Vous êtes sur la page 1sur 14

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

net/publication/225071934

The Psychopathy Checklist-Revised (PCL-R), Low Anxiety, and Fearlessness:


A Structural Equation Modeling Analysis

Article  in  Personality Disorders: Theory, Research, and Treatment · May 2012


DOI: 10.1037/a0027886 · Source: PubMed

CITATIONS READS
95 4,632

3 authors, including:

Craig S Neumann Robert Hare


University of North Texas University of British Columbia - Vancouver
160 PUBLICATIONS   8,551 CITATIONS    229 PUBLICATIONS   30,941 CITATIONS   

SEE PROFILE SEE PROFILE

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

Examining the links between attachment style, personality traits, and sexual behavior View project

Early Schizophrenia research View project

All content following this page was uploaded by Craig S Neumann on 28 March 2018.

The user has requested enhancement of the downloaded file.


Personality Disorders: Theory, Research, and
Treatment
The Psychopathy Checklist-Revised (PCL-R), Low Anxiety,
and Fearlessness: A Structural Equation Modeling
Analysis
Craig S. Neumann, Robert D. Hare, and Peter T. Johansson
Online First Publication, May 28, 2012. doi: 10.1037/a0027886

CITATION
Neumann, C. S., Hare, R. D., & Johansson, P. T. (2012, May 28). The Psychopathy
Checklist-Revised (PCL-R), Low Anxiety, and Fearlessness: A Structural Equation Modeling
Analysis. Personality Disorders: Theory, Research, and Treatment. Advance online
publication. doi: 10.1037/a0027886
Personality Disorders: Theory, Research, and Treatment © 2012 American Psychological Association
2012, Vol. ●●, No. ●, 000 – 000 1949-2715/12/$12.00 DOI: 10.1037/a0027886

The Psychopathy Checklist-Revised (PCL-R), Low Anxiety, and


Fearlessness: A Structural Equation Modeling Analysis

Craig S. Neumann Robert D. Hare


University of North Texas University of British Columbia

Peter T. Johansson
Örebro University

The current study employed a large representative sample of violent male offenders
within the Swedish prison system to examine the factor structure of the PCL-R and the
latent variable relations between the PCL-R items and clinical ratings of low trait
anxiety and trait fearlessness (LAF). Consistent with previous research, confirmatory
factor analysis (CFA) revealed strong support for the four-factor model of psychopathy
(Interpersonal, Affective, Lifestyle, and Antisocial). Also, a series of CFAs revealed
that the LAF items could be placed on any of the PCL-R factors without any changes
in model fit. Finally, structural equation modeling results indicated that a PCL-R
superordinate factor was able to account for most of the variance of a separate LAF
factor. Taken together, the results indicate that if low anxiety and fearlessness, as
measured via clinical ratings, are part of the psychopathy construct they are compre-
hensively accounted for by extant PCL-R items.

Keywords: psychopathy, PCL-R, low anxiety, fearlessness, structural properties

Many hundreds of empirical articles have ments. This belief is based, in part, on the
described the psychometric properties, struc- interpretation of clinical accounts of psychopa-
ture, and correlates of the Psychopathy Check- thy by Cleckley (1976) and other early clini-
list-Revised (PCL-R; Hare, 1991, 2003) and its cians (see Hervé, 2007), and on findings from
derivatives (see www.hare.org). The reliability, early laboratory experiments (e.g., Hare, 1965;
validity, and utility of these measures of the Lykken, 1957). Hare and Neumann (2008) have
psychopathy construct are firmly established in argued that the role of anxiety and fear in clin-
a wide variety of contexts, basic and applied. ical accounts of psychopathy is not entirely
Nonetheless, some investigators believe that clear.
specific measures of low trait anxiety and trait
fearlessness should be included in these instru- Low Anxiety

Cleckley (1976, p. 340) wrote of the psycho-


path, “Within himself he appears almost as in-
capable of anxiety as of profound remorse”
Craig S. Neumann, Department of Psychology, Univer-
sity of North Texas; Peter T. Johansson, Center for Devel- (Cleckley, 1976, p. 340), a statement usually
opmental Research, Department of Behavioral, Social, and quoted by those who believe that lack of trait
Legal Sciences, Örebro University, Örebro, Sweden; Robert anxiety should have been included in the
D. Hare, PhD, Department of Psychology, University of PCL-R. But, in the previous sentence Cleckley
British Columbia, and Darkstone Research Group, Vancou- commented that psychopaths experience ten-
ver, BC.
Robert Hare receives royalties from the sale of PCL-R sion or uneasiness but that it “seems provoked
and its derivatives. Financial support for this research was entirely by external circumstances, never by
generously provided by the William H. Donner Foundation, feelings of guilt, remorse, or intrapersonal inse-
Inc. (CSN). curity.” As Hare and Neumann (2008, p. 228)
Correspondence concerning this article should be ad-
dressed to Craig S. Neumann, Psychology Department, put it, “This psychodynamic perspective sug-
1155 Union Circle, No. 311280, University of North Texas, gests that it is not so much a lack of anxiety that
Denton, TX 76203. E-mail: craig.neumann@unt.edu differentiates psychopaths from others as it is
1
2 NEUMANN, HARE, AND JOHANSSON

