You are on page 1of 16

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/271772967

Variants of Psychopathy in Adult Male Offenders: A Latent Profile Analysis.

Article  in  Journal of Abnormal Psychology · February 2015


DOI: 10.1037/abn0000042 · Source: PubMed

CITATIONS READS
69 1,952

6 authors, including:

Andreas Mokros Robert Hare


FernUniversität in Hagen University of British Columbia - Vancouver
111 PUBLICATIONS   1,962 CITATIONS    228 PUBLICATIONS   30,217 CITATIONS   

SEE PROFILE SEE PROFILE

Craig S Neumann Pekka Santtila


University of North Texas Åbo Akademi University
155 PUBLICATIONS   8,191 CITATIONS    246 PUBLICATIONS   4,802 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Age Preferences View project

Incest aversion, Inbreeding avoidance, and Intra-familial sexual abuse View project

All content following this page was uploaded by Pekka Santtila on 06 January 2016.

The user has requested enhancement of the downloaded file.


Journal of Abnormal Psychology © 2015 American Psychological Association
2015, Vol. 124, No. 2, 372–386 0021-843X/15/$12.00 http://dx.doi.org/10.1037/abn0000042

Variants of Psychopathy in Adult Male Offenders:


A Latent Profile Analysis

Andreas Mokros Robert D. Hare


University Hospital of Psychiatry Zurich, Zurich, Switzerland University of British Columbia and Darkstone Research Group,
Vancouver, Canada

Craig S. Neumann Pekka Santtila


University of North Texas Abo Akademi University
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Elmar Habermeyer Joachim Nitschke


University Hospital of Psychiatry Zurich, Zurich, Switzerland Ansbach District Hospital, Ansbach, Germany

As measured with the Psychopathy Checklist⫺Revised (PCL-R), psychopathy is a dimensional construct


underpinned by 4 correlated factors: Interpersonal, Affective, Lifestyle, and Antisocial. Theorists and
clinicians (e.g., Karpman and Arieti) have posited 2 variants of “primary” or “true” psychopathy, both
distinct from so-called “secondary” or “pseudopsychopathy.” We used latent profile analysis to deter-
mine if homogeneous classes exist within a sample of 1,451 male offenders with high PCL-R scores (ⱖ
27). The 4 PCL-R factors were the dependent variables for clustering. A solution with 3 latent classes
showed a better fit to the data than did a unitary model without latent classes. Tentative labels for the
latent classes are Manipulative (Latent Class 1 [LC1]), Aggressive (Latent Class 2 [LC2]), and
Sociopathic (Latent Class 3 [LC3]). The latter class represented an antisocial group that lacked the
emotional detachment observed in the other 2 groups. We propose that LC1 and LC2 reflect phenotypic
variations on a theme of the traditional construct of psychopathy, and that LC3 is consistent with
conceptions of antisocial personality disorder and sociopathy. Replication and external classification with
an independent data set of 497 adult male offenders again yielded clearly separable clusters, as well as
meaningful differences or trends among latent classes on education, intelligence, symptoms of antisocial
personality disorder, and self-reported psychopathic traits and negative affect. The conceptualization of
psychopathy in terms of manipulative and aggressive variants is consistent with clinical theory and is
empirically grounded.

Keywords: psychopathy, manipulative psychopathy, aggressive psychopathy, subtype, sociopathy

Supplemental materials: http://dx.doi.org/10.1037/abn0000042.supp

The current concept of psychopathy combines aspects that com- psychopathic features. The PCL-R was built, in part, on earlier
prise (a) narcissistic grandiosity and the ability to manipulate work by Hervey Cleckley (1964, 1976) who had described a set of
others; (b) a lack of emotional resonance and sincerity; (c) a 16 features that he regarded as characteristic of psychopathic
reckless and impulsive lifestyle; and (d) a tendency to ignore or personalities (see Hare & Neumann, 2008, for a critical analysis of
violate social conventions and mores (Hare & Neumann, 2006). Cleckley’s work). Because the PCL-R factors are integral to the
These four foci represent the first-order factors underlying the analyses conducted in the present study, they and the items that
Psychopathy Checklist⫺Revised (PCL-R; Hare, 2003), a 20-item comprise them are listed here: Interpersonal (i.e., glibness/super-
rating scale that uses file and interview information to assess ficial charm, grandiose sense of self-worth, pathological lying,

This article was published Online First February 2, 2015. Robert D. Hare receives royalties from the sale of the PCL-R and its
Andreas Mokros, Department for Forensic Psychiatry, University Hospital of derivatives. We thank Joseph P. Newman, University of Wisconsin⫺Madi-
Psychiatry Zurich, Zurich, Switzerland; Robert D. Hare, Department of Psychol- son, for kindly providing the data used in the external classification
ogy, University of British Columbia, and Darkstone Research Group, Vancouver, analysis. We thank Kylie Neufeld for her editorial assistance with this
Canada; Craig S. Neumann, Department of Psychology, University of North article.
Texas; Pekka Santtila, Department of Psychology and Logopedics, Abo Akademi Correspondence concerning this article should be addressed to Andreas
University; Elmar Habermeyer, Department for Forensic Psychiatry, University Mokros, University Hospital of Psychiatry Zurich, Department for Foren-
Hospital of Psychiatry Zurich; Joachim Nitschke, Department of Forensic Psychi- sic Psychiatry, Lenggstrasse 31, P.O. Box, 1931, 8032 Zurich, Switzerland.
atry, Ansbach District Hospital, Ansbach, Germany. E-mail: andreas.mokros@puk.zh.ch

372
VARIANTS OF PSYCHOPATHY 373

cunning/manipulative); Affective (i.e., lack of remorse or guilt, Some Current Subtype Theories
shallow affect, callous/lack of empathy, failure to accept respon-
Since the writings of Karpman, Arieti, Cleckley, and their
sibility for own actions); Lifestyle (i.e., need for stimulation/prone-
contemporaries, theorists and researchers have pursued efforts to
ness to boredom, parasitic lifestyle, lack of realistic long-term
parse the psychopathy construct. In many cases, the results bear a
goals, impulsivity, irresponsibility); and Antisocial (i.e., poor be-
strong resemblance to the idea of primary and secondary (or rather,
havioral controls, early behavior problems, juvenile delinquency,
sociopathic) subtypes. Although many investigators view psy-
revocation of conditional release, criminal versatility). Two other
chopathy as pathology (i.e., abnormality, deficit), evolutionary
items (promiscuous sexual behavior, many short-term relation-
psychologists take the view that psychopathy is an evolved, adap-
ships) do not load on any factor but do contribute to the total
tive life-strategy for reproductive and survival success (Book &
PCL-R score. The Interpersonal/Affective dimensions and the
Quinsey, 2004; Harpending & Sobus, 1987; Harris, Rice, Hilton,
Lifestyle/Antisocial dimensions comprise, respectively, the Lalumiére, & Quinsey, 2007; Mealey, 1995a, 1995b; Ward &
second-order PCL-R Factors 1 and 2 described by Hare (2003; also Durrant, 2011; also see Buss, 2009, for a discussion of personality
see Hare & Neumann, 2008).
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

traits as forms of strategic individual differences in adaptation).


This document is copyrighted by the American Psychological Association or one of its allied publishers.

Book and Quinsey (2004) have suggested that psychopaths


achieve their ends through the use of cheater (e.g., manipulation,
Clinical Background
deception, selfishness, callousness) and warrior-hawk strategies
As early as the 1920s Karpman (e.g., Karpman, 1929) began to (e.g., impulsivity, aggression, violence). Presumably, psychopaths
describe a clinical disorder very similar to that later outlined by differ among themselves (and across context and time) in the
Cleckley (1941). Like Cleckley, Karpman was concerned that relative use of manipulative or aggressive strategies.
many different conditions were lumped under the umbrella of Along these lines, and drawing on extensive research from
psychopathic personality, many characterized by antisocial behav- behavior genetics, psychology, sociobiology, and game theory,
ior. He took pains to differentiate “primary, essential, or idiopathic Mealey (1995b) proposed an evolutionary model to account for the
psychopathy” from “symptomatic or secondary psychopathy,” on similarities and differences between primary and secondary psy-
the grounds that the underlying etiology, motivations, dynamics, chopathy (she originally used the term sociopathy; Mealey,
and response to treatment of the former are quite unlike those of 1995a). She stated: “My basic premise is that sociopaths are
the latter. Karpman described two forms of primary or idiopathic designed for the successful execution of social deception and that
psychopathy that shared similar motivations and dynamics but they are the product of evolutionary pressures which, through a
differed in their interactions with others: aggressive/predatory and complex interaction of environmental and genetic factors, lead
passive/parasitic (Karpman, 1955, p. 46). some individuals to pursue a life-history strategy of manipulative
Karpman’s distinctions between primary and secondary psy- and predatory social interactions” (Mealey, 1995b, p. 524). For
chopathy, and between variants of the former, were shared by primary psychopaths, the etiological pathway to this life-history
several other clinicians (see Hervé, 2007, for a review). Like strategy is heavily influenced by genetically based biological,
Karpman, Arieti (1963) also described several kinds of the “true” personality, and behavioral dispositions, whereas for secondary
psychopathy, the pathway is influenced more by early adverse
psychopath that differed from one another in their interpersonal
social and environmental conditions. This latter pathway captures
and aggressive behaviors. They included the simple psychopath
the meaning of earlier conceptions of sociopathy or pseudopsy-
and the complex psychopath, the latter described by Arieti (1963,
chopathy (e.g., Arieti, 1963; Lykken, 1995; McCord & McCord,
pp. 307–308) as having the mental set, wit, and Machiavellian
1964), but adds dimensions derived from models of evolutionary
orientation needed to transform the simple psychopath’s “how to
psychology. For example, primary psychopaths should have highly
do it” into “how to do it and get away with it.” Arieti (p. 308) also
heritable psychopathic traits, be predominantly male, be affection-
referred to dyssocial individuals who are not socialized in the usual
less and emotionally hypoaroused, stem from all socioeconomic
sense, are antisocial with respect to society at large, but loyal to
backgrounds, exhibit behavioral propensities that are persistent
members of their own group. Another term for such individuals throughout the life span, excel in social exchanges that reward
would be sociopath, as described by Lykken (1995). successful defection more than cooperation, and be relatively
It is clear that Karpman, Arieti, and other early influential uncommon. In contrast, the heritability of most of the traits of
clinicians did not consider secondary or symptomatic types to be secondary psychopathy is presumed be low. Furthermore, Mealey
part of the prototypical psychopathy construct described so vividly (1995b) suggested that secondary psychopaths should have few
in their writings. A common view was that psychopathy is rooted heritable psychopathic traits, come predominantly from lower-
in genetic predispositions and social/environmental forces that are class backgrounds, and exhibit patterns of antisocial conduct that
quite different from those that lead to secondary psychopathy. In vary across the life span. Finally, secondary psychopaths should be
this sense, diagnostic labels, such as secondary or symptomatic found among both men and women, and their prevalence should
psychopathy, are problematic and misleading because they imply fluctuate according to societal conditions across time or culture. In
that individuals so labeled are psychopaths in the traditional sense many respects, her suggestions concerning primary psychopathy
of the term (Hare, 1970). Pseudopsychopath (Arieti, 1963; Mc- are consistent with the current empirical literature on psychopathy
Cord & McCord, 1964) or sociopath may be more appropriate (e.g., Hare, 2003; Hare, Neumann, & Widiger, 2012; Hervé &
descriptors for such individuals. It also would facilitate theory and Yuille, 2007; Kiehl & Sinnott-Armstrong, 2013; Millon, Krueger,
research on the real issue: Do different manifestations of psychop- & Simonsen, 2010; Patrick, 2006; Widiger, 2012). They also are
athy reflect meaningful subtypes or are they variations on a theme? consistent with the theoretical outline by Yildirim and Derksen
374 MOKROS ET AL.

