Você está na página 1de 6

Psychological Assessment

2006, Vol. 18, No. 2, 225230

Copyright 2006 by the American Psychological Association


1040-3590/06/$12.00 DOI: 10.1037/1040-3590.18.2.225

Examining the Utility of the PCL:SV as a Screening Measure Using


Competing Factor Models of Psychopathy
Laura S. Guy and Kevin S. Douglas
Simon Fraser University
The correspondence between the Hare Psychopathy Checklist: Screening Version (PCL:SV; S. D. Hart,
D. N. Cox, & R. D. Hare, 1995) and the Hare Psychopathy ChecklistRevised (PCLR; R. D. Hare,
1991, 2003) was examined in forensic (N 175) and correctional (N 188) samples. Intermeasure
correlations for Total scores (.95 forensic, .94 correctional) and the original 2-factor, D. J. Cooke and C.
Michies (2001) 3-factor, and R. D. Hares (2003) 4-facet models (range .87.95) were high. Area
under the curve values for the PCL:SV were .98 in both samples (cutoff PCLR Total score of 25).
The PCL:SV performed well as a screen, maximizing false positive relative to false negative errors. Close
correlations for prediction of violent recidivism in the correctional sample were obtained for the PCLR
(.42) and PCL:SV (.37). Results indicate the robust relation between the measures is maintained whether
they are completed on the basis of file review only or file plus interview and whether the same or different
raters score the measures.
Keywords: PCL:SV, PCLR, psychopathy factor models, scoringassessment procedures

alcohol abuse (Hart, Hare, & Forth, 1994; Hill, Rogers, & Bickford, 1996), treatment noncompliance (Hill et al., 1996), and
antisocial behavior and aggression (Belfrage, Fransson, & Strand,
2000; Belfrage & Rying, 2004; Douglas, Ogloff, Nicholls, &
Grant, 1999; Doyle, Dolan, & McGovern, 2002; Hill et al., 1996;
Kropp & Hart, 2000; Monahan et al., 2001; Skeem & Mulvey,
2001; Steadman et al., 2000; Strand, Belfrage, Fransson, &
Levander, 1999). It also relates strongly to antisocial personality
disorder (Howard, Payamal, & Neo, 1997) and bears conceptually
meaningful relationships with personality tests measuring constructs such as arrogant calculating, or cold-heartedness (Hart &
Hare, 1994). Item response theory analyses of the PCL:SV compelled Cooke, Michie, Hart, and Hare (1999) to conclude that the
PCL:SV can be considered a short or parallel form of the PCLR
(p. 11).
Recent structural analyses have suggested that either a three- or
a four-factor model may fit the data for the PCLR and PCL:SV
more adequately than does the original two-factor model. These
developments have implications for the PCL:SV as a screen and
parallel measure, especially at the factor level. Cooke and Michie
(2001) concluded that a superordinate psychopathy factor is underpinned by three distinct, lower order factors. In this model,
items from the original Factor 1 load on two separate factors:
Arrogant and Deceitful Interpersonal Style (glibness/superficial
charm, grandiosity, pathological lying, conning/manipulativeness)
and Deficient Affective Experience (shallow affect, lacks empathy, lacks remorse, failure to accept responsibility). The third
factor, Impulsive and Irresponsible Behavioral Style, comprises a
subset of items from the original Factor 2 (need for stimulation,
impulsivity, irresponsibility, parasitic lifestyle, lacks realistic
goals). Seven PCLR items are not included in this model.
Hare (2003) subsequently proposed a four-factor model in
which two factors each