the source of the anxiety (intra- or extrapsy- anxiety were positively correlated with several
chic).” Further, in the first edition of The Mask measures of psychopathy, including the Psy-
of Sanity, the Clinical Profile contained only chopathy Checklist: Youth Version (PCL: YV;
half a sentence on the topic: “[The psychopath] Forth, Kosson, & Hare, 2003) and the Antiso-
. . . is usually free from any marked nervousness cial Process Screening Device (APSD; Frick &
or other symptoms of psychoneurosis” (Cleck- Hare. 2001). Kubak and Salekin (2009, p. 280)
ley, 1941, p. 239; emphasis added). In a later noted that the “. . . findings suggest that adoles-
edition he said, “The true psychopaths person- cent psychopathy co-occurs with anxiety at a
ally observed have usually been free, or as free higher rate than expected and that psychopathic
as the general run of humanity, from real symp- youth experience anxiety symptomatology . . .”
toms of psychoneurosis” (Cleckley, 1976, p. It is possible that the determinants and subjec-
259; emphasis added). These statements would tive experiences of anxiety, and the way in
apply to the average person. They suggest “an which they are reported, are different in adoles-
absence of problematic anxiety, rather than a cent and adult psychopathy, but this would pres-
problematically low level of anxiousness, which ent several conceptual/developmental problems,
are really very different in their implications for some of which Kubak and Salekin (2009) ad-
personality disorder” (Hare, Neumann, & Wi- dress (see also Hemphala & Tengström, 2010).
diger, in press). Part of the difficulty lies in the almost exclusive
Sylvers, Lillienfeld, and LaPrairie (2011, p. use of self-reports in the investigation of psy-
134) define trait anxiety as “the persistent and chopathy and anxiety.
pervasive experience of state anxiety across sit-
uations,” and say that it is “unclear which, if Fearlessness
any, psychopathological syndromes [including
psychopathy] are characterized by low trait anx- Fearlessness was mentioned only once in the
iety.” The empirical literature (largely self- fifth edition of the Mask of Sanity (Cleckley,
report) on the topic certainly is ambiguous. In 1976, p. 319), in reference to a literary character
the development of the PCL-R (see Hare, 1991; about whom Cleckley wrote, “. . . his suicide,
Hare, 2003) measures of manifest trait anxiety his capacity to admit his misdeed with what
did not negatively covary with the other items impresses one as a measure of sincerity, his
being evaluated, and similar results have been warmth, and his depicted strength and fearless-
reported by other investigators who used a ness all stand out in contrast to the personality
Cleckley-based measure of psychopathy (Lo- patterns discussed in this book” (emphasis
ney, Taylor, Butler, & Iacono, 2007). These added). That is, Cleckley considered fearless-
findings are consistent with a literature indicat- ness to be counter indicative of psychopathy.
ing that most self-report measures of anxiety Sylvers et al. (2011, p. 134) define trait fear
and personal distress are related to negative as “the persistent and pervasive experience of
affect, and generally are only modestly associ- state fear across situations.” The idea that psy-
ated with the PCL-R (cf. Table 1 of Blonigen chopaths are low in trait fear, or are fearless,
et al., 2010). For example, Hale, Goldstein, stems in large part from an early laboratory
Abramowitz, Calamari, & Kosson (2004, p. study in which Lykken (1957) used the concepts
705), concluded that the PCL-R was unrelated of low fear arousal and conditioned fear to
to contemporary measures of anxiety and that account for the poor performance of psycho-
the “finding raises questions about traditional paths in classical (electrodermal) conditioning
conceptualizations of psychopathy that posit an and in passive avoidance learning paradigms.
attenuated capacity for anxiety.” In his analysis The model for linking these two paradigms was
of the issue, Dahl (1998) argued that Cleckley Mowrer’s (1947) two-stage theory of passive
(1976) “was wrong when he stated that psycho- avoidance learning in which anticipatory fear is
paths did not show manifest anxiety” (p. 298). an essential element in learning to avoid pun-
Complicating the matter further is the recent ishment (passive avoidance). There is a research
finding by Kubak and Salekin (2009) that, literature consistent with the suggestion that the
among male and female juvenile offenders, self- apparent social poise, sensation seeking, and
report measures of generalized anxiety disorder, difficulty in staying out of trouble that charac-
posttraumatic stress disorder, and separation terizes psychopathy may be interpreted in terms
PCL-R AND LAF 3

of low fear arousal. However, measures of fear- To evaluate these issues we conducted a se-
lessness typically are heavily laden with excite- ries of analyses of data from a representative
ment-seeking, sensation-seeking and impulsiv- national sample of violent Swedish male of-
ity, making it difficult to determine whether fenders who had been assessed both with the
psychopathy is associated with fearlessness per PCL-R and with clinical ratings of low trait
se or with impulsive disinhibition (Hare et al., in anxiety and trait fearlessness (referred to here as
press; Kubak & Salekin, 2010). Complicating LAF). The first step was to determine if the
the picture is evidence that psychopathy may be four-factor PCL-R model (Neumann, Hare, &
associated with a general attenuation of emo- Newman, 2007; see below) could be replicated
tional processes and not specifically with poor on this Swedish sample. The second step, and
fear conditioning (see Flor, Birbaumer, Her- the primary purpose of the study, was to deter-
mann, Ziegler, & Patrick, 2002; Hare, 2003; mine if the addition of LAF items alters the
Herpertz & Sass, 2000). Further, research by item-to-factor structural properties or factor in-
Newman and his colleagues (e.g., Newman, terpretations of the PCL-R, and if the PCL-R
Curtin, Bertsch, & Baskin-Sommers, 2010; items could substantially account for these
Zeier & Newman, in press) suggests that psy- traits. Note that some of the PCL-R data have
chopathic behaviors may be characterized less been used in two previous studies (Johansson &
by trait fearlessness than by idiosyncrasies in Kerr, 2005; Skeem, Johansson, Andershed,
attention that limit the processing of emotion- Kerr, & Louden, 2007), though none has in-
related cues associated with response modula- volved latent variable models of the PCL-R or
tion. For example, using a blink-startle para- LAF items.
digm Wolf et al. (2012, p. 108) concluded that
attentional blink attenuation “was correlated Method
with the affective-interpersonal features of psy-
chopathy tentatively [suggesting] that an atten- Participants
tion abnormality contributes to the expression
In Sweden all adult men who receive a sen-
of the affective and interpersonal symptoms of
tence of four years to life for violent crimes are
psychopathy as well as to their impulsive and
sent to the National Assessment Unit in Kumla
antisocial symptoms.” In their meta-analytic re-
Prison for psychological evaluation. These men
view of the literature, Sylvers et al. (2011, p.
are considered to be generally representative of
134) concluded that the argument by Lykken Swedish male, long-term violent offenders (see
and others “that psychopathy is characterized by Johansson & Kerr, 2005). For various reasons
low trait fear remains controversial.” not all offenders who pass through the Unit
actually receive a complete psychological as-
The Present Study sessment (e.g., occasional refusals to participate
in some evaluations, lack of resources due to
However one interprets the clinical and em- illness or vacation, etc.). Thus, while the total
pirical literature on the role of anxiety and fear sample consisted of 1,031 offenders, the sample
in the explication of psychopathy, the question Ns for most analyses the sample Ns varied from
under consideration here is whether or not the
PCL-R is compromised by its omission of items 1
There is evidence that PCL-R psychopaths with low or
that specifically measure these traits.1 Hare and high levels of self-reported anxiety may differ in perfor-
Neumann (2008; see also Hare et al., in press) mance of some laboratory tasks (e.g., Newman, MacCoon,
argued that both low trait anxiety and trait fear- Vaughn, & Sadeh, 2005; but see Zeier & Newman, in
lessness already are subsumed under PCL-R press). Some investigators have suggested that those with
low self-report anxiety scores are primary psychopaths,
items associated with more general emotional whereas those with high self- report anxiety scores are
experiences and processes, including lack of secondary psychopaths (e.g., Skeem et al., 2007), but this
remorse or guilt, shallow affect, callous/lack of mistakenly assumes that the role of anxiety in psychopathy
empathy, behaviors and dispositions reflecting is well-understood. The recent tendency to equate PCL-R
Factor 1 (F1) and Factor 2 (F2) with primary and secondary
impulsivity, irresponsibility, conning and ma- psychopathy, respectively, is unfortunate and misleading.
nipulation of others, sensation-seeking, and en- These topics are beyond the scope of the present paper and
gagement in overt antisocial behavior. will be addressed in detail elsewhere.
4 NEUMANN, HARE, AND JOHANSSON