(2013) on the emotional, behavioral, and neurophysiological dif- self-reported anxiety; the differences included the PCL-R Inter-
ferences between psychopathy and sociopathy. personal, Affective, and Lifestyle factors. That is, the “variants”
did differ in their pattern of PCL-R traits, with those in the
secondary psychopathy cluster having lower scores than those in
Empirical Studies
the primary psychopathy cluster on the very elements (i.e., Inter-
Since the writings of Karpman, Arieti, Cleckley, and their personal, Affective, Lifestyle) considered by Skeem and col-
contemporaries, the notion of psychopathic subgroups has given leagues (e.g., Skeem & Cooke, 2010) to be at the core of the
rise to a considerable amount of research. Many of the earlier psychopathy construct.
studies relied on conventional techniques of cluster analysis, Hicks et al. (2004) obtained two clusters from a model-based
known to be prone to overidentifying subgroups (Beauchaine, analysis of scores on the 11 scales of the self-report Multidimen-
2003). Some more recent studies (e.g., Claes et al., 2014; Drislane sional Personality Questionnaire⫺Brief Form (MPQ-BF; Patrick,
et al., 2014; Falkenbach, Poythress, & Creevy, 2008; Falkenbach, Curtin, & Tellegen, 2002). The scores were obtained from 96 male
Stern, & Creevy, 2014; Hicks, Markon, Patrick, Krueger, & New- offenders, each with a PCL-R score of at least 30. One cluster was
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

man, 2004; Hicks, Vaidyanathan, & Patrick, 2010; Lee & Salekin, delineated by MPQ-BF scores indicative of low stress reaction,
This document is copyrighted by the American Psychological Association or one of its allied publishers.

2010) have used more rigorous methods—such as latent class social dominance, lack of close attachments, capacity for strategic
analysis (Vermunt & Magidson, 2006) or model-based cluster action, and proneness to taking risks. Hicks et al. (2004) labeled
analysis (Fraley & Raftery, 2002)—that allow the relative fit of individuals in this cluster (31% of the sample) as emotionally
competing models to be compared. These studies differed widely stable psychopaths, and noted that their features were consistent
in their conceptual orientation and definition of psychopathy (e.g., with conceptions of the primary psychopath or “classic” psycho-
clinical ratings, self-reports), involved a diversity of participants path. The second cluster (69% of the sample) was delineated by
(e.g., offender, community, and student samples), and used differ- MPQ-BF scores indicative of negative emotionality, and aggres-
ent methods of measurement and analysis (see Blackburn, 2009; sive, externalizing, and disinhibited behavior. The authors labeled
Hervé, 2007; Poythress & Skeem, 2006, for reviews). individuals in this cluster as aggressive psychopaths.
Poythress et al. (2010) performed a model-based cluster analysis The authors of the three studies discussed in detail above (Hicks
of the data from an entire sample of 691 adult male offenders, all et al., 2004; Poythress et al., 2010; Skeem et al., 2007) interpreted
of whom had been diagnosed with antisocial personality disorder their results in terms of Karpman’s (1941, 1948) primary⫺second-
(ASPD). The data subjected to clustering consisted of scores on ary distinction. Furthermore, the results of one of these studies
three (of the four) PCL-R factors (i.e., Interpersonal, Affective, (Hicks et al., 2004) can be understood in terms of the views of
and Lifestyle), along with other measures such as self-reports of Karpman (1955) and Arieti (1963) that there are two variants of
harm avoidance, reward responsivity, and experiences of child- primary psychopathy. Therefore, we should be able to identify
hood trauma. They obtained five clusters, including two clusters individuals with high levels of psychopathic and/or antisocial traits
labeled as “primary psychopath” (20.4% of the sample) and “Karp- as psychopathic and sociopathic types. It also should be possible to
man’s secondary psychopath” (22.1% of the sample), respectively. distinguish passive/parasitic/complex psychopaths from aggres-
Individuals in the primary psychopath cluster had relatively high sive/predatory/simple psychopaths.
mean scores on the Interpersonal and Affective factors, and rela- The studies discussed above derived their subtypes from a
tively low mean scores on the Lifestyle factor and self-reported combination of PCL-R factors and various self-report scales, or
anxiety. Those assigned to the secondary psychopath cluster had from self-report scales alone. An alternative approach would be to
relatively high mean scores on self-report scales measuring anxi- search for classes defined by the traits or dimensions that comprise
ety, antisociality, and childhood trauma. These two clusters did not the psychopathy construct. Analyses based on individual traits
differ on the PCL-R Antisocial factor, a variable that was not would be far more complex and difficult to interpret than analyses
included in the cluster analysis. based on dimensions of psychopathy. The logical candidates for
Skeem, Johansson, Andershed, Kerr, and Louden (2007) per- such analyses are the four dimensions or factors that underpin the
formed a model-based cluster analysis on a sample of 123 Swedish PCL-R. Hervé (2007) described some preliminary cluster analyses
violent male offenders whose PCL-R scores were in the top one of the factor scores of 202 male offenders with a PCL-R score of
third of the distribution of scores in a group of 367 offenders. Each at least 27.
offender in the clustering sample had a PCL-R score of 29 or
higher, close to the commonly used threshold for psychopathy (ⱖ
Current Study
30). The data subjected to clustering included scores on all four
PCL-R factors and a self-report measure of trait anxiety. Two The sample in the current study consisted of a large number of
clusters emerged, which the authors interpreted as consistent with adult male offenders with high trait levels of psychopathy and
Karpman’s (1941) conception of primary and secondary psycho- antisociality. On the basis of clinical theory and previous research,
pathic types. The primary subtype (60% of the sample) had a we expected that latent class analysis of the four PCL-R factors
higher mean PCL-R total score and higher mean scores on PCL-R would yield two variants of psychopathy and a separate socio-
Factors 1⫺3 (i.e., Interpersonal, Affective, and Lifestyle), about pathic subgroup. We expected that the psychopathy variants would
the same score on Factor 4 (Antisocial), and a lower mean trait be consistent with clinical and empirical descriptions of manipu-
anxiety score, than did the secondary subtype (40% of the sample). lative (passive/parasitic/complex) and aggressive (predatory/sim-
Skeem et al. (2007) stated that “. . .we found no differences ple) psychopathy. We also expected that these variants would be
between variants in their patterns of psychopathic traits” (p. 404). distinguishable from sociopathy (pseudopsychopathy/dyssociality/
But, the differences between the “variants” were not confined to “secondary psychopathy”). The sociopathic cluster should be
VARIANTS OF PSYCHOPATHY 375

mostly antisocial and would share many psychopathic features, but A Mardia (1974) test showed that the subtotals on the four
with less extreme emotional components (e.g., lack of empathy, PCL-R factors were not distributed as multivariate normal due to
guilt or remorse, shallow emotions) so characteristic of the psy- skewness (␹2(20) ⫽ 498.48, p ⬍ .001) even though the distribution
chopathy construct. In keeping with the method chosen by Hervé was mesokurtic (z ⫽ ⫺1.62, p ⫽ .11). Skewness is a result of
(2007), we included offenders with a PCL-R total score of 27 or truncating the data by only including slightly more than the upper
higher, which is 1 SEM unit below the conventional PCL-R thresh- quartile of the reference group.
old for psychopathy of 30. The rationale was to increase sensitivity Statistical analyses. LPA is a variant of latent class analysis
(i.e., not missing “psychopaths”) at the cost of decreasing speci- based on observed continuous rather than categorical variables.
ficity (i.e., possibly including “nonpsychopaths”), as argued by LPA serves to identify homogeneous subgroups within a sample
Hervé (2007). We expected that evidence for latent classes or through maximum likelihood (ML) estimation. Consequently,
variants would emerge as different patterns of PCL-R factor scores LPA is a probabilistic or model-based alternative to conven-
(profiles). tional cluster analysis (Vermunt & Magidson, 2006). As Ver-
Second, we aimed to provide a cross-validation of the results of munt and Magidson pointed out, LPA is basically equivalent to
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

the development sample in an independent sample by (a) replicat- model-based clustering. Mathematically, LPA as well as model-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

ing the structure of the latent profile analysis (LPA), and (b) based clustering are variants of finite-mixture modeling
validating the latent classes through an external classification with (McLachlan & Peel, 2000). As a mixture-distribution model,
respect to meaningful variables not included in the clustering LPA seeks to identify nominal variables that underlie the con-
procedure. tinuous data and whose identification allows “demixing” of the
data (Rost, 2006). It should be noted that individual cases
Experiment 1 commonly have probabilities of belonging to more than one
latent class. The more distinct these average latent class prob-
abilities for the most likely class membership are, the more
Method
useful will be a latent class solution. As Rost (2006) noted:
Sample. Participants were part of the pooled sample (referred “The average probability of group membership for all persons
to here as the reference group) of 5,408 North American adult tells something about the quality of the class allocation . . .. In
male offenders described in the PCL-R manual (Hare, 2003). The fact, the values in useful analyses are mostly above .80, indeed
PCL-R rating of each offender in the reference group was based on often even above .90” (p. 278; translation by first author).
the standard PCL-R assessment procedure, which involves use of Both information criteria (e.g., the Bayesian information crite-
a semistructured interview, case-history information, and specific rion [BIC]) as well as modified likelihood ratio tests (LRTs; Lo,
scoring criteria to rate each item on a 3-point scale (range: 0⫺2) Mendell, & Rubin, 2001; Nylund, Asparouhov, & Muthén, 2007)
according to the extent to which it applies to a given person. Total can be used to decide on the number of latent classes. For the BIC
scores can range from 0⫺40, with 30 typically used by researchers (Schwarz, 1978), a minimal value would indicate best model fit in
as a convenient threshold for psychopathy (see Hare, Black, & terms of the optimal trade-off between model parsimony and
Walsh, 2013; Forth, Bo, & Kongerslev, 2013). The mean PCL-R residuals. Reductions in BIC value of less than 3 are considered
score for the reference group was 22.1 (SD ⫽ 7.9). Reliability of negligible (Kass & Raftery, 1995). For the LRTs, a nonsignificant
the total scores (as indexed with Guttman’s ␭2 coefficient of outcome for a solution with k latent classes indicates that model fit
internal consistency) was .86. The PCL-R items can be accounted did not improve substantially compared with a more parsimonious
for in terms of a superordinate latent variable (Neumann, Hare, & (k ⫺ 1) model; hence, the simpler (k ⫺ 1) model should be chosen.
Newman, 2007), and this superordinate factor is underpinned by Nylund et al. (2007) has recommended using the BIC index as
four correlated first-order factors or dimensions, labeled as Inter- well as an LRT. Because the bootstrap LRT is more strongly
personal, Affective, Lifestyle, and Antisocial. Guttman’s ␭2 coef- affected by nonsymmetrical data distributions (Nylund et al.,
ficients for these four factors in the reference group were, respec- 2007) and often remains inconclusive (Kupzyk, 2011), we opted
tively, .73, .73, .74, and .70. Assigning an individual to a group for the combination of the LRT of Lo et al. (2001) and the BIC
profile based on the four factors would thus occur at an estimated index instead.
battery reliability of rP equal to .57 (using the Mahalanobis D as a Data were analyzed with the SPSS statistics program, Version
measure of similarity; cf. Conger & Lipshitz, 1973). 19.0.0 (IBM Corp., Somers, NY), and LPA was carried out with
Participants were 1,451 offenders (27% of the reference group) Mplus for Mac, Version 6.12 (Muthén & Muthén, 2011). To
with a PCL-R score of at least 27. The scores for this development prevent local maxima in the ML estimation procedure for the LPA,
sample ranged from 27⫺40 (M ⫽ 30.5, SD ⫽ 2.7). The mean several random starts were conducted in the ML procedure. We
score on each factor was as follows: Interpersonal, 5.4 (SD ⫽ 1.6); specified 600 random sets of starting values and 120 optimizations
Affective, 6.5 (SD ⫽ 1.2); Lifestyle, 7.9 (SD ⫽ 1.4); and Antiso- of the expectation maximization (EM) algorithm for the solutions
cial, 8.0 (SD ⫽ 1.7). Note that the scores for the Lifestyle/ with two to five latent classes. These values were raised to 5,000
Antisocial factors are higher than those for the Interpersonal/ and 1,000, respectively, for models with six or more latent classes.
Affective factors because the former are based on five items per Furthermore, the default settings for the maximum number of
factor (scores from 0⫺10) whereas the latter are based on four iterations within the initial stage (10) and the convergence criterion
items per factor (scores from 0⫺8). Also, scores on two PCL-R (0.1e-5) were used. In the bootstrap variant of the LRT (Nylund et
items (promiscuous sexual behavior, many short-term marital re- al., 2007), the number of random starts and of optimizations within
lationships) do not load on any factor but do contribute to the total the EM algorithm was set to 1,000 and 100, respectively, for the
PCL-R score. model with k classes. Scripts for the estimation of battery reliabil-
376 MOKROS ET AL.

