PCL scales factors 3 and 4). Still, research
the three-factor model to be superior to the four-factor model. is indicating that purposed-designed measure to assess violence
With the use of SEM analyses we further tested the four-factor risk is superior to the PCL scales, for example due to focus on
model. If Factor 4/Antisocial was a factor of the underlying specific violence types as well as management and treatment
construct we would expect no direct effect of exogenous variables targets. However, this does not mean excluding measures of
on Factor 4, but only an indirect effect mediated through the psychopathy. When applying risk assessment measures in clinical
underlying construct. The likelihood ratio tests reported in practice psychopathy does have a central role. A key part of a
Table 3 showed direct effects of age and education on Factor 4/ violence risk assessment is to understand the reasons why an
Antisocial. Similar analyses for the three other factors showed no individual may be violent. Psychopathic traits, alongside symptoms
evidence of direct effects. The direct effects constitute evidence of other mental disorders are vital in the understanding of an

© 2019 Scandinavian Psychological Associations and John Wiley & Sons Ltd
454 S. Bo et al. Scand J Psychol 60 (2019)

individual’s potential violent behavior. Such a case Andersen, H. S., Sestoft, D., Lillebaek, T., Mortensen, E. L. & Kramp, P.
conceptualization primarily requires an understanding of traits (1999). Psychopathy and psychopathological profiles in prisoners on
remand. Acta Psychiatrica Scandinavica, 99, 33–39.
related to the personality constellation and not a calculation of
Anderson, D. R. (2008). Model based inference in the life sciences : A
prior antisocial behavior. Furthermore, as part of violence risk primer on evidence. New York: Springer.
assessment psychopathic traits may play a vital role in informing Bateman, A. & Fonagy, P. (2009). Randomized controlled trial of
management plans and treatment strategies in order to reduce outpatient mentalization-based treatment versus structured clinical
future violence. In clinical practice this must be done regardless management for borderline personality disorder. American journal of
Psychiatry, 166(12), 1355–1364.
that only little research exists as to whether psychopathy is
Bengtson, S. & Pedersen, L. (2008). Implementation of evidence-based
treatable. In the research literature, there is evidence that treatment practices in forensic psychiatric clinical practice in Denmark: Are we
reduces violent behavior among individuals with psychopathy there? Scandinavian Journal of Forensic Science, 2, 48–53.
(Caldwell, McCormick, Umstead & Van Rybroek, 2007; Caldwell, Blackburn, R. (2007). Personality disorder and psychopathy: Conceptual
Skeem, Salekin & Van Rybroek, 2006; Skeem, Monahan & and empirical integration. Psychology, Crime & Law, 13, 7–18.
Boduszekab, D. & AgataDebowskac. (2016). Critical evaluation of
Mulvey, 2002) challenging the notion of “untreatable.” However,
psychopathy measurement (PCL-R and SRP-III/SF) and
research has yet to examine to what extent core traits of recommendations for future research. Journal of Criminal Justice, 44,
psychopathy are changeable. In relation to treatment of other PDs 1–12.
evidence indicate that core traits of for example borderline PD such Bolt, D. A., Hare, R. D., Vitale, J. E. & Newman, J. P. (2004). A
as impulsivity are changeable (Bateman & Fonagy 2009). multigroup item response theory analysis of the Psychopathy
Checklist-Revised. Psychological Assessment, 16, 155–168.
There are certain limitations to the present study. Primarily the
Burnham, K. P. (2004). Multimodel inference: Understanding AIC and
retrospective design in regards to some of the psychopathy BIC in model selection. Sociological Methods & Research, 33, 261–
assessments may have affected the rating of the measures. 304.
However, patients with incomplete file material were excluded Burnham, K. P. & Anderson, D. R. (2002). Model selection and
from the study. Hence, the included patients were all judged to multimodel inference: A practical information-theoretic approach.
New York: Springer.
have sufficient information. Furthermore, a whole group of
Caldwell, M., McCormick, D., Umstead, D., & Van Rybroek, G. J.
clinicians took part in the assessment of psychopathy. However, (2007). Evidence of treatment progress and therapeutic outcomes
training and group exercises were prioritized throughout the data among adolescents with psychopathic features. Criminal Justice and
collection and inter-rater reliability was high (Pedersen & Behavior, 34, 573–587.