398 to 989. The LAF items (see below) were scribed by Hare (2003; also see Hare & Neu-
recent additions to the clinical assessment bat- mann, 2008).
tery and only 291 offenders had complete Raters were six psychologists who completed
PCL-R and LAF files (no missing data). a two day PCL–R workshop and a Darkstone
postworkshop training course that included
Measures working through a minimum of eight standard-
ized videotaped cases. PCL-R data were avail-
Psychopathy Checklist-Revised (PCL-R).
The PCL-R is the most widely used interna- able for 973 offenders assessed during the pe-
tional measure for psychopathy. It is a 20-item riod 1997–2010. In an early analysis Johansson
clinical construct rating scale that uses a semi- and Kerr (2005) reported that the intraclass cor-
structured interview, case-history information, relation (ICC) for two sets of independent rat-
and specific scoring criteria to rate each item on ings (N ⫽ 32), using a two-way mixed effects
a 3-point scale (0, 1, 2) according to the extent model, was .84 for a single rating, and .91 for
to which it applies to a given person. Total the average of two ratings.
scores can vary from zero to 40 and reflect the Low Anxiety and Fearlessness (LAF).
degree to which the person matches the proto- Andershed, Douglas, and Skeem (2004) devel-
typical psychopathic person. This is in line with oped two research items for Low Anxiety and
strong evidence that, at the measurement level, Fearlessness (LAF) in order to look for “poten-
the construct underlying the PCL-R (and its tial subtypes/groups of inmates with PCL-
derivatives) is dimensional in nature rather than psychopathy” among referrals to the Swedish
taxonic (e.g., Guay, Ruscio, Knight, & Hare, national assessment center (Henrik Andershed,
2007). Evidence for the reliability and validity personal communication to R.D.H., October 28,
of the PCL-R as a measure of psychopathy is 2010). Scoring of these LAF items began in
extensive (Hare, 2003; Patrick, 2006). There is 2005, with data available for 398 violent of-
strong evidence (e.g., Neumann et al., 2007) fenders (81 were charged with a violent sexual
that the PCL-R measures a superordinate con- offense). We used the LAF items in this study to
struct underpinned by four correlated factors evaluate the role of low anxiety and fearlessness
(see Table 1). The Interpersonal/Affective di- in the PCL-R measurement of psychopathy. To
mensions and the Lifestyle/Antisocial dimen- our knowledge, there are no formal reports of
sions comprise, respectively, the original their use for this purpose. Briefly, Low Anxiety
PCL-R Factor 1 (F1) and Factor 2 (F2) de- is described in terms of a range of cognitive,

Table 1
Items and Factors in the Hare PCL-R
Interpersonal Affective
1. Glibness/superficial charm 6. Lack of remorse
2. Grandiose self-worth 7. Shallow affect
4. Pathological lying 8. Lack of empathy
5. Conning/manipulative 16. Will not accept responsibility

Lifestyle Antisocial
3. Need for stimulation 10. Poor behavioral controls
9. Parasitic lifestyle 12. Early behavioral problems
13. Lack of goals 18. Juvenile delinquency
14. Impulsivity 19. Revocation conditional release
15. Irresponsibility 20. Criminal versatility
Note. The items and factors are from the Hare Psychopathy Checklist-Revised, 2nd Edition
(2003, p. 83). Copyright, 2003 R.D. Hare and Multi-Health Systems, 3770 Victoria Park
Avenue, Toronto, Ontario, M2H 3M6. All rights reserved. Reprinted by permission. Note that
the item titles cannot be scored without reference to the formal criteria contained in the PCL-R
Manual. Two items, Promiscuous sexual behavior (Item 11), and Many short-term marital
relationships (Item 17), contribute to the Total PCL-R score but do not load on any factors.
PCL-R AND LAF 5