ity (Conger & Lipshitz, 1973) and for Mardia’s (1974) test of For display purposes, we converted the subtotals per PCL-R factor
multivariate normality were implemented in Maple, Version 14.01 into z score deviations from the reference group. The z score means
(Waterloo Maple Inc., Waterloo, ON, Canada). The scripts are of the three latent classes are depicted in Figure 1a, with a z score
available from the first author on request. Effect sizes were ex- of 0 representing the mean of the reference group.
pressed as Cohen’s d, computed as the absolute difference between As Figure 1a shows, individuals with the highest allocation
means, divided by the pooled standard deviation. Values of d equal probability for LC1 had a very high mean score on the Interper-
to or larger than 0.2, 0.5, or 0.8 are considered indicative of small, sonal factor (indicative of deceitfulness, manipulativeness, and
medium, or large effects, respectively (Cohen, 1992). egocentricity). Those with the highest allocation probability for
LC2 had a very high mean score on the Antisocial factor (indic-
Results ative of a propensity for violating social norms). LC1 and LC2 did
not differ on the Affective and Lifestyle factors, and their mean
According to the LPA, solutions with latent classes fit the data
scores on these factors were much higher than those of the refer-
better than did a unitary solution without latent classes (see Table
ence group.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

1, top). The LRT of Lo et al. (2001) indicated that the fit did not
The individuals most likely allocated to LC3, in contrast, had a
This document is copyrighted by the American Psychological Association or one of its allied publishers.

improve significantly when increasing the number of latent classes


mean score on the Affective factor that was on par with the mean
to four (p ⫽ .272). The minimum BIC value was observed for a
solution with eight latent classes, however. In the sense of parsi- of the reference group, and much lower than the corresponding
mony, the three-latent-classes solution was chosen as the frame of means of the other two groups. They also had a high mean score
reference for further inspection. on the Lifestyle and Antisocial factors. That is, on average, the
For the solution with three latent classes, the average probabil- latter resembled the LC2 individuals in terms of antisociality and
ities for the most likely class membership were .88, .86, and .73 in indicators of a lifestyle that includes impulsivity, sensation
(see Table 2, top), with a relative entropy of .67. Hence, the seeking, and irresponsibility.
allocation probabilities were substantial with regard to Latent Table 3 provides the means (standard deviations) of the three
Classes 1 (LC1) and 2 (LC2) and somewhat lower for Latent Class latent classes on the four PCL-R factors as well as the corre-
3 (LC3). Cases with most likely latent class allocation toward LC1 sponding pairwise effect sizes between the clusters. LC1 had a
had the second-highest mean probability for LC2. Similarly, for higher mean score on the Interpersonal factor than did LC2 (by
cases with a most likely latent class allocation toward LC3, the a medium effect size; d ⫽ 0.58), whereas LC2 a higher mean on
second-highest mean probability concerned LC2. the Antisocial factor than did LC1 (by a very large effect size;
Assigning cases to clusters based on their most likely class d ⫽ 2.58).
membership, three groups of different size did emerge. These LC1 and LC2, pooled as a single group (LC1/LC2), had a
clusters contained 287, 948, and 216 individuals, respectively. We much higher average score than LC3 on the Affective factor
tentatively labeled these clusters as LC1 (n ⫽ 287; 19.8% of the (d ⫽ 2.79), a moderately higher average score on the Interper-
development sample), LC2 (n ⫽ 948; 65.3% of the development sonal factor (d ⫽ 0.43), and somewhat lower average scores on
sample), and LC3 (n ⫽ 216; 14.9% of the development sample). the Lifestyle (d ⫽ 0.35) and Antisocial (d ⫽ 0.39) factors.

Table 1
Model Fit of the Latent Profile Analyses for the Development Sample and the Replication Sample

(⫺2) ⴱ log-
No. of free Adjusted likelihood Bootstrap
No. of latent classes Log-likelihood parameters BIC BIC AIC differencea LMR LRTb pd LRTc pd 1 – entropy

Development sample (N ⫽ 1,451)


1 ⫺10,360.21 8 20,778.66 20,753.25 20,736.42 — — — —
2 ⫺10,247.86 13 20,590.36 20,549.06 20,521.72 224.70 ⬍.001 ⬍.001 0.730
3 ⫺10,196.10 18 20,523.25 20,466.07 20,428.21 103.52 0.012 ⬍.001 0.674
4 ⫺10,166.12 23 20,499.68 20,426.62 20,378.24 59.96 0.272 ⬍.001 0.727
5 ⫺10,138.44 28 20,480.72 20,391.77 20,332.88 55.36 0.433 ⬍.001 0.659
6 ⫺10,116.09 33 20,472.42 20,367.59 20,298.18 44.70 0.062 ⬍.001 0.744
7 ⫺10,086.25 38 20,449.14 20,328.43 20,248.50 59.68 0.277 ⬍.001 0.638
8 ⫺10,061.57 43 20,436.17 20,299.58 20,209.13 49.37 0.084 ⬍.001 0.650
9 ⫺10,044.39 48 20,438.22 20,285.74 20,184.78 34.35 0.259 ⬍.001 0.665
Replication sample (N ⫽ 497)
1 ⫺3,583.948 8 7,217.57 7,192.17 7,183.90 — — — —
2 ⫺3,525.559 13 7,131.83 7,090.57 7,077.12 116.78 <.001 ⬍.001 0.797
3 ⫺3,506.575 18 7,124.90 7,067.77 7,049.15 37.97 0.087 ⬍.001 0.789
4 ⫺3,495.563 23 7,133.92 7,060.92 7,037.13 22.02 0.013 ⬍.001 0.833
Note. Optimal models according to LMR LRT and BIC are highlighted in boldface. Other fit indices are reported for completeness. AIC ⫽ Akaike’s
information criterion; BIC ⫽ Bayesian information criterion; LMR ⫽ Lo–Mendell–Rubin; LRT ⫽ likelihood ratio test.
a
Difference between models of (k – 1) and k classes. b LRT according to Lo et al. (2001). c LRT according to Nylund et al. (2007). d If less than .05,
a model with k latent classes will fit significantly better than a model with (k – 1) latent classes. The relevant criteria for deciding on the number of latent
classes are LMR LRT and BIC (see text for details).
VARIANTS OF PSYCHOPATHY 377

Table 2 are manipulative and aggressive psychopathy, respectively. In


Cluster Allocation Based on Maximum Posterior Probability for each case, the theoretical implication is that we are dealing with
Three Latent Classes: Mean Probabilities of Latent Class variations on the theme of psychopathy rather than with clear
Membership in the Development Sample and in the subtypes, in-line with the LPA results.
Replication Sample The lowest average classification probability was observed for
LC3. The pattern observed for this latent class was similar to the
Latent class
pseudopsychopathic or sociopathic profile described by Hervé
Cluster allocation n % LC1 LC2 LC3 (2007). It also is consistent with Arieti’s (1963) description of
Development sample (N ⫽ 1,451) dyssocial individuals and, to a degree, with the diagnostic criteria
LC1 287 19.8 .86 .12 .03 for ASPD. Because it does not seem to reflect psychopathy, at least
LC2 948 65.3 .06 .88 .06 in the narrow sense of the term, and for convenience, we refer to
LC3 216 14.9 .04 .23 .73
LC3 as sociopathic. However, as Table 3 indicates, the boundary
Replication sample (N ⫽ 497)
LC1 96 19.3 .87 .11 .02 between LC3 and LC2 seems rather fuzzy. Some offenders might
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

LC2 364 73.2 .04 .93 .03 fit into either latent class, depending in large part on how high their
This document is copyrighted by the American Psychological Association or one of its allied publishers.

LC3 37 7.4 .05 .15 .80 score is on the Affective factor.