Rasmussen, 2006; Pedersen et al., 2010). Furthermore, the sample Caldwell, M., Skeem, J., Salekin, R. & Van Rybroek, G. (2006).
Treatment response of adolescent offenders with psychopathy features:
size was smaller than generally considered desirable. However, the
A 2-year follow-up. Criminal Justice and Behavior, 33, 571–596.
stable solutions for the IRT models, suggest that the low sample Comprehensive Assessment of Psychopathic Personality (CAPP). (2008).
size apparently had no substantial adverse impact on the quality of [Website] Retrieved from http://www.gcu.ac.uk/capp/index.html.
the data analyzed. Also, when trying to disentangle the nature of Cooke, D. J. (1995). Psychopathic disturbance in the Scottish prison
psychopathy, we have to be aware that the majority of studies population: The cross-cultural generalisability of the Hare Psychopathy
Checklist. Psychology, Crime & Law, 2, 101–118.
investigating the construct have been overly targeting a range of
Cooke, D. J. & Michie, C. (1997). An item response theory analysis of the
forensic populations inherently focusing on antisocial behavioral Hare Psychopathy Checklist - Revised. Psychological Assessment, 9,
aspects. Hence, in an effort to “carve psychopathy at its joints” we 3–14.
need to acknowledge that we may only be approaching one side of Cooke, D. J. & Michie, C. (2001). Refining the construct of psychopathy:
the phenomenon. Finally, we applied the PCL:SV in a Danish Towards a hierarchical model. Psychological Assessment, 13, 171–
188.
sample and there could potentially be cross-cultural issues biasing
Cooke, D. J., Michie, C., Hart, S. D. & Hare, R. D. (1999). Evaluating the
the results. In a recent study there has been reported cross-cultural screening version of the An item response theory analysis of the Hare
differences in the phenotypic structure of psychopathy (Verschuere Psychopathy Checklist - Revised (PCL : SV): An item response theory
et al., 2018) between a north-American and a Dutch sample. Thus, analysis. Psychological Assessment, 11, 3–13.
it is important for future research to explore if cross-cultural Cooke, D. J., Michie, C. & Skeem, J. (2007). Understanding the structure
of the Psychopathy Checklist - Revised: An exploration of
differences exists, or if differences found in the structure of
methodological confusion. British Journal of Psychiatry, 190(Suppl.
psychopathy is due to measurement variance in the PCL, or both. 49), 39–50.
Cooke, D. J. & Sellbom, M. (2018). An examination of Psychopathy
We deeply thank all the patients who participated in this study and the Checklist-Revised latent factor structure via exploratory structural
various staff members, psychiatrists and psychologists who helped equation modeling. Psychological Assessment, 31, 581–591.
organizing the data collection. The authors have declared that there are no Det Centrale Kriminalregister. (2009). Det Centrale Kriminalregister[The
conflicts of interest in relation to the subject of this study. National Crime Register] [online]. Retrieved 15 May 2019 from http://
www.politi.dk/da/borgerservice/straffeattest/kriminalregisteret/
Edens, J. F., Campbell, J. S. & Weir, J. M. (2007). Youth psychopathy
and criminal recidivism: A meta-analysis of the Psychopathy Checklist
REFERENCES
measures. Law and Human Behavior, 31, 53–75.
Acheson, S. (2005). Review of the Hare Psychopathy Checklist-Revised. Fischer, G. H. & Molenaar, I. W. (1995). Rasch models – foundations,
In S. Robert & P. Barbara (Eds.), The sixteenth Mental Measurements recent developments, and applications. Berlin: Springer-Verlag.
Yearbook (2nd edn., pp. 429–431). Lincoln, Ne: University of Floyd, F. J. & Widaman, K. F. (1995). Factor analysis in the development
Nebraska Press. and refinement of clinical assessment instruments. Psychological
American Psychiatric Association (2013). Diagnostic and statistical Assessment, 7, 286–299.
manual and mental disorders: DSM-5. Washington, DC: American Forth, A. E., Kosson, D. & Hare, R. D. (2003). The Hare PCL: Youth
Psychiatric Association. Version. Toronto: Multi-Health Systems.