emotional, and behavioral features associated tional ways of assessing differences in the ␹2 fit
with the relative absence of anxiousness. Fear- statistic between nested models. For the models
lessness is defined primarily as engagement in a described below, all nested models (#’s 1– 6)
variety of risky behaviors, with little evidence were compared to a more elaborated model (#7)
of fear. Each item follows the PCL-R item- in which the LAF items were specified as indi-
description and scoring format. Raters used a cators for a separate LAF latent variable.
semistructured interview and a review of file The default in Mplus (v. 6.1) is to estimate
information to score each item on a 3-point latent models using all available data (including
scale according to specific criteria, summarized cases that have some missing values for the
as follows: 0 ⫽ the item does not apply; 1 ⫽ variables within the model) in accordance with
item may apply or applies in some respects/ missing data theory. Listwise deletion (using
conflicts between interview and file that cannot only cases that have a value on all variables) is
be resolved in favor of a 0 or 2; 2 ⫽ the item correct only under MCAR (missing completely
applies to the individual; a reasonably good at random), while MAR (missing at random) is
match in most essential respects.2 much more flexible than MCAR (Little & Ru-
Raters scored the LAF items after they had bin, 2002), even when MAR may be only ap-
scored the PCL-R items. Only single ratings proximately true (e.g., some offenders were
were available. administered the LAF items later in the assess-
ment battery). In most clinical data sets there is
Data Analytic Plan a good chance that missing cases are not miss-
ing at random (NMAR) and a function of many
Consistent with recent research on the latent unobserved variables, but MAR may still be a
relations between personality disorder and reasonable approximation (Muthén & Muthén,
symptoms of mental disorders (Eaton et al., 1998⫺2010). Therefore to remain in accor-
2011), the majority of analyses for the current dance with best practices for handling missing
study involved structural equation modeling data, we used all cases for our primary analyses
(SEM), given its methodological rigor (e.g., in Mplus. The proportion of missing values for
modeling error separate from common variance, the current study was examined by a covariance
unambiguous specification of item-to-factor re- coverage matrix, which provides an estimate of
lations), and capacity to provide evidence of available observations for each pair of vari-
construct validity (Strauss & Smith, 2009). ables. The minimum recommended coverage is
All model analyses were conducted with .10 (Muthén & Muthén, 1998⫺2010). In this
Mplus (Muthén & Muthén, 1998⫺2010), using study, coverage ranged from .39 to .99. Given
the robust weighted least squares estimation adequate coverage, the primary SEM analyses
procedure, given the ordinal nature of the were run on the larger sample of male offenders
PCL-R and LAF items. As recommended by Hu with available data (Ns ⫽ 973–989), though
and Bentler (1999), we used a two-index strat- parallel model analyses using only cases with
egy to assess model fit: The incremental Tuck- complete data (N ⫽ 291) revealed results that
er-Lewis index (TLI), and an absolute fit index, were substantively unchanged.
the Root Mean Square Error of Approximation
(RMSEA). Generally, TLI at or above .90 and
RMSEA at or below .08 is considered accept- Results
able model fit (Hoyle, 1995). To assess statisti-
cal differences in model fit we used the guide- Descriptive statistics for the sample were as
lines laid out by Cheung and Rensvold (2002). follows: mean age 32.89 (SD ⫽ 10.39), age of
If the incremental change in the comparative fit first crime 21.88 (SD ⫽ 10.34), and sentence in
index (⌬CFI) between a superordinate model years 5.92 (SD ⫽ 2.28). The most frequent
and a nested, more-constrained model is ⱕ .01, types of offenses committed were: Murder
then it is reasonable to hold that the two models
within the comparison do not differ statistically 2
The complete item descriptions and scoring instructions
in terms of fit. Based on an extensive simulation are available on request from P. Johansson
study, Cheung and Rensvold recommended that (petert.johansson@kriminalvarden.se) or H. Andershed
⌬CFI is more appropriate than the more tradi- (henrik.andershed@oru.se).
6 NEUMANN, HARE, AND JOHANSSON

(18.3%), Robbery (17.5%), Attempted Murder PCL-R and LAF Model Results
(16.2%), Aggravated Assault (8.6%), Man-
slaughter (5.9%), Aggravated Robbery (3.7%), We determined if the LAF items could be
and Arson (2.6%). Preliminary missing cases placed with one or other of the traditional two
analyses revealed no meaningful differences be- PCL-R factors (F1 ⫽ Interpersonal and Affec-
tween those with and without complete data in tive items; F2 ⫽ Lifestyle and Antisocial
terms of age, t(1021) ⫽ 1.69, p ⬎ .05, age of items). The results indicated good fit (Models 1
first crime, t(246) ⫽ ⫺0.75, p ⬎ .05, sentence and 2 in Table 2). There was little substantive
in years, t(927) ⫽ ⫺1.09, p ⬎ .05, Low Anxiety difference as a function of LAF item placement,
(␹2(2) ⫽ 0.42, p ⬎ .05), or Fearlessness though these two models were associated with
(␹2(2) ⫽ 1.85, p ⬎ .05). Those with versus CFI change values slightly greater than .01.
without complete data differed slightly on three This is likely due to the fact that four-factor
PCL-R facet scores, but this difference resulted PCL-R models had better model fit than the
in small effect sizes (eta2): Interpersonal (.01), two-factor model (see Table 2).
Affective (.02), Lifestyle (.01). To determine if the LAF items would “fit”
better with one or another of the four PCL-R
Four-Factor PCL-R Model factors, we conducted a series of SEMs, alter-
natively placing the LAF items with each one of
The mean PCL-R score for the full sample
these PCL-R factors. We examined model fit,
was 18.9 (SD ⫽ 10.3), with an alpha coefficient
the extent to which the LAF item loadings
of .92 and a mean interitem correlation (MIC)
changed as a function of their placement within
of .37. Mean PCL-R facet scores were as fol-
lows: Interpersonal 2.70 (SD ⫽ 2.48), Affec- a model, and change in CFI. As can be seen in
tive 4.60 (SD ⫽ 2.72), Lifestyle 5.59 Table 2 (Models 3– 6), model fit was excellent
(SD ⫽ 3.36), and Antisocial 4.60 (SD ⫽ 3.33). (TLI’s ⫽ .96⫺.97, RMSEA’s ⫽ .06⫺.07).
In a CFA we specified PCL-R items to load There was little change in the statistical (model
onto their respective factors (see Table 1): In- fit) or substantive (factor loading) results across
terpersonal (1, 2, 4, 5), Affective (6, 7, 8, 16), models, despite the fact that the LAF items were
Lifestyle (3, 9, 13, 14, 15) or Antisocial (10, 12, being tested with the different PCL-R factors.
18, 19, 20). The factors were allowed to corre- Most importantly, all CFI change values were
late freely in each CFA. The results indicated below the recommended cut-off of .01. These
excellent fit for the four-factor model (TLI ⫽ results suggest that each of the PCL-R factors
.97, RMSEA ⫽ .06). All factor loadings and shares some element of the features rated by the
factor correlations were significant ( ps ⬍ LAF items, and that the latter can go with any of
.05⫺.001). Figure 1 displays the items, the fac- the PCL-R factors.
tors they comprise, and the standardized model As shown in Table 2 and implied from the
parameters. CFI change results, Model 7, which involved
Alpha coefficients and MICs for the four the separate LAF factor represented by the low
PCL-R factors all exceeded .82 and .48, respec- anxiety and fearless items, (“five-factor model”)
tively, indicating good internal consistency and had excellent fit and the LAF item loadings
unidimensionality for each factor (cf. Schmitt, were large. Also noteworthy is that we found
1996; Smith, McCarthy, & Zapolski, 2009).3 excellent support (TLI ⫽ .97, RMSEA ⫽ .07)
for this model when we employed a relatively
LAF Items new modeling approach referred to as explor-
atory structural equation modeling (ESEM),
The mean Low Anxiety score was .68 which essentially involves estimation of model
(SD ⫽ .77); 18.2% of the offenders had a fit across a series of EFAs.
rating of 2. The mean Fearlessness score was
.80 (SD ⫽ .84); 26.9% of the offenders had a
3
rating of 2. The two LAF items were signif- The mean PCL-R score was somewhat lower for those
icantly correlated (r ⫽ .64, p ⬍ .001), and offenders with complete PCL-R and LAF data (17.2,
SD ⫽ 10.2), primarily because of the addition of sex of-
each item was significantly correlated with fenders to the sample when LAF data collection began.
each of the PCL-R facets (rs ⫽ .40⫺.60, ps ⬍ Alpha coefficients and MICs were virtually the same as
.001). those for the larger sample.
PCL-R AND LAF 7