Note. LC1 ⫽ Latent Class 1; LC2 ⫽ Latent Class 2; LC3 ⫽ Latent Class 3. The mean PCL-R score of the offenders in LC3 was almost as
high as that of those in the other latent classes. But, the suggestion
that LC3 is a variant of psychopathy is at odds with the very low
Mean (standard deviation) PCL-R total scores were 29.2 (2.0), score on the Affective factor, and with the widely held view and
31.2 (2.8), and 28.9 (1.7) for LC1, LC2, and LC3, respectively.
empirical evidence that shallow emotions and lack of empathy,
Pairwise comparisons of the means yielded a very small effect size
guilt, and remorse are essential components of the psychopathy
for LC1 versus LC3 (d ⫽ 0.13), and large effect sizes for LC1
construct (e.g., Cleckley, 1976; Hare, 2003; Herpertz & Sass,
versus LC2 (d ⫽ 0.80) and for the LC2 versus LC3 (d ⫽ 0.89). The
2000). Item response theory analyses of the data from the reference
similarity of the average PCL-R total scores observed for LC1 and
group have indicated that the Affective factor of the PCL-R is
LC3 indicates that they may be difficult to tell apart without
information about their pattern (profile) of factor scores. more discriminating of the latent trait of psychopathy than are the
other factors (Bolt, Hare, Vitale, & Newman, 2004). Further, the
offenders in LC3 seem very much like those who meet the criteria
Discussion
for ASPD but who do not have the affective features that charac-
We used LPA to determine whether homogeneous classes with terize psychopathy (Hare et al., 2012). We might expect that such
relatively unique PCL-R four-factor profiles exist within a sample offenders would display more affect, and perhaps even symptoms
of male offenders with high PCL-R scores. A solution with three of affective disturbance, than would those who match the LC1 and
latent classes showed better model fit than did a single-group LC2 PCL-R factor profiles. Finally, the emergence of LC3 seems
version or a solution with two latent classes, attesting to a degree to be largely due to the choice of a lower-than-customary cut-score
of phenomenological heterogeneity of the psychopathy construct of 27 points on the PCL-R because a similar type did not emerge
as measured with the PCL-R. Model fit did not improve signifi- once the cut-score was raised to the customary 30 points.1
cantly when we increased the number of latent classes to four.
Hence, we focused on the three-latent-classes solution for inter-
pretation and further analysis. Experiment 2
The average classification probabilities when assigning individ-
uals to their most likely latent class were relatively high, particu- The sample used for the LPA in Experiment 1 (development
larly for LC1 and LC2. This suggests that the two latent classes sample) consisted of offenders with high PCL-R scores in a
represent separable variations of the psychopathic theme. This reference group made up of data sets provided by several research-
observation counters Eysenck’s (1995) criticism of Mealey’s ers (Hare, 2003). Although we had PCL-R item and factor scores,
(1995b) theory on the grounds that he expected low agreement for we did not have ready access to the concomitant data that the
any attempts to allocate individuals to such groups: “From expe- researchers may have gathered. For this reason, we were unable to
rience, I would be surprised if agreement between professional
observers would be greater than 0.3 or thereabouts; clearly 1
Several investigators have searched for latent classes in samples of
insufficient for scientific or practical purposes” (Eysenck, 1995, offenders selected on the basis of a PCL-R threshold higher than 27 (e.g.,
p. 556). 29 by Skeem et al., 2007; 30 by Hicks et al., 2004). For comparative
LC1 and LC2, respectively, seem in keeping with Karpman’s purposes, we performed a supplementary LPA on a sample of 856 offend-
(1955) passive/parasitic and aggressive/predatory psychopathy, ers with a PCL-R score of at least 30 (the upper 15% of the reference
group). A two-group solution provided sufficient fit to the data. The two
Arieti’s (1963) complex and simple psychopathy, Book and Quin- latent classes were virtually identical with LC1 (manipulative; 23.7% of the
sey’s (2004) cheater and warrior-hawk psychopathy, and Hervé’s sample) and LC2 (aggressive; 76.3% of the sample) described here. In
(2007) manipulative and aggressive psychopathy. Similarly, Hicks statistical terms, the three-class solution obtained with a PCL-R threshold
et al. (2004) referred to emotionally stable and aggressive psycho- of 27 was more robust than the two-class solution obtained in this supple-
mentary analysis. Also, use of the lower threshold likely ameliorated some
paths. Given the profiles in this study, it seems plausible to refer to of the issues associated with conducting LPA on more extreme portions of
both LC1 and LC2 as variants of (primary) psychopathy. We a distribution of scores (Bauer & Curran, 2004). Details of the 30⫹
propose that clinically meaningful descriptors for LC1 and LC2 analyses are provided in the supplemental materials for this article.
378 MOKROS ET AL.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 1. Mean z scores of each latent class on each PCL-R factor, referenced against the mean factor scores
of the reference group of 5,408 male offenders, depicted on the bottom line as a z score of 0. The latent classes
consist of offenders with a PCL-R score of 27 or higher. The reference group is from Hare (2003, p. 59).
Descriptive labels for the three latent classes are: manipulative psychopaths (Latent Class 1), aggressive
psychopaths (Latent Class 2), and sociopaths (Latent Class 3). Results for the development sample (N ⫽ 1,451)
(a). Results for the replication sample (N ⫽ 497) (b). Vertical bars indicate standard deviations. PCL-R ⫽
Psychopathy Checklist⫺Revised (Hare, 2003).
VARIANTS OF PSYCHOPATHY 379

Table 3
Means (Standard Deviations) and Effect Sizes (Cohen’s d) on Four Factors of the PCL-R in Three Latent Classes for the
Development Sample (N ⫽ 1,451) and the Replication Sample (N ⫽ 497)

Pairwise comparisons between latent classes


LC1 LC2 LC3 (Cohen’s d)
M (SD) M (SD) M (SD) 1 vs. 2 1 vs. 3 2 vs. 3

PCL-R factors n ⫽ 287 [96] n ⫽ 948 [364] n ⫽ 216 [37]


1. Interpersonal 6.2 (1.5) [6.0 (1.6)] 5.3 (1.5) [4.8 (1.6)] 4.8 (1.5) [5.6 (1.6)] 0.58 [0.73] 0.90 [0.22] 0.32 [0.51]
2. Affective 6.8 (0.9) [6.9 (1.0)] 6.8 (0.9) [6.7 (1.0)] 4.8 (0.9) [4.1 (1.0)] 0.06 [0.21] 2.17 [2.89] 2.22 [2.68]
3. Lifestyle 8.0 (1.4) [8.3 (1.3)] 7.8 (1.4) [7.8 (1.3)] 8.3 (1.4) [8.3 (1.3)] 0.09 [0.41] 0.25 [0.05] 0.34 [0.37]
4. Antisocial 5.9 (1.1) [6.0 (1.1)] 8.6 (1.1) [8.9 (1.1)] 8.5 (1.1) [8.2 (1.1)] 2.58 [2.84] 2.46 [2.18] 0.12 [0.68]
Note. LC1 ⫽ Latent Class 1; LC2 ⫽ Latent Class 2; LC3 ⫽ Latent Class 3; PCL-R ⫽ Psychopathy Checklist–Revised.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

explore how the empirically generated, and theoretically coherent, fenders from the Wisconsin replication sample to three latent
variants/subtypes might differ on various demographic and other classes. Consequently, the profiles for the latent classes were the
psychological variables. We therefore are grateful to Joseph New- same as those depicted in Figure 1a. The three latent classes were
man for providing us with a data set obtained from his ongoing labeled in the same way as in the development sample: LC1, LC2,
research in Wisconsin prisons. The set included the scores of male and LC3 (i.e., external classification). Finally, we used the Wis-
offenders on each of the PCL-R items and on a variety of variables consin data to explore several broad assumptions about how the
relevant to testing our latent class model. The PCL-R data used for three latent classes might differ on certain external variables (i.e.,
external validation were not part of the development sample used validation). The restricted range of the PCL-R scores (27– 40) used
to conduct the LPA analysis described in Experiment 1. to generate the latent classes likely served to reduce the strength of
The Wisconsin data set allowed us to determine (a) if the three these associations. Nonetheless, we expected that the latent classes
latent classes obtained in the development sample also could be derived from the external classification would differ from one
identified in an independent set of PCL-R data (the replication another in their associations with some of the external variables,
sample); (b) if there was substantial overlap when the same cases particularly education, intelligence, and measures of antisociality
from the replication sample were allocated to clusters on the and affect. First, we expected psychopathic variants (i.e., manip-
classification routine derived from the development sample (ex- ulative and aggressive psychopaths) to differ from each other in
ternal classification); and (c) if the clusters derived in Step b terms of education, IQ, symptoms of ASPD, and self-reported
differed in meaningful ways from one another with respect to criminal tendencies. More specifically, we expected higher aver-
scores on relevant demographic and psychological variables (ex- age levels of education and IQ for manipulative psychopaths than
ternal validation). We note, however, that these variables were part for aggressive psychopaths, whereas ASPD symptom load and
of a “convenience” sample, and that their use here is preliminary self-reported criminal tendencies should be, on average, higher
to more extensive and planned research with variables selected to among aggressive psychopaths than among manipulative ones.
be of direct relevance to interpretation and validation of the latent Second, we assumed that there would be meaningful differences
classes described herein. between psychopaths (i.e., manipulative and aggressive variants,
or LC1 and LC2, combined), on the one hand, and sociopaths (i.e.,
Method LC3), on the other. More specifically, we hypothesized that psy-
chopaths would display lower mean scores than sociopaths on
Sample. Each of the 497 male offenders in the Wisconsin
sample (referred to here as the replication sample) had a PCL-R self-report measures of negative affectivity and behavioral disin-
score of at least 27 (M ⫽ 30.7, SD ⫽ 2.6), almost identical to that hibition, whereas the reverse would be true for a self-report scale
of the development sample. The scores were based on interview of Callous Affect.
and file information, with intraclass correlations greater than .90 Education and IQ. The external demographic variables in-
for subsamples (see Baskin-Sommers, Baskin, Sommers, & New- cluded years of education and Wechsler Adult Intelligence Scale–
man, 2013). Mean age was 30.9 years (SD ⫽ 7.3). Data on the Revised (Wechsler, 1981) Full Scale IQ estimated from the revised
Wisconsin sample were reported previously in other publications Shipley Institute of Living Scale (Zachary, 1986).
(e.g., Newman, MacCoon, Vaughn, & Sadeh, 2005). ASPD symptoms. ASPD symptom count was indexed accord-
Procedure. First, the data from the replication sample were ing to the total number of corresponding Diagnostic and Statistical
subjected to LPA analogous to the LPA described in Experiment 1
for the development sample (i.e., replication; see Figure 1b). 2
The supplemental materials provide an alternative allocation algorithm
Subsequently, and following the suggestion of Linda Muthén based on R for researchers who do not have access to Mplus. Note that the
(personal communication, April 12, 2013), the latent class proba- R algorithm is a quadratic classification function that rests on manifest
bilities for the replication sample data set were estimated again cluster allocations in the first place. The differences between the algorithms
in Appendixes A and B within the supplemental material seem negligible,
based on the fixed parameter estimates obtained with the 1,451 though: The correlations between the corresponding posterior probabilities
offenders from the development sample (see see Appendix A were around .98 for all three latent classes, and Cohen’s kappa for agree-
within online supplemental material).2 This method assigned of- ment was .96 for manifest allocation.
380 MOKROS ET AL.

Manual of Mental Disorders, 4th Edition (American Psychiatric (7.4% of the replication sample), respectively. These values are
Association, 2000), criteria fulfilled. close to those obtained in the development sample (19.8%, 65.3%,
Welsh Anxiety Scale. The Welsh Anxiety Scale (WAS; and 14.9 for LC1, LC2, and LC3, respectively). As Table 3
Welsh, 1956) is a 39-item inventory (true/false) that yields infor- indicates, the means and standard deviations of the corresponding
mation on four scales: Trouble Concentrating; Negative Affect; latent classes in the development and replication samples were
Low Energy/Pessimism; and Personal Sensitivity (cf. Newman et nearly on par. The respective means differed by 0.5 SD or more in
al., 2005). Newman and colleagues (e.g., Newman et al., 2005; only two cases (i.e., the Interpersonal and Affective factors of
Zeier & Newman, 2013) routinely use WAS scores to subdivide LC3), whereas the average difference (in standard deviation units)
offenders with high PCL-R scores into primary (low anxiety) and was 0.18 for the remaining comparisons, which is a very small
secondary (high anxiety) psychopathy groups. We used the WAS effect.
total score as a measure of anxiety and negative affect. Figure 1b shows the outcome of the LPA for the replication
Behavioral Inhibition System. The Behavioral Inhibition Sys- sample. As visual inspection shows, there was considerable over-
tem (BIS; Carver & White, 1994) was used to assess sensitivity to lap with the results from the development sample (see Figure 1a).
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

aversive stimuli based on seven items. Each of the seven items is Calculating the Euclidean distances between the profiles ob-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