© 2019 Scandinavian Psychological Associations and John Wiley & Sons Ltd
Scand J Psychol 60 (2019) PCL:SV in a Danish forensic sample 455

Guy, L. S. & Douglas, K. S. (2006). Examining the utility of the PCL:SV Psychiatric Setting. International Journal of Forensic Mental Health,
as a screening measure using competing factor models of psychopathy. 9, 308–315.
Psychological Assessment, 18, 225–230. Pedersen, L. & Rasmussen, K. (2006). Reliability of the Danish version of
Guy, L. S., Douglas, K. S. & Hendry, M. C. (2010). The role of the HCR-20 risk assessment scheme. Scandinavian Journal of
psychopathic personality disorder in violence risk assessments using Forensic Science, 2, 45–96.
the HCR-20. Journal of Personality Disorders, 24, 551–580. Pedersen, L. & Simonsen, E. (2011). The role of psychopathy in assessing
Guy, L. S., Edens, J. F., Anthony, C. & Douglas, K. S. (2005). Does risk of violence - predictive and incremental validity. Melbourne:
psychopathy predict institutional misconduct among adults? A meta- ISSPD.
analytic investigation. Journal of Consulting and Clinical Psychology, Preszler, J., Marcus, D. K., Edens, J. F. & McDermott, B. E. (2018).
73, 1056–1064. Network analysis of psychopathy in forensic patients. Journal of
Hare, R. (1980). A research scale for the assessment of psychopathy in Abnormal Psychology, 127, 171–182.
criminal populations. Personality and Individual Differences, 1, 111– Salekin, R., Rogers, R. & Sewell, K. W. (1996). A review and meta-
119. analysis of the Psychopathy Checklist-Revised: Predictive validity of
Hare, R. D. (2003). The Hare Psychopathy Checklist - Revised (2nd edn.) dangerousness. Clinical Psychology: Science and Practice, 3, 203–215.
(Hogrefe, Trans.). Toronto: Multi-Health Systems. Singh, J., Grann, M. & Fazel, S. (2011). A comparative study of violence
Hare, R. & Neumann, C. (2008). Psychopathy as a clinical and empirical risk assessment tools: A systematic review and metaregression analysis
construct. Annual Review of Clinical Psychology, 4, 217–246. of 68 studies involving 25,980 participants. Clinical Psychology
Hare, R. D., Neumann, C. S. & Mokros, A. (2018). The PCL-R Review, 31, 499–513.
assessment of psychopathy: Development, properties, debates and new Skeem, J. L. & Cooke, D. J. (2010a). Is criminal behavior a central
directions. In C. J. Patrick (Ed.), Handbook of psychopathic component of psychopathy? Conceptual directions for resolving the
personality disorder (pp. 39–79). New York: Guilford Press. debate. Psychological Assessment, 22, 433–445.
Harpur, T. J., Hakstian, A. R. & Hare, R. D. (1988). Factor structure of Skeem, J. L. & Cooke, D. J. (2010b). One measure does not a construct
the Psychopathy Checklist. Journal of Consulting and Clinical make: Directions toward reinvigorating psychopathy research – reply
Psychology, 56, 741–747. to Hare and Neumann (2010). Psychological Assessment, 22, 455–459.
Hart, S. D., Cox, D. N. & Hare, R. D. (1995). The Hare Psychopathy Skeem, J. L., Monahan, J. & Mulvey, E. (2002). Psychopathy, treatment
Checklist: Screening Version. Toronto: Multi-Health Systems. involvement, and subsequent violence among civil psychiatric patients.