Figure 1. Standardized Parameters: Four-factor Model Results (TLI ⫽ .97, RMSEA ⫽ .06).

Drawing on our prior research with the PCL-R PCL-R superordinate factor was able to predict
(Neumann et al., 2007), we formed a superordi- 73% of the variance of the LAF factor, suggesting
nate PCL-R factor by loading the four first-order the superordinate factor comprehensively ac-
factors onto a second-order “psychopathy” factor. counts for the LAF factor. The model results were
For this SEM the two LAF items served as indi- essentially the same when we used only cases with
cators for a separate LAF factor. The LAF factor complete data (69% of the variance in LAF pre-
was specified as a criterion variable and the su- dicted by the superordinate factor). Finally, we
perordinate factor as the predictor. Using the full examined which of the four first-order PCL-R
sample (Figure 2A), this model indicated that the factors was able to uniquely predict the LAF fac-

Table 2
Structural Equation Model Fit When Low Anxiety and Fearlessness (LAF) Items are Placed on Individual
PCL-R Factors (Models 1– 6) or as a Separate LAF Factor (Model 7)
Loading of LAF Items on PCL-R Factors
Model TLI RMSEA Low-Anxiety Fearlessness ⌬CFIa
Two-Factor Models
1. F1 .955 .078 .75 .85 .012
2. F2 .951 .081 .75 .83 .015
Four-Factor Models
3. Interpersonal .965 .069 .77 .88 .002
4. Affective .967 .067 .75 .85 .001
5. Lifestyle .964 .069 .75 .83 .003
6. Antisocial .960 .074 .77 .86 .007
“Five-Factor” Model
7. Four-factor ⫹ LAF .967 .066 .82 .93
Note. N ⫽ 989. F1 ⫽ Factor 1 (Interpersonal/Affective); F2 ⫽ Factor 2 (Lifestyle/Antisocial).
a
The ⌬CFI column represents the change in comparative fit index (CFI) of a given model (1– 6) compared to model #7.
The strategy for assessing statistical differences in model fit draws on the guidelines by Cheung and Rensvold (2002). Based
on an extensive simulation study, these authors recommended that ⌬CFI is more appropriate than the more traditional way
of assessing differences in the ␹2 fit statistic between nested models.
8 NEUMANN, HARE, AND JOHANSSON

Figure 2. PCL-R Superordinate Psychopathy Factor (A) and Individual PCL-R Factors (B)
as Predictors of LAF (Low Anxiety, Fearlessness).

tor, controlling for PCL-R factor overlap. The Discussion


results (Figure 2B) indicated that only the Inter-
personal (.54) and Antisocial (.42) factors Based on a large sample of violent Swedish
uniquely predicted the LAF factor (71% of the offenders, the results revealed excellent fit for
variance). The model analyses were substantively the four-factor PCL-R model, consistent with
unchanged when we used only those cases with the results of other studies in a variety of con-
complete (listwise) data. texts and diverse national samples (e.g., Hare &
PCL-R AND LAF 9