scored on an ordinal rating scale, ranging from 1 (strongly agree) tained from the development sample, on the one hand, and the
to 4 (strongly disagree). Wallace, Malterer, and Newman (2009) replication sample, on the other, the corresponding profiles
reported that BIS scores negatively correlate with PCL-R Factor 1 clearly had the smallest distances from each other, with values
(Interpersonal/Affective) scores. of 0.45 (manipulative psychopaths), 0.64 (aggressive psycho-
Self-Report Psychopathy Scale. The Self-Report Psychopathy paths), and 1.11 (sociopaths), respectively. The Euclidean dis-
Scale⫺III (SRP-III; Neumann & Pardini, 2014; Paulhus, Neu- tances of the cross-modal pairings (e.g., comparing manipula-
mann, & Hare, in press) consists of 64 items, each scored on tive psychopaths from the development sample with sociopaths
5-point scale, ranging from 1 (disagree strongly) to 5 (agree from the replication sample) consistently yielded larger values,
strongly). It has a stable four-factor structure similar to that of the from 2.05⫺3.77. Judging from the Euclidean distances, the
PCL-R (Neal & Sellbom, 2012; Neumann & Pardini, 2014). The latent class of manipulative psychopaths is preserved best in the
labels for the factors are Interpersonal Manipulation, Callous Af- replication study because it has the smallest Euclidean distance,
fect, Erratic Lifestyle, and Criminal Tendencies. The conceptual followed by the aggressive psychopaths and the sociopathic
and empirical associations between the SRP-III and PCL-R factors cluster (in this order). The only major discrepancy concerns the
are well-established (Neumann, Hare, & Pardini, 2014; Paulhus et Interpersonal factor between the last two: There is a reversal of
al., in press). the aggressive psychopath and sociopath groups in this regard.
Statistical analyses. We conducted a series of seven planned Although aggressive psychopaths on average had higher scores
comparisons in which either manipulative psychopaths were com- on the Interpersonal factor than sociopaths in the development
pared with aggressive ones (i.e., LC1 vs. LC2) or in which psy- sample (d ⫽ 0.58), the converse was true for the replication
chopaths were compared with sociopaths (i.e., LC1/LC2 pooled sample (d ⫽ ⫺0.51).
vs. LC3). For these comparisons, the family wise error rate (FWE) When the cases from the replication sample were allocated to
for the Type I error was set at .05 (one-sided) using the sequential latent classes following the specifications for the development
Bonferroni adjustment by Holm (1979). Mean differences were sample described in Experiment 1 (i.e., external classification), 82
tested for statistical significance using parametric independent- of 497 cases (16.5% of the replication sample) were assigned to
samples t tests. Because some of the variables were not distributed LC1, with a mean posterior probability of cluster membership of
normally according to a Kolmogorov–Smirnov goodness-of-fit 0.84 (SD ⫽ 0.17; range: 0.38⫺1.00). About two thirds (n ⫽ 341
test, all independent-samples t tests were carried out in a nonpara- [68.8%]) were assigned to LC2, at an average posterior probability
metric way. For this purpose, permutation tests were used, each of class membership of 0.90 (SD ⫽ 0.12; range: 0.48⫺1.00).
with 10,000 draws for the Monte Carlo. We also were interested in Finally, 73 cases (14.7% of the sample) were allocated to LC3,
how LC1, LC2, and LC3 differed from each other. For this reason, with a mean allocation probability of 0.77 (SD ⫽ 0.18; range:
we conducted the full set of pairwise comparisons in an explor- 0.48⫺1.00). Maximum class allocation probabilities were rela-
atory manner (i.e., without adjusting p values for multiple testing). tively high and on par with those obtained in the development
sample, listed in Table 2 (top).
To assess the degree of overlap between the internal classifica-
Results
tion based on an LPA of the Wisconsin replication sample (see
The outcome of the replication analysis is described in Table 1 above) and this external classification, the outcome of both solu-
(bottom). The results of the LRT of Lo et al. (2001) pointed to a tions was compared. The majority of cases (88.7% [n ⫽ 441])
solution with two latent classes, whereas the BIC indicated that were allocated to corresponding clusters (i.e., LC1/LC1, LC2/LC2,
a three-latent-classes solution was superior. Choosing the latter as or LC3/LC3). Cohen’s kappa coefficient of agreement between the
a frame of reference (in line with the number of classes chosen for two allocation routines calculated to .751, 90% confidence interval
the development sample), the average probabilities for most likely [.69, .81]—a value that indicates an excellent level of agreement
latent class membership were all in excess of .80 (see Table 2, (Fleiss, 1981).
bottom). Assigning cases to manifest clusters based on their prob- Mean (standard deviation) PCL-R total scores for offenders in
abilities of most likely latent class membership yielded clusters LC1, LC2, and LC3 were 29.7 (2.2), 31.3 (2.6), and 29.1 (1.9),
that consisted of 96 offenders (19.3% of the replication sample), respectively. These values are very close to those obtained for the
364 offenders (73.2% of the replication sample), and 37 offenders three latent classes in the development sample. We used the cluster
VARIANTS OF PSYCHOPATHY 381

Table 4
External Classification Results: Mean (SD) Scores of the Three Clusters and Pairwise Comparisons (Cohen’s d) for Each
External Variable

Latent Class 1 Latent Class 2 Latent Class 3 1 vs. 2 1 vs. 3 2 vs. 3


a a
Variables M SD n M SD n M SD n d d da

Education (years) 10.96 1.43 81 9.81 1.74 338 10.10 1.43 71 0.68 0.60 0.17
Estimated IQ 98.80 11.42 76 96.26 12.29 323 96.64 14.45 71 0.21 0.17 0.03
BIS 17.67 3.53 51 17.53 3.93 197 17.82 3.58 44 0.04 0.04 0.07
WAS 11.93 6.99 44 13.39 9.32 160 17.20 9.11 40 0.17 0.66 0.41
ASPD symptoms
Total 8.88 2.80 82 12.07 3.31 342 11.67 3.14 72 0.99 0.94 0.12
Adult 5.37 1.34 82 5.58 1.19 342 5.40 1.11 72 0.17 0.02 0.15
CD 1–7 (Aggressive) 1.11 1.14 82 2.40 1.60 342 2.10 1.31 72 0.85 0.81 0.19
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

CD 8–15 (Antisocial) 2.40 1.67 82 4.09 2.00 342 4.17 1.76 72 0.87 1.03 0.04
This document is copyrighted by the American Psychological Association or one of its allied publishers.

SRP-III
Interpersonal Manipulation 47.19 15.46 31 52.17 14.83 145 51.90 13.89 29 0.33 0.32 0.02
Callous Affect 39.48 7.34 31 40.97 7.33 145 39.21 8.47 29 0.20 0.03 0.23
Erratic Life Style 53.29 9.54 31 56.48 8.03 145 55.76 7.14 29 0.38 0.29 0.09
Criminal Tendencies 45.84 8.99 31 51.76 8.14 145 54.66 7.35 29 0.71 1.08 0.36
Total score 185.81 31.02 31 201.38 26.98 145 201.52 27.35 29 0.56 0.54 0.01
Note. Variables from planned comparisons (see main body of text) are highlighted in boldface. Variables not included in the planned comparisons are
given for completeness. ASPD ⫽ antisocial personality disorder; BIS ⫽ Behavioral Inhibition Scale; CD ⫽ ASPD conduct disorder symptoms; SRP-III ⫽
Self-Report Psychopathy–III scale; WAS ⫽ Welsh Anxiety Scale.
a
Cohen’s d effect size measure (mean difference in pooled SD units).

allocation based on the external classification method as a frame- sociality (ASPD symptom count, Criminal Tendencies), followed
work for the validation analyses (see below) because it provides a by years of education and negative affectivity (WAS), with small
more rigorous test than any clustering based on an internal anal- effects for IQ and Callous Affect.
ysis. External classification entails the transfer of a prescription to
a new sample (i.e., from the development to the replication sam-
Discussion
ple) whereas an internal analysis would run the risk of capitalizing
on the specifics of the replication sample itself. In Experiment 2, we attempted to cross-validate the classifica-
Table 4 contains the mean (standard deviation) scores of the tion model derived in the development sample, using a new sample
three clusters on each of the external variables. In the following, of male offenders with PCL-R scores of at least 27. First, a
the notation p ⬍ FWE represents statistical p values from permu- separate LPA carried out on the replication sample was commen-
tation t tests for the planned comparisons that turned out to be surate with the results for the development sample even though
smaller than the adjusted p values according to an FWE of .05. As there was some ambiguity of fit indices concerning whether a two-
far as the expected differences among the psychopathic variants or three-latent-classes solution was preferable. Settling on the
are concerned, all mean differences were in the expected direction. solution with three latent classes, the allocation probabilities for
On average, manipulative psychopaths had more years of educa- the most appropriate cluster were as high as they were in the
tion (d ⫽ 0.68, t ⫽ 5.50, p ⬍ .001 ⬍ FWE) and a higher IQ than development sample, attesting to the reliability of the classification
aggressive psychopaths (d ⫽ 0.21, t ⫽ 1.64, p ⫽ .050, ns), whereas routine. Furthermore, the relative frequencies of cluster allocation
the aggressive psychopaths had a higher ASPD symptom count reflected those from the development sample. More cogently, the
(d ⫽ 0.99, t ⫽ ⫺8.05, p ⬍ .001 ⬍ FWE) and a higher mean score cluster profiles were highly similar in both instances, as evidenced
on the SRP-III Criminal Tendencies subscale than the former (d ⫽ by their respective Euclidean distances. Finally, the overlap be-
0.71, t ⫽ ⫺3.61, p ⬍ .001 ⬍ FWE). tween classifications derived from the internal analysis of the
Regarding the presumed differences between psychopaths, on replication sample and an external analysis using the specifications
the one hand (i.e., LC1/LC2 pooled), and sociopaths (LC3), on the from the development sample showed a very high level of agree-
other, all mean differences were in the expected direction, but only ment. Consequently, the distinction between manipulative and
one was statistically significant. Sociopaths had a higher mean aggressive psychopaths and between sociopaths appears to be
score on the BIS (d ⫽ 0.07, t ⫽ ⫺0.42, p ⫽ .326, ns) and the WAS robust and replicable.
(d ⫽ 0.46, t ⫽ ⫺2.67, p ⫽ .004 ⬍ FWE) than did the psychopaths, Second, planned comparisons between the groups derived from
whereas the psychopaths on average had higher scores than the external classification highlighted several meaningful differences
former on the SRP-III Callous Affect subscale (d ⫽ 0.20, t ⫽ 1.00, that were generally consistent with the descriptive labels suggested
p ⫽ .164, ns). for the clusters. As expected, manipulative psychopaths were more
In sum, all planned comparisons, except one (behavioral disin- educated than aggressive ones, but the two had relatively similar
hibition), entailed at least small effect sizes in the expected direc- IQ scores. Mirroring the strongest difference between the two
tion, but only four of seven was statistically significant after clusters on the PCL-R factors (i.e., Antisocial), the most pro-
Bonferroni correction. The largest effects were observed for anti- nounced differences were observed for the ASPD symptom count
382 MOKROS ET AL.

and self-reported Criminal Tendencies, with aggressive psycho- would be in line with the cheater and warrior-hawk strategies
paths having higher average scores on these measures than ma- described by Book and Quinsey (2004). Presumably, these are not
nipulative psychopaths, by medium-to-large effect sizes. Although mutually exclusive (or exhaustive) strategies, but rather more a
both the PCL-R Antisocial factor and the ASPD criteria partly rely matter of preference or emphasis influenced by interactions be-
on the same information, the occurrence of a difference on a tween the immediate context and personal characteristics, such as
self-report scale (i.e., Criminal Tendencies) supports the distinc- physical appearance, verbal facility, intelligence, and social back-
tion between the two clusters in terms of social deviance. ground. Psychopaths may be able to draw on a repertoire of
The differences observed between the psychopathic variants and behaviors that is wider and less encumbered by emotional and
the sociopathic cluster, however, were weaker. Only one of three conventional mores than is the case with most people. That is,
corresponding hypotheses (concerning greater negative affectivity charm, manipulation, and persuasion readily give way to intimi-
of sociopaths) could be upheld, whereas neither a higher BIS for dation, threats, and violence, according to the nature of the situa-
sociopaths (as expected, for instance, based on the work of New- tion.
man et al., 2005; but see critique by Poythress et al., 2008; and The third latent class, LC3, is reminiscent of the concepts of
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

response by Newman & Malterer, 2009), nor a higher score of secondary psychopathy, sociopathy, and ASPD, and seems to
This document is copyrighted by the American Psychological Association or one of its allied publishers.