Hemphill, J. F., Hare, R. D. & Wong, S. (1998). Psychopathy and Law and Human Behavior, 26, 577–603.
recidivism: A review. Legal and Criminological Psychology, 3, 139– Skeem, J. L., Mulvey, E. & Grisso, T. (2003). Applicability of traditional
170. and revised models of psychopathy to the psychopathy checklist:
Higgs, T., Tully, R. & Browne, K. D. (2018). Psychometric properties in Screening Version. Psychological Assessment, 15, 41–55.
forensic application of the screening version of the Psychopathy Storey, J. E., Hart, S. D., Cooke, D. J. & Michie, C. (2016). Psychometric
Checklist. International Journal of Offender Therapy and Comparative properties of the Hare Psychopathy Checklist-Revised (PCL-R) in a
Criminology, 62, 1869–1887. representative sample of Canadian federal offenders. Law and Human
Kim, S. & Hagtvet, K. A. (2003). The impact of misspecified item Behavior, 40, 136–146.
parceling on representing latent variables in covariance structure Thissen, D. & Steinberg, L. (1986). A taxonomy of item response models.
modeling: A simulation study. Structural Equation Modeling, 10, 101– Psychometrika, 51, 567–577.
127. Verschuere, B., van Ghesel Grothe, S., Waldorp, L., Watts, A. L.,
Kreiner, S. (2011). A note on item–restscore association in Rasch models. Lilienfeld, S. O., Edens, J. F. et al. (2018). What features of
Applied Psychological Measurement, 35, 557–561. psychopathy might be central? A network analysis of the Psychopathy
Leistico, A. M. R., Salekin, R. T., DeCoster, J. & Rogers, R. (2008). A Checklist-Revised (PCL-R) in three large samples. Journal of
large-scale meta-analysis relating the hare measures of psychopathy to Abnormal Psychology, 127, 51–65.
antisocial conduct. Law and Human Behavior, 32, 28–45. Vitacco, M., Rogers, R., Neumann, C., Harrison, K. & Vincent, G. (2005).
Masters, G. N. (1982). A Rasch model for partial credit scoring. A comparison of factor models on the PCL-R with mentally
Psychometrika, 47, 149–174. disordered offenders - The development of a four-factor model.
Morrissey, C., Cooke, D., Michie, C., Hollin, C., Hogue, T., Lindsay, W. Criminal Justice and Behavior, 32, 526–545.
R. & Taylor, J. L. (2010). Structural, item, and test generalizability of Wilson, M. J., Abramowitz, C., Vasilev, G., Bozgunov, K. & Vassileva, J.
the psychopathy checklist–revised to offenders with intellectual (2014). Psychopathy in Bulgaria: The cross-cultural generalizability of
disabilities. Assessment, 17, 16–29. the Hare Psychopathy Checklist. Journal of Psychopathological and
Muraki, E. (1992). A Generalized Partial Credit Model: Application of an Behavioral Assessment, 36, 389–400.
EM algorithm. Applied Psychological Measurement, 16, 159–176. Yang, M., Wong, S. & Coid, J. (2010). The efficacy of violence
Neumann, C. S., Hare, R. D. & Pardini, D. A. (2015). Antisociality and prediction: A meta-analytic comparison of nine risk assessment tools.
the construct of psychopathy: Data from across the globe. Journal of Psychological Bulletin, 136, 740–767.
Personality, 83, 678–692. 
Zukauskien_  esnien_e, I. (2010). Testing factorial
e, R., Laurinavicius, A. & C_
Pedersen, L., Kunz, C., Rasmussen, K. & Elsass, P. (2010). Psychopathy structure and validity of the PCL:SV in Lithuanian prison population.
as a Risk Factor for Violent Recidivism: Investigating the Psychopathy Journal of Psychopathology and Behavioral Assessment, 32, 363–372.
Checklist Screening Version (PCL: SV) and the Comprehensive
Assessment of Psychopathic Personality (CAPP) in a Forensic Received 13 February 2019, accepted 13 May 2019

© 2019 Scandinavian Psychological Associations and John Wiley & Sons Ltd
Copyright of Scandinavian Journal of Psychology is the property of Wiley-Blackwell and its
content may not be copied or emailed to multiple sites or posted to a listserv without the
copyright holder's express written permission. However, users may print, download, or email
articles for individual use.

You might also like