Neumann, 2008; Mokros et al., 2011; Neu- previously suggested (Hare & Neumann, 2008).
mann, 2007; Neumann et al., 2007; Vitacco, Moreover, a PCL-R superordinate factor was
Neumann, & Jackson, 2005; Walters, Wilson, able to account for most of the variance of a
& Glover, in press). Thus, the results provide separate LAF factor. It is possible that the
continued support for a PCL-based four-factor strong association between the PCL-R and LAF
model of the psychopathy construct. was due, in part, to a shared methodology, but it
In addition, consistent with past research also is possible that clinical ratings of LAF are
(Patrick, 2006), the four first-order PCL-R fac- more informative than self-reports, at least in
tors showed a pattern of differential (predictive) forensic settings.
associations with external variables, in this case The significant associations between the
the LAF factor (Figure 1b). This latter finding is PCL-R and LAF factors provide additional con-
important to consider. Specifically, while the struct validity for the PCL-R, and suggest that
four first-order factors can be accounted for by when assessed with a PCL-type format, low
a single superordinate factor (Neumann & Hare, anxiety and fearlessness may be part of the
2008; Neumann et al., 2007; Neumann, Kosson, larger psychopathy construct. Still, we suggest
Forth, & Hare, 2006), suggesting the PCL-R that psychopathy involves, and that the PCL-R
can be safely treated as a unidimensional mea- captures, a more general “dysregulation” of af-
sure, the evidence of differential associations of fective experience, not just limited to anxiety
the four factors with a range of external corre- and fear. In this regard, Sylvers and colleagues
lates should not be ignored.4 In other words, the (2011) suggested that trait anxiety and trait fear-
superordinate nature of the PCL-R should not lessness are largely distinct emotions. If this
be equated with a simplistic notion of a unidi- supposition is correct we would expect clinical
mensional psychopathy construct, as some ratings of fearlessness and low anxiety gener-
authors have suggested (Skeem, Polaschek, Pat- ally to be uncorrelated among psychopathic and
rick, & Lilienfeld, 2012); such a mistake in- other individuals. On the other hand, if psy-
volves both confusion between the measure and chopathy involves a general attenuation of
the construct it is designed to assess, and failure affective experience, then we would expect
to recognize that latent first-order PCL-based fearlessness and low anxiety to be positively
factors have differential associations with a correlated, and more importantly, we would ex-
range of external (etiologically relevant) corre- pect that the PCL-R should be linked to clinical
lates (e.g., Carré, Hyde, Neumann, Vidin, & ratings of both traits. The latter expectation was
Hariri, 2012; Neumann & Hare, 2008; Neu- met. Nevertheless, to more fully evaluate our
mann & Pardini, in press; Salekin, Neumann, “attenuation” hypothesis, a more wide-ranging
Leistico, & Zalot, 2004; Seara-Cardoso, Neu- and in-depth assessment of emotional experi-
mann, Roiser, McCrory, & Viding, 2012; Vi- ence will be necessary.
tacco, Neumann, & Caldwell, 2010; Vitacco et The use of a two-item factor in the current
al., 2005). Unidimensionality of a measure study only represented a narrow portion of
(PCL-R) does not necessarily imply unidimen- LAF. At minimum, latent variables generally
sionality of the construct it is designed to assess should have four to five items. Yet, even if we
(psychopathy). expand the LAF factor to a 5-item set, it is
Model fit also was excellent when the LAF possible that its associations with the PCL-R
items were brought into the model. Based on the
series of modeling results, our study addressed
two questions: “Is LAF separate from the other 4
The empirical modeling results for all of the PCL Scales
four aspects of PCL-R?” and “Is LAF subsumed provide evidence of scale unidimensionality, which is a
further strength of the PCL Scales that no doubt contributes
under the general PCL-based psychopathy con- to clear and robust findings with their use. See Smith et al.
struct?” The answer to these questions appears (2009) for an in-depth discussion of the strengths of such
to be “no” and “yes,” respectively. measures. At the same time, there are other psychopathy
The SEM analyses suggest that the two LAF scales (e.g., PPI-1 and PPI-2) that are the habitually treated
items may not “belong” with any particular as unidimensional measures, when in fact the evidence
clearly suggests this is inappropriate, given that they are
PCL-R factor. In other words, each of the multidimensional in nature (Neumann, Uzieblo, Grombez,
PCL-R factors (and their underlying items) taps & Hare, in press). This latter situation is likely to result in
aspects of low anxiety and fearlessness, as we ambiguous findings.
10 NEUMANN, HARE, AND JOHANSSON

and its factors would not be much different from items reflective of Antagonism (cf., PCL-R Factor 1) and
those described here. The addition of low anx- Disinhibition (cf., PCL-R Factor 2); they did not include
low anxiousness or fearlessness.
iety and fearlessness to the PCL-R may serve to
increase its face validity, but we doubt that it
would add very much to its incremental validity
(e.g., in neuroimaging, cognitive functioning, References
criminal justice, etc.).5 Nonetheless, this re- American Psychiatric Association. (June 21, 2011).
mains an empirical question. Meanwhile, the Personality disorders. Retrieved from http://
present results may have implications for the www.dsm5.org/PROPOSEDREVISIONS/Pages/
putative role of other traits and behaviors that PersonalityandPersonalityDisorders.aspx
some investigators believe should have been Andershed, H., Douglas, K., & Skeem, J. (2004).
included in the PCL-R assessment of psychop- Clinical low anxiety, fearlessness item ratings
athy. If, as we contend, the extant PCL-R items (Version 1.0). Örebro, Sweden: Department of Be-
account for low anxiety and fearlessness it is havioral, Social, and Legal Sciences, Örebro Uni-
possible that they also account for other features versity, SE-701 82.
Blonigen, D. M., Patrick, C. J., Douglas, K. S.,
of psychopathy, such as dominance, lack of
Poythress, N. G., Skeem, J. L., Lilienfeld, S. O.,
insight, and so forth. That is, the PCL-R item set . . . Krueger, R. F. (2010). Multimethod assess-
may be more comprehensive than some com- ment of psychopathy in relation to factors of in-
mentators claim. ternalizing and externalizing from the Personality
There are several limitations to the present Assessment Inventory: The impact of method vari-
research. First, we had only single ratings for ance and suppressor effects. Psychological Assess-
the LAF items, and new research is needed to ment, 22, 96 –107. doi:10.1037/a0017240
determine whether or not their interrater reli- Carré, J., Hyde, L., Neumann, C. S., Viding, E., &
ability is similar to that of the PCL-R items. Hariri, A. (2012). The Neural signatures of distinct
Second, raters scored the LAF items after they psychopathic traits. Manuscript under review.
Cheung, G. W., & Rensvold, R. B. (2002). Evaluat-
had scored the PCL-R items, with the result that
ing goodness-of-fit indexes for testing measure-
the LAF scores may have been influenced by ment invariance. Structural Equation Modeling, 9,
impressions formed by the raters while they 233–255. doi:10.1207/S15328007SEM0902_5
scored the PCL-R items. This issue is not con- Cleckley, H. (1941). The mask of sanity. St. Louis,
fined to the LAF items, but is inherent in the MO: Mosby.
protocol for scoring the PCL-R. In theory, Cleckley, H. (1976). The mask of sanity (5th edition).
though rather impractical, it would be possible St. Louis, MO: Mosby.
to have each PCL-R item scored by a different Dahl, A. A. (1998). Psychopathy and psychiatric
rater. With respect to the LAF items, investiga- comorbidity. In T. Millon, E. Simonson, M. Bir-
tors could add independently scored LAF items ket-Smith, & R. D. Davis (Eds.), Psychopathy:
to their existing PCL-R data sets. This would Antisocial, criminal, and violent behavior (pp.
291–303). New York, NY: Guilford Press.
help to clarify the role of low anxiety and fear- Eaton, N. R., Krueger, R. F., Keyes, K. M. Skodol,
lessness in the psychopathy construct. More- A. E., Markon, K. E. Grant, B. F., & Hasin, D. S.
over, it would be helpful in future research to (2011). Borderline personality disorder co-
investigate the potential incremental validity of morbidity: Relationship to the internalizing–
the LAF items relative to the PCL-R facets in externalizing structure of common mental disor-
predicting critical correlates of psychopathy. ders. Psychological Medicine, 41, 1041–1050. doi:
10.1017/S0033291710001662
5
It is worth noting that the DSM-V Work Group has not Flor, H., Birbaumer, N., Hermann, C., Ziegler, S., &
included traits of fearlessness or low anxiousness in the Patrick, C. (2002). Aversive Pavlovian condition-
DSM-V diagnosis of antisocial personality disorder/ ing in psychopaths: Peripheral and central corre-
dissocial personality disorder (American Psychiatric Asso- lates. Psychophysiology, 39, 505–518. doi:
ciation, 2011; also see Hare et al., in press, for a historical 10.1111/1469-8986.3940505
overview of PCL-R psychopathy and the DSM criteria for Forth, A. E., Kosson, D., & Hare, R. D. (2003). The
antisocial personality disorder). Further, in their expert con-
sensus study, Samuel, Lynam, Widiger, and Ball (2012, p.
Hare PCL: Youth version. Toronto, Canada:
6) concluded, “The Work Group and the experts surveyed in Multi-Health Systems.
the current study appeared to be in complete agreement Frick, P. J., & Hare, R. D. (2001). The antisocial
about traits that should and should not be used to diagnose process screening device. Toronto, Canada: Multi-
the antisocial/psychopathic type.” The included traits were Health Systems.
PCL-R AND LAF 11