psychopaths on self-reported Callous Affect could be substantiated represent an entity different from LC1 and LC2. Whereas the
by statistical significance. It is conceivable, though, that the dif- latter can be conceived of as variants of prototypical, primary or
ferences between psychopaths and sociopaths would be more idiopathic psychopathy, LC3 was characterized by social devi-
considerable if individuals with lower PCL-R total scores than in ance, but not by the affective features integral to psychopathy.
the present sample were scrutinized as well. The appearance of this latent class was largely a consequence of
Surprisingly, the mean score of manipulative psychopaths on the the relatively low PCL-R threshold (27) set for psychopathy. It
SRP-III Manipulative factor was somewhat lower than that of the did not appear when the threshold was raised to 30 (see Foot-
other groups, perhaps suggesting that the offenders in the former note 1). The warrior-hawk strategy may not be limited to
were more adept than others at controlling their responses to items aggressive psychopaths but may also be prevalent among so-
that tap deception and manipulation. In a meta-analysis, Ray et al. ciopaths, as suggested by Mealey’s (1995b) theory. This is
(2013) did not find a positive linear association between self-rated reflected by both aggressive psychopaths and sociopaths dis-
psychopathic traits and response distortion in research studies. The playing high trait levels of social deviance or antisociality,
findings of Ray et al. (2013), however, may not generalize to according to both observer ratings and self-report.
assessment in real-world clinical settings, especially forensic set- The latent classes identified in this study bear some resemblance
tings, in which there may be incentives for response distortion. In to several of the classes obtained in previous studies that used
the present study, it is plausible that the individuals in LC1 were similar model testing methodology and similar samples of psycho-
more careful in monitoring their responses to the SRP items than pathic male offenders, although somewhat different variables
were those in LC2 or LC3. Still, the self-report inventories we used (Hicks et al., 2004; Poythress et al., 2010; Skeem et al., 2007). For
were not impressive in their ability to discriminate among latent example, Hicks et al. (2004) applied model-based cluster analysis
classes. As suggested above, this may also be because we were to the scores of self-report personality inventories obtained from
dealing with a sample of offenders at the upper end of the psy- male offenders with scores of at least 30 on the PCL-R. They
chopathy continuum, with the reduced variance making it difficult labeled the two latent classes that emerged as emotionally stable
to uncover associations between the latent classes and external psychopaths and aggressive psychopaths, respectively. As in the
variables.3 present study, aggressive psychopaths were more abundant than
emotionally stable (or manipulative) ones. We found a ratio of
General Discussion about 1:3 whereas Hicks et al. noted a ratio of about 1:2 between
manipulative (emotionally stable) and aggressive psychopaths.
This article describes two sets of analyses on putative variants of Two of the four clusters described by Hervé (2007), namely,
psychopathy in male offenders. Our purpose was to clarify and manipulative and macho, were similar to LC1 and LC2, respec-
possibly reconcile the sometimes confusing findings and positions tively, in this study, while a third cluster (pseudopsychopathic)
reported in the literature. Using a large sample of individuals with resembled LC3.
high levels of psychopathic (PCL-R) traits, we first identified three The Wisconsin data set allowed us to start testing the viability of
latent classes: LC1, LC2, and LC3. LC1 and LC2 mirror early our latent class model, first by showing that the variants obtained
clinical conceptualizations of psychopathy, akin to the aggressive/ in the development sample could be identified in an independent
predatory/simple and passive/parasitic/complex types described by sample, and second by demonstrating that there were several
Karpman (1955) and Arieti (1963). We labeled LC1 and LC2 as, expected differences among classes in their associations with
respectively, manipulative and aggressive variants of psychopathy. external variables. Of course, our efforts at external validation
These two variants of psychopathy may reflect different modes of were based as much on the availability of variables as on their
adaptive behavior that are plausible from sociobiological and theoretical plausibility. Nonetheless, the results were encouraging
developmental perspectives (Book & Quinsey, 2004; Harpending
& Sobus, 1987; Hicks et al., 2004; Mealey, 1995b). That is,
3
individuals in LC1 and LC2 may differ primarily in the strategies For example, the correlation (r) between the SRP-III and PCL-R total
scores was .54 for the total sample of 549 offenders in the Wisconsin
with which they attain their goals, those in LC1 more through external validation dataset (i.e., PCL-R total score range: 2⫺39). When
manipulation and deception, and those in LC2 more through ag- only the 206 offenders with a PCL-R score of 27 or higher were used, the
gression and force. In this respect, their interactions with society correlation (r) between the SRP-III and the PCL-R dropped to .33.
VARIANTS OF PSYCHOPATHY 383

and should lead to attempts by other researchers to replicate and very limited data on correlates that should be relevant to evaluating
test our latent class model with diverse samples and an array of our model. However, many other researchers will have the data
relevant variables. Because of the dramatic increase in research on needed to test our model and to examine in detail its external
psychopathy and use of the PCL-R over the past decade or so, correlates.
many large sample data sets are available to evaluate both the The present study dealt exclusively with adult male offenders.
structure and the correlates of our latent class model. Similar analyses should be conducted with female offenders
For instance, the observed differences between psychopaths and (Hicks et al., 2010; Lee & Salekin, 2010), possibly testing sex-
sociopaths in terms of emotional facilitation effects in lexical limitation models within behavioral genetics, following Mealey’s
decision tasks, as measured by response latencies and event-related (1995b) suggestions.
brain potentials (Kosson, Lorenz, & Newman, 2006; Verona, The latent class analyses identified in this study were based
Sprague, & Sadeh, 2012; Williamson, Harpur, & Hare, 1991), on the dimensions of a single measure of psychopathy, the
should also become apparent if the present classification algorithm PCL-R. However, it seems more reasonable to conduct an LPA
was adopted. Particularly interesting would be research that looks on empirically derived dimensions of psychopathy than to in-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

at the latent classes through the lenses of, for example, molecular/ clude other dimensions or variables (especially self-reports)
This document is copyrighted by the American Psychological Association or one of its allied publishers.

behavioral genetics and gene⫺environment interactions (Eme, with uncertain or hypothetical relevance to psychopathy. Thus,
2013; Viding & McCrory, 2012), developmental psychopathology we sought to identify clusters derived from the PCL-R alone,
(Frick, Ray, Thornton, & Kahn, 2014; Lynam & Gudonis, 2005), and to evaluate their relevance by comparing them on other
aggression (Blair, 2010; Glenn & Raine, 2009), cognitive⫺emo-
correlates (as was done by Hicks et al., 2004). An added
tional integration (Brook, Brieman, & Kosson, 2013; Hare, 1998;
advantage of limiting the dependent variables for clustering to
Newman et al., 2006), attentional processes (Zeier & Newman,
the PCL-R factors is that many researchers already have such
2013), and neuroimaging (Anderson & Kiehl, 2013; Boccardi,
data, making it relatively easy for them to use the classification
2013). In terms of molecular genetics, it would be intriguing to test
algorithm from Appendix A within the online supplemental
the supposition of Viding and McCrory (2012) that different
material to test a variety of hypotheses about the external
alleles might be at play with regard to the psychopathic variants
correlates of our model.
(LC1 and LC2), on the one hand, and the sociopathic/antisocial
variant (LC3), on the other. The long allele variant (rs25531) at the The generalizability of our latent class model and its correlates
promoter region (5HTTLPR) of the serotonin transporter gene readily can be evaluated in extant samples of adolescent and
(SLC6A4) may be more prevalent among LC1/LC2 psychopaths female offenders, forensic and civil psychiatric patients, and mem-
than among LC3 sociopaths (cf. Glenn, 2011), for instance. In bers of the community. We believe that initial evaluations of our
contrast, the low-activity polymorphism of the monoamine oxidase model should be based on similar four-factor models of psychop-
A gene (MAOA-L) may be more strongly associated with LC3 athy. This would allow researchers to work from a common
given the link between this polymorphism and reactive aggression framework, thereby facilitating comparisons across conceptually
(Buckholtz & Meyer-Lindenberg, 2008). Indeed, Sadeh, Javdani, similar studies. This also would allow researchers to compare our
and Verona (2013) recently reported an association between the latent class model with those obtained from the use of other
long-allele polymorphism (5-HTT) at the serotonin transporter measures of psychopathy, including those that are based on or
gene promoter region with the Affective factor of the Psychopathy include self-reports. Psychopathic variants should become appar-
Checklist: Screening Version (PCL:SV; Hart, Cox, & Hare, 1995) ent when looking at the traits that constitute psychopathy measured
and a link between MAOA-L and the Lifestyle factor of the in other reliable ways.
PCL:SV in a forensic sample.4 As a final point, our model is based on analysis of offenders
Various personal and demographic variables also may contrib- with very high PCL-R scores. This approach allowed us to inden-
ute to different behavioral styles and strategies among the latent tify meaningful variants of psychopathy. An alternative approach
classes. Individuals in LC1 may be more physically and personally involves LPA of an entire distribution of psychopathy scores, thus
attractive, better educated, more intelligent, or verbally facile, and allowing for the emergence of a variety of clinical profiles. Either
less likely to have come from a disadvantaged or abusive back- approach would be useful to clinicians and researchers, depending
ground, than are those in LC2 or LC3. Presumably, such individ- on whether they are interested in a nuanced description of indi-
uals akin to LC1 are more likely to be attracted to and successful viduals with high PCL-R scores or in obtaining a general picture of
in the corporate and political worlds than are those in the other the entire set of individuals. This issue is beyond the scope of the
latent classes (Babiak & Hare, 2006; Babiak, Neumann, & Hare, current article, but is discussed in detail elsewhere (Hare, Neu-
2010). As an aid for future research, the supplemental materials mann, & Mokros, in press).
provide the code for allocation routines in Mplus or R toward the
latent classes presented herein (see Footnote 2).
4
PCL-R total scores typically vary from about 3⫺10 among various
Limitations samples of noncriminals, with scores on its four factors being proportion-
ately lower (Hare, 2003, pp. 52-53). Additionally, a derivative of the
There are several limitations to the present study. Because we PCL-R, the 12-item PCL:SV (Hart et al., 1995) is conceptually and
empirically equivalent to the 20-item PCL-R, making it possible to esti-
lacked information on adverse or traumatic experiences of the
mate PCL-R scores from PCL:SV scores by multiplying the latter by
individuals in the sample, we could not test hypotheses that would 20/12 ⫽ 1.67 (see Hare et al., 2013, p. 238). For example, Neumann and
follow from Mealey’s (1995b) etiological speculations about psy- Hare (2008) obtained a mean PCL:SV score of 3.5 for a community male
chopathy and sociopathy. Also, for the current analyses, we had sample. The PCL-R score “equivalent” to this PCL:SV score is 5.9.
384 MOKROS ET AL.