Guay, J. P., Ruscio, J., Knight, R. A., & Hare, R. D. ment, 31, 271–284. doi:10.1007/s10862-009-
(2007). A taxometric analysis of the latent struc- 9144-2
ture of psychopathy: Evidence for dimensionality. Little, R. J. A., & Rubin, D. B. (2002). Statistical
Journal of Abnormal Psychology, 116, 701–716. analysis with missing data (2nd ed.). New York,
doi:10.1037/0021-843X.116.4.701 NY: John Wiley.
Hale, L. R. Goldstein, D. S., Abramowitz, C. S., Loney, B. R., Taylor, J., Butler, M. A., & Iacono,
Calamari, J. E., & Kosson, D. S. (2004). Psychop- W. G. (2007). Adolescent psychopathy features:
athy is related to negative affectivity but not to 6-year temporal stability and the prediction of ex-
anxiety sensitivity. Behaviour Research and Ther- ternalizing symptoms during the transition to
apy, 42, 697–710. doi:10.1016/S0005- adulthood. Aggressive Behavior, 33, 242–252. doi:
7967(03)00192-X 10.1002/ab.20184
Hare, R. D. (1965). Temporal gradient of fear arousal Lykken, D. T. (1957). A study of anxiety in the
in psychopaths. Journal of Abnormal Psychol- sociopathic personality. Journal of Abnormal and
ogy, 70, 442– 445. doi:10.1037/h0022775 Social Psychology, 55, 6 –10. doi:10.1037/
Hare, R. D. (1991). The Hare Psychopathy Checklist- h0047232
revised. Toronto, ON: Multi-Health Systems. Mokros, A., Neumann, C. S., Stadtland, C., Oster-
Hare, R. D. (2003). The Hare Psychopathy Checklist- heider, M., Nedopil, N., & Hare, R. D. (2011).
revised (2nd ed.). Toronto, ON: Multi-Health Sys- Assessing measurement invariance of PCL-R as-
tems. sessments from file reviews of North American
Hare, R. D., & Neumann, C. S. (2008). Psychopathy and German offenders. International Journal of
as a clinical and empirical construct. Annual Re- Law and Psychiatry, 34, 56 – 63. doi:10.1016/
view of Clinical Psychology, 4, 217–246. doi: j.ijlp.2010.11.009
10.1146/annurev.clinpsy.3.022806.091452 Mowrer, O. H. (1947). On the dual nature of learn-
Hare, R. D., Neumann, C. S., & Widiger, T. A. (in ing—A reinterpretation of “conditioning” and
press). Psychopathy. In T. A. Widiger (Ed.), The “problem-solving.” Harvard Educational Re-
Oxford handbook of personality disorders. New view, 17, 102–148.
York, NY: Oxford University Press. Muthén, L. K., & Muthén, B. O. (1998 –2010). Mplus
Hemphala, M., & Tengström, A. (2010). Associa- user’s guide (6th ed.). Los Angeles, CA: Muthén
tions between psychopathic traits and mental dis- & Muthén.
orders among adolescents with substance use prob- Neumann, C. S. (2007). Psychopathy. British Journal
lems. British Journal of Clinical Psychology, 49, of Psychiatry, 191, 357–358.
109 –122. doi:10.1348/014466509X439216 Neumann, C. S., & Hare, R. D. (2008). Psychopathic
Herpertz, S. C., & Sass, H. (2000). Emotional defi- traits in a large community sample: Links to vio-
ciency and psychopathy. Behavioral Sciences and lence, alcohol use, and intelligence. Journal of
the Law, 18, 567–580. doi:10.1002/1099- Consulting and Clinical Psychology, 76, 893– 899.
0798(200010)18:5⬍567::AID-BSL410⬎3.0.CO; doi:10.1037/0022-006X.76.5.893
2-8 Neumann, C. S., Hare, R. D., & Newman, J. P.
Hervé, H. F. (2007). Psychopathy across the ages: A (2007). The super-ordinate nature of the Psychop-
history of the Hare psychopath. In H. F. Hervé & athy Checklist-revised. Journal of Personality Dis-
J. C. Yuille (Eds.), The psychopath: Theory, re- orders, 21, 102–117. doi:10.1521/pedi.2007
search and practice. Mahwah, NJ: Lawrence Erl- .21.2.102
baum Associates. Neumann, C. S., Kosson, D. S., Forth, A. E., & Hare,
Hoyle, R. H. (1995). Structural equation modeling: R. D. (2006). Factor structure of the Hare Psychop-
Concepts, issues, and applications. Thousand athy Checklist: Youth version (PCL: YV) in incar-
Oaks, CA: Sage. cerated adolescents. Psychological Assess-
Hu, L., & Bentler, P. M. (1999). Cut-score criteria for ment, 18, 142–154. doi:10.1037/1040-3590.18
fit indexes in covariance structure analysis: Con- .2.142
ventional criteria versus new alternatives. Struc- Neumann, C. S., & Pardini, D. (in press). Factor
tural Equation Modeling, 6, 1–55. doi:10.1080/ structure and construct validity of the Self-Report
10705519909540118 Psychopathy (SRP) Scale and the Youth Psycho-
Johansson, P., & Kerr, M. (2005). Psychopathy and pathic Traits Inventory (YPI) in young men. Jour-
intelligence: A second look. Journal of personality nal of Personality Disorders.
disorders, 19, 357–369. doi:10.1521/ Neumann, C. S., Uzieblo, K., Grombez, G., & Hare,
pedi.2005.19.4.357 R. D. (in press). Understanding the Psychopathic
Kubak, F. A., & Salekin, R. T. (2009). Psychopathy Personality Inventory (PPI) in terms of the unidi-
and anxiety in children and adolescents: New in- mensionality, orthogonality, and construct validity
sights on developmental pathways to offending. of PPI-I and -II. Personality Disorders: Theory,
Journal of Psychopathology and Behavior Assess- Research, and Treatment.
12 NEUMANN, HARE, AND JOHANSSON

Newman, J. P., Curtin, J. J., Bertsch, J. D., & Baskin- Public Interest, 12, 95–162. doi:10.1177/
Sommers, A. R. (2010). Attention moderates the 1529100611426706
fearlessness of psychopathic offenders. Biological Smith, G. T., McCarthy, D. M., & Zapolski, T. C.
Psychiatry, 67, 66 –70. doi:10.1016/j.bio- (2009). On the value of homogeneous constructs
psych.2009.07.035 for construct validation, theory testing, and the
Newman, J. P., MacCoon, D. G., Vaughn, L. J., & description of psychopathology. Psychological As-
Sadeh, N. (2005). Validating a distinction between sessment, 21, 272–284. doi:10.1037/a0016699
primary and secondary psychopathy with measures Strauss, M., & Smith, G. (2009). Construct validity:
of Gray’s BIS and BAS constructs. Journal of Advances in methodology. Annual Review of Clin-
Abnormal Psychology, 114, 319 –323. doi: ical Psychology, 5, 1–25. doi:10.1146/annurev
10.1037/0021-843X.114.2.319 .clinpsy.032408.153639
Patrick, C. J. (Ed.). (2006). Handbook of psychopa- Sylvers, P., Lilienfeld, S. O., & LaPrairie, J. L.
thy. New York, NY: Guilford Press. (2011). Differences between trait fear and trait
Salekin, R. T., Neumann, C. S., Leistico, A. M., & anxiety: Implications for psychopathology. Clini-
Zalot, A. A. (2004). Psychopathy in youth and cal Psychology Review, 132, 122–137. doi:
intelligence: An investigation of Cleckley’s Hy- 10.1016/j.cpr.2010.08.004
pothesis. Journal of Clinical Child and Adolescent Vitacco, M. J., Neumann, C. S., & Caldwell, M. F.
Psychology, 33, 731–742. doi:10.1207/ (2010). Predicting antisocial behavior in high-risk
s15374424jccp3304_8 male adolescents: Contributions of psychopathy
Samuel, D. B., Lynam, D. R., Widiger, T. A., & Ball, and instrumental violence. Criminal Justice and
S. A. (2012). An expert consensus approach to Behavior, 37, 833– 846. doi:10.1177/
relating the proposed DSM-5 types and traits. Per- 0093854810371358
sonality Disorders: Theory, Research, and Treat- Vitacco, M. J., Neumann, C. S., & Jackson, R. L.
ment, 3, 1–16. doi:10.1037/a0023787 (2005). Testing a four-factor model of psychopa-
Schmitt, N. (1996). Uses and abuses of coefficient thy and its association with ethnicity, gender, in-
alpha. Psychological Assessment, 8, 350 –353. doi: telligence, and violence. Journal of Consulting and
10.1037/1040-3590.8.4.350 Clinical Psychology, 73, 466 – 476. doi:10.1037/
Seara-Cardoso, A., Neumann, C. S., Roiser, J., Mc- 0022-006X.73.3.466
Crory, E., & Viding, E. (2012). Investigating as- Walters, G. D., Wilson, N. J., & Glover, A. J. J. (in
sociations between empathy, morality and psycho- press). Predicting recidivism with the Psychopathy
pathic personality traits in the general population. Checklist: Are factor score composites really nec-
Personality and Individual Differences, 52, 67–71. essary? Psychological Assessment.
doi:10.1016/j.paid.2011.08.029 Wolf, R. C., Carpenter, R. W., Warren, C. M., Zeier,
Skeem, J., Johansson, P., Andershed, H., Kerr, M., & J. D., Baskin-Sommers, A. R., & Newman, J. P.
Louden, J. E. (2007). Two subtypes of psycho- (2012). Reduced susceptibility to the attentional
pathic violent offenders that parallel primary and blink in psychopathic offenders: Implications for
secondary variants. Journal of Abnormal Psychol- the attention bottleneck hypothesis. Neuropsychol-
ogy, 116, 395– 409. doi:10.1037/0021-843X.116 ogy, 26, 102–109. doi:10.1037/a0026000
.2.395 Zeier, J. D., & Newman, J. P. (in press). Both
Skeem, J. L., Polaschek, L. L., Patrick, C. J., & self-report and interview-based measures of psy-
Lilienfeld, S. O. (2012). Psychopathic personal- chopathy predict attention abnormalities in
ity: Bridging the gap between scientific evidence criminal offenders. Assessment. doi:10.1177/
and public policy. Psychological Science in the 1073191111415364

View publication stats

Vous aimerez peut-être aussi