References Cleckley, H. (1976). The mask of sanity (5th ed.). St. Louis, MO: Mosby.
Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155–159.
American Psychiatric Association. (2000). Diagnostic and statistical man- http://dx.doi.org/10.1037/0033-2909.112.1.155
ual of mental disorders (4th ed., text rev.). Washington, DC: Author. Conger, A. J., & Lipshitz, R. (1973). Measures of reliability for profiles
Anderson, N. E., & Kiehl, K. A. (2013). Functional neuroimaging and and test batteries. Psychometrika, 38, 411– 427. http://dx.doi.org/
psychopathy. In K. A. Kiehl & W. P. Sinnott-Armstrong (Eds.), Hand- 10.1007/BF02291663
book on psychopathy and law (pp. 131–149). New York, NY: Oxford Drislane, L. E., Patrick, C. J., Sourander, A., Sillanmäki, L., Aggen, S. H.,
University Press.
Elonheimo, H., . . . Kendler, K. S. (2014). Distinct variants of extreme
Arieti, S. (1963). Psychopathic personality: Some views on its psychopa-
psychopathic individuals in society at large: Evidence from a
thology and psychodynamics. Comprehensive Psychiatry, 4, 301–312.
population-based sample. Personality Disorders: Theory, Research, and
http://dx.doi.org/10.1016/S0010-440X(63)80056-5
Treatment, 5, 154 –163. http://dx.doi.org/10.1037/per0000060
Babiak, P., & Hare, R. D. (2006). Snakes in suits: When psychopaths go to
Eme, R. (2013). MAOA and male antisocial behavior: A review. Aggres-
work. New York, NY: Harper Collins.
sion and Violent Behavior, 18, 395–398. http://dx.doi.org/10.1016/j.avb
Babiak, P., Neumann, C. S., & Hare, R. D. (2010). Corporate psychopathy:
.2013.02.001
Talking the walk. Behavioral Sciences & the Law, 28, 174 –193.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Baskin-Sommers, A. R., Baskin, D. R., Sommers, I. B., & Newman, J. P. Eysenck, H. J. (1995). Psychopathology: Type or trait? Behavioral and
This document is copyrighted by the American Psychological Association or one of its allied publishers.

(2013). The intersectionality of sex, race, and psychopathology in pre- Brain Sciences, 18, 555–556. http://dx.doi.org/10.1017/S0140525X00039765
dicting violent crimes. Criminal Justice and Behavior, 40, 1068 –1091. Falkenbach, D., Poythress, N., & Creevy, C. (2008). The exploration of
http://dx.doi.org/10.1177/0093854813485412 subclinical psychopathic subtypes and the relationship with types of
Bauer, D. J., & Curran, P. J. (2004). The integration of continuous and aggression. Personality and Individual Differences, 44, 821– 832. http://
discrete latent variable models: Potential problems and promising op- dx.doi.org/10.1016/j.paid.2007.10.012
portunities. Psychological Methods, 9, 3–29. http://dx.doi.org/10.1037/ Falkenbach, D. M., Stern, S. B., & Creevy, C. (2014). Psychopathy
1082-989X.9.1.3 variants: Empirical evidence supporting a subtyping model in a com-
Beauchaine, T. P. (2003). Taxometrics and developmental psychopathol- munity sample. Personality Disorders: Theory, Research, and Treat-
ogy. Development and Psychopathology, 15, 501–527. http://dx.doi.org/ ment, 5, 10 –19. http://dx.doi.org/10.1037/per0000021
10.1017/S0954579403000270 Fleiss, J. L. (1981). Statistical methods for rates and proportions (2nd ed.).
Blackburn, R. (2009). Subtypes of psychopath. In M. McMurran (Ed.), New York, NY: Wiley.
Personality, personality disorder and violence: An evidence based ap- Forth, A. E., Bo, S., & Kongerslev, M. (2013). Assessment of psychopathy:
proach (pp. 113–132). Chichester, England: Wiley. The Hare Psychopathy Checklists. In K. A. Kiehl & W. P. Sinnott-
Blair, R. J. R. (2010). Psychopathy, frustration, and reactive aggression: Armstrong (Eds.), Handbook on psychopathy and law (pp. 5–33). New
The role of ventromedial prefrontal cortex. British Journal of Psychol- York, NY: 4 Oxford University Press.
ogy, 101, 383–399. http://dx.doi.org/10.1348/000712609X418480 Fraley, C., & Raftery, A. E. (2002). Model-based clustering, discriminant
Boccardi, M. (2013). Structural brain abnormalities and psychopathy. In analysis, and density estimation. Journal of the American Statistical
K. A. Kiehl & W. P. Sinnott-Armstrong (Eds.), Handbook on psychop- Association, 97, 611–631. http://dx.doi.org/10.1198/016214502760047131
athy and law (pp. 150 –157). New York, NY: Oxford University Press. Frick, P. J., Ray, J. V., Thornton, L. C., & Kahn, R. E. (2014). Can
Bolt, D. M., Hare, R. D., Vitale, J. E., & Newman, J. P. (2004). A callous-unemotional traits enhance the understanding, diagnosis, and
multigroup item response theory analysis of the Psychopathy Checklist– treatment of serious conduct problems in children and adolescents? A
Revised. Psychological Assessment, 16, 155–168. http://dx.doi.org/ comprehensive review. Psychological Bulletin, 140, 1–57. http://dx.doi
10.1037/1040-3590.16.2.155 .org/10.1037/a0033076
Book, A. S., & Quinsey, V. L. (2004). Psychopaths: Cheaters or warrior Glenn, A. L. (2011). The other allele: Exploring the long allele of the
hawks? Personality and Individual Differences, 36, 33– 45. http://dx.doi serotonin transporter gene as a potential risk factor for psychopathy: A
.org/10.1016/S0191-8869(03)00049-7 review of the parallels in findings. Neuroscience and Biobehavioral
Brook, M., Brieman, C. L., & Kosson, D. S. (2013). Emotion processing in Reviews, 35, 612– 620. http://dx.doi.org/10.1016/j.neubiorev.2010.07
Psychopathy Checklist⫺assessed psychopathy: A review of the litera-
.005
ture. Clinical Psychology Review, 33, 979 –995. http://dx.doi.org/10.1016/
Glenn, A. L., & Raine, A. (2009). Psychopathy and instrumental aggres-
j.cpr.2013.07.008
sion: Evolutionary, neurobiological, and legal perspectives. Interna-
Buckholtz, J. W., & Meyer-Lindenberg, A. (2008). MAOA and the neu-
tional Journal of Law and Psychiatry, 32, 253–258. http://dx.doi.org/
rogenetic architecture of human aggression. Trends in Neurosciences,
10.1016/j.ijlp.2009.04.002
31, 120 –129. http://dx.doi.org/10.1016/j.tins.2007.12.006
Hare, R. D. (1970). Psychopathy: Theory and research. New York, NY:
Buss, D. M. (2009). How can evolutionary psychology successfully ex-
Wiley.
plain personality and individual differences? Perspectives on Psycho-
logical Science, 4, 359 –366. http://dx.doi.org/10.1111/j.1745-6924.2009 Hare, R. D. (1998). Psychopathy, affect, and behavior. In D. Cooke, A.
.01138.x Forth, & R. Hare (Eds.), Psychopathy: Theory, research, and implica-
Carver, C. S., & White, T. L. (1994). Behavioral inhibition, behavioral tions for society (pp. 105–137). Dordrecht, The Netherlands: Kluwer.
activation, and affective responses to impending reward and punishment: http://dx.doi.org/10.1007/978-94-011-3965-6_6
The BIS/BAS Scales. Journal of Personality and Social Psychology, 67, Hare, R. D. (2003). The Hare Psychopathy Checklist⫺Revised (2nd ed.).
319 –333. http://dx.doi.org/10.1037/0022-3514.67.2.319 Toronto, ON: Multi-Health Systems.
Claes, L., Tavernier, G., Roose, A., Bijttebier, P., Smith, S. F., & Lilien- Hare, R. D., Black, P., & Walsh, Z. (2013). The PCL-R: Forensic appli-
feld, S. O. (2014). Identifying personality subtypes based on the five- cations and limitations. In R. P. Archer & E. M. Archer (Eds.), Forensic
factor model dimensions in male prisoners: Implications for psychopa- uses of clinical assessment instruments (2nd ed., pp. 266 –290). New
thy and criminal offending. International Journal of Offender Therapy York, NY: Routledge.
and Comparative Criminology, 58, 41–58. http://dx.doi.org/10.1177/ Hare, R. D., & Neumann, C. S. (2006). The PCL-R assessment of psy-
0306624X12462013 chopathy: Development, structural properties, and new directions. In
Cleckley, H. (1941). The mask of sanity. St. Louis, MO: Mosby. C. J. Patrick (Ed.), Handbook of psychopathy (pp. 58 – 88). New York,
Cleckley, H. (1964). The mask of sanity (4th ed.). St. Louis, MO: Mosby. NY: Guilford Press.
VARIANTS OF PSYCHOPATHY 385

Hare, R. D., & Neumann, C. S. (2008). Psychopathy as a clinical and Lee, Z., & Salekin, R. T. (2010). Psychopathy in a noninstitutional sample:
empirical construct. Annual Review of Clinical Psychology, 4, 217–246. Differences in primary and secondary subtypes. Personality Disorders:
http://dx.doi.org/10.1146/annurev.clinpsy.3.022806.091452 Theory, Research, and Treatment, 1, 153–169. http://dx.doi.org/10.1037/
Hare, R. D., Neumann, C. S., & Mokros, A. (in press). The PCL-R a0019269
assessment of psychopathy: Development, properties, debates, and new Lo, Y., Mendell, N., & Rubin, D. (2001). Testing the number of compo-
directions. In C. J. Patrick (Ed.), Handbook of psychopathy (2nd ed.). nents in a normal mixture. Biometrika, 88, 767–778. http://dx.doi.org/
New York, NY: Guilford Press. 10.1093/biomet/88.3.767
Hare, R. D., Neumann, C. S., & Widiger, T. A. (2012). Psychopathy. In Lykken, D. T. (1995). The antisocial personalities. Hillsdale, NJ: Erlbaum.
T. A. Widiger (Ed.), The Oxford handbook of personality disorders (pp. Lynam, D. R., & Gudonis, L. (2005). The development of psychopathy.
478 –504). New York, NY: Oxford University Press. Annual Review of Clinical Psychology, 1, 381– 407. http://dx.doi.org/
Harpending, H., & Sobus, J. (1987). Sociopathy as an adaptation. Ethology 10.1146/annurev.clinpsy.1.102803.144019
and Sociobiology, 8(Supp.), 63–72. http://dx.doi.org/10.1016/0162- Mardia, K. V. (1974). Applications of some measures of multivariate
3095(87)90019-7 skewness and kurtosis in testing normality and robustness studies.
Harris, G. T., Rice, M. E., Hilton, N. Z., Lalumiére, M. L., & Quinsey, Sankhya៮ : The Indian Journal of Statistics, Series B, 36, 115–128. Re-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

V. L. (2007). Coercive and precocious sexuality as a fundamental aspect trieved from http://www.jstor.org/stable/25051892
This document is copyrighted by the American Psychological Association or one of its allied publishers.

of psychopathy. Journal of Personality Disorders, 21, 1–27. http://dx McCord, W., & McCord, J. (1964). The psychopath: An essay on the
.doi.org/10.1521/pedi.2007.21.1.1 criminal mind. Princeton, NJ: D. Van Nostrand.
Hart, S. D., Cox, D. N., & Hare, R. D. (1995). Manual for the Psychopathy McLachlan, G., & Peel, D. (2000). Finite mixture models. New York, NY:
Checklist: Screening Version (PCL:SV). Toronto, ON: Multi-Health Wiley. http://dx.doi.org/10.1002/0471721182
Systems. Mealey, L. (1995a). Primary sociopathy (psychopathy) is a type, secondary
Herpertz, S. C., & Sass, H. (2000). Emotional deficiency and psychopathy. is not. Behavioral and Brain Sciences, 18, 579 –599. http://dx.doi.org/
Behavioral Sciences & the Law, 18, 567–580. http://dx.doi.org/10.1002/ 10.1017/S0140525X00040024
1099-0798(200010)18:5⬍567::AID-BSL410⬎3.0.CO;2-8 Mealey, L. (1995b). The sociobiology of sociopathy: An integrated evo-
Hervé, H. (2007). Psychopathic subtypes: Historical and contemporary lutionary model. Behavioral and Brain Sciences, 18, 523–540. http://dx
perspectives. In J. C. Yuille & H. Hervé (Eds.), The psychopath: Theory, .doi.org/10.1017/S0140525X00039595
research, and practice (pp. 431– 460). Mahwah, NJ: Erlbaum. Millon, T., Krueger, R. F., & Simonsen, E. (2010). Contemporary direc-
Hervé, H., & Yuille, J. C. (Eds.). (2007). The psychopath: Theory, re- tions in psychopathology. New York, NY: Guilford Press.
search, and practice. Mahwah, NJ: Erlbaum. Muthén, L. K., & Muthén, B. O. (2011). Mplus (Version 6.12 for Mac)
Hicks, B. M., Markon, K. E., Patrick, C. J., Krueger, R. F., & Newman, [Computer software]. Los Angeles, CA: Author.
J. P. (2004). Identifying psychopathy subtypes on the basis of person- Neal, T. M. S., & Sellbom, M. (2012). Examining the factor structure of the
ality structure. Psychological Assessment, 16, 276 –288. http://dx.doi Hare Self-Report Psychopathy Scale. Journal of Personality Assessment,
.org/10.1037/1040-3590.16.3.276 94, 244 –253. http://dx.doi.org/10.1080/00223891.2011.648294
Hicks, B. M., Vaidyanathan, U., & Patrick, C. J. (2010). Validating female Neumann, C. S., & Hare, R. D. (2008). Psychopathic traits in a large
psychopathy subtypes: Differences in personality, antisocial and violent community sample: Links to violence, alcohol use, and intelligence.
behavior, substance abuse, trauma, and mental health. Personality Dis- Journal of Consulting and Clinical Psychology, 76, 893– 899. http://dx
orders: Theory, Research, and Treatment, 1, 38 –57. http://dx.doi.org/ .doi.org/10.1037/0022-006X.76.5.893
10.1037/a0018135 Neumann, C. S., Hare, R. D., & Newman, J. P. (2007). The super-ordinate
Holm, S. (1979). A simple sequentially rejective multiple test procedure. nature of the Psychopathy Checklist⫺Revised. Journal of Personality
Scandinavian Journal of Statistics, 6, 65–70. Retrieved from http:// Disorders, 21, 102–117. http://dx.doi.org/10.1521/pedi.2007.21.2.102
www.jstor.org/stable/4615733 Neumann, C. S., Hare, R. D., & Pardini, D. A. (2014). Antisociality and the
Karpman, B. (1929). The problem of psychopathies. Psychiatric Quarterly, construct of psychopathy: Data from across the globe. Journal of Per-
3, 495–525. http://dx.doi.org/10.1007/BF01585441 sonality. Advance online publication. http://dx.doi.org/10.1111/jopy
Karpman, B. (1941). On the need of separating psychopathy into distinct .12127
clinical types: The symptomatic and the idiopathic. Journal of Criminal Neumann, C. S., & Pardini, D. (2014). Factor structure and construct
Psychopathology, 3, 112–137. validity of the Self-Report Psychopathy (SRP) scale and the Youth
Karpman, B. (1948). The myth of the psychopathic personality. The Psychopathic Traits Inventory (YPI) in young men. Journal of Person-
American Journal of Psychiatry, 104, 523–534. http://dx.doi.org/ ality Disorders, 28, 419–433. http://dx.doi.org/10.1521/pedi_2012_26_063
10.1176/appi.ajp.104.9.523 Newman, J. P., MacCoon, D. G., Buckholtz, J., Bertsch, J., Hiatt, K. D., &
Karpman, B. (1955). Criminal psychodynamics: A platform. Archives of Vaughn, L. J. (2006). Deficient integration of top-down and bottom-up
Criminal Psychodynamics, 1, 3–100. influences on attention in psychopaths: Potential contribution of the
Kass, R., & Raftery, A. (1995). Bayes factors. Journal of the American septal-hippocampal system. In D. Barch (Ed.), Cognitive and affective
Statistical Association, 90, 773–795. http://dx.doi.org/10.1080/ neuroscience of psychopathology. New York, NY: Oxford University
01621459.1995.10476572 Press.
Kiehl, K. A., & Sinnott-Armstrong, W. (Eds.). (2013). Oxford handbook on Newman, J. P., MacCoon, D. G., Vaughn, L. J., & Sadeh, N. (2005).
psychopathy and the law. New York, NY: Oxford University Press. Validating a distinction between primary and secondary psychopathy
Kosson, D. S., Lorenz, A. R., & Newman, J. P. (2006). Effects of comorbid with measures of Gray’s BIS and BAS constructs. Journal of Abnormal
psychopathy on criminal offending and emotion processing in male Psychology, 114, 319 –323.
offenders with antisocial personality disorder. Journal of Abnormal Newman, J. P., & Malterer, M. B. (2009). Problems with the BIS/BAS
Psychology, 115, 798 – 806. http://dx.doi.org/10.1037/0021-843X.115.4 scales or Lykken’s model of primary psychopathy? A reply to Poythress
.798 et al. (2008). Personality and Individual Differences, 46, 673– 677.
Kupzyk, K. A. (2011). Introduction to mixture modeling. Retrieved from http://dx.doi.org/10.1016/j.paid.2009.01.037
http://r2ed.unl.edu/presentations/2011/RMS/012111_Kupzyk/ Nylund, K. L., Asparouhov, T., & Muthén, B. O. (2007). Deciding on the
012111_Kupzyk.pdf number of classes in latent class analysis and growth mixture modeling:
386 MOKROS ET AL.

A Monte Carlo simulation study. Structural Equation Modeling, 14, Vermunt, J. K., & Magidson, J. (2006). Latent class cluster analysis. In
535–569. http://dx.doi.org/10.1080/10705510701575396 J. A. Hagenaars & A. L. McCutcheon (Eds.), Applied latent class
Patrick, C. J. (Ed.). (2006). Handbook of psychopathy. New York, NY: analysis (pp. 89 –106). New York, NY: Cambridge University Press.
Guilford Press. Verona, E., Sprague, J., & Sadeh, N. (2012). Inhibitory control and
Patrick, C. J., Curtin, J. J., & Tellegen, A. (2002). Development and negative emotional processing in psychopathy and antisocial personality
validation of a brief form of the Multidimensional Personality Question- disorder. Journal of Abnormal Psychology, 121, 498 –510. http://dx.doi
naire. Psychological Assessment, 14, 150 –163. http://dx.doi.org/ .org/10.1037/a0025308
10.1037/1040-3590.14.2.150 Viding, E., & McCrory, E. J. (2012). Genetic and neurocognitive contri-
Paulhus, D. L., Neumann, C. S., & Hare, R. D. (in press). Manual for the butions to the development of psychopathy. Development and Psycho-
Self-Report Psychopathy Scale. Toronto, ON. Multi-Health Systems. pathology, 24, 969 –983. http://dx.doi.org/10.1017/S095457941200048X
Poythress, N. G., Edens, J. F., Landfield, K., Lilienfeld, S. O., Skeem, J. L., Wallace, J. F., Malterer, M. B., & Newman, J. P. (2009). Mapping Gray’s
& Douglas, K. S. (2008). A critique of Carver and White’s (1994) BIS and BAS constructs onto Factor 1 and Factor 2 of Hare’s Psychop-
Behavioral Inhibition Scale (BIS) for investigating Lykken’s (1995) athy Checklist⫺Revised. Personality and Individual Differences, 47,
theory of primary psychopathy. Personality and Individual Differences, 812– 816. http://dx.doi.org/10.1016/j.paid.2009.06.019
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

45, 269 –275. http://dx.doi.org/10.1016/j.paid.2008.04.014 Ward, T., & Durrant, R. (2011). Evolutionary behavioural science and
Poythress, N. G., Edens, J. F., Skeem, J. L., Lilienfeld, S. O., Douglas, crime: Aetiological and intervention implications. Legal and Crimino-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

K. S., Frick, P. J., . . . Wang, T. (2010). Identifying subtypes among logical Psychology, 16, 193–210. http://dx.doi.org/10.1111/j.2044-8333
offenders with antisocial personality disorder: A cluster-analytic study. .2011.02020.x
Journal of Abnormal Psychology, 119, 389 – 400. http://dx.doi.org/ Wechsler, D. (1981). Wechsler Adult Intelligence Scale-Revised. San An-
10.1037/a0018611 tonio, TX: Psychological Corporation.
Poythress, N. G., & Skeem, J. L. (2006). Disaggregating psychopathy: Welsh, G. (1956). Factor dimensions A and R. In G. S. Welsh & W. G.
Where and how to look for subtypes. In C. J. Patrick (Ed.), Handbook of Dahlstrom (Eds.), Basic readings on the MMPI in psychology and
psychopathy (pp. 172–192). New York, NY: Guilford Press. medicine (pp. 264 –281). Minneapolis, MN: University of Minnesota
Ray, J. V., Hall, J., Rivera-Hudson, N., Poythress, N. G., Lilienfeld, S. O., Press.
& Morano, M. (2013). The relation between self-reported psychopathic Widiger, T. A. (Ed.). (2012). The Oxford handbook of personality disor-
traits and distorted response styles: A meta-analytic review. Personality ders. New York, NY: Oxford University Press. http://dx.doi.org/
Disorders: Theory, Research, and Treatment, 4, 1–14. http://dx.doi.org/ 10.1093/oxfordhb/9780199735013.001.0001
10.1037/a0026482 Williamson, S., Harpur, T. J., & Hare, R. D. (1991). Abnormal processing
Rost, J. (2006). Latent-Class-Analyse [Latent class analysis]. In F. Peter- of affective words by psychopaths. Psychophysiology, 28, 260 –273.
mann & M. Eid (Eds.), Handbuch der psychologischen Diagnostik http://dx.doi.org/10.1111/j.1469-8986.1991.tb02192.x
[Handbook of psychological assessment] (pp. 275–287). Göttingen, Ger- Yildirim, B. O., & Derksen, J. J. L. (2013). Systematic review, structural
many: Hogrefe. analysis, and new theoretical perspectives on the role of serotonin and
Sadeh, N., Javdani, S., & Verona, E. (2013). Analysis of monoaminergic associated genes in the etiology of psychopathy and sociopathy. Neuro-
genes, childhood abuse, and dimensions of psychopathy. Journal of science and Biobehavioral Reviews, 37, 1254 –1296. http://dx.doi.org/
Abnormal Psychology, 122, 167–179. http://dx.doi.org/10.1037/ 10.1016/j.neubiorev.2013.04.009
a0029866 Zachary, R. A. (1986). Shipley Institute of Living Scale⫺Revised manual.
Schwarz, G. (1978). Estimating the dimension of a model. Annals of Los Angeles, CA: Western Psychological Services.
Statistics, 6, 461– 464. http://dx.doi.org/10.1214/aos/1176344136 Zeier, J. D., & Newman, J. P. (2013). Feature-based attention and conflict
Skeem, J., Johansson, P., Andershed, H., Kerr, M., & Louden, J. E. (2007). monitoring in criminal offenders: Interactive relations of psychopathy
Two subtypes of psychopathic violent offenders that parallel primary with anxiety and externalizing. Journal of Abnormal Psychology, 122,
and secondary variants. Journal of Abnormal Psychology, 116, 395– 797– 806. http://dx.doi.org/10.1037/a0033873
409. http://dx.doi.org/10.1037/0021-843X.116.2.395
Skeem, J. L., & Cooke, D. J. (2010). Is criminal behavior a central
component of psychopathy? Conceptual directions for resolving the Received February 6, 2014
debate. Psychological Assessment, 22, 433– 445. http://dx.doi.org/10.1037/ Revision received December 1, 2014
a0008512 Accepted December 8, 2014 䡲

View publication stats

You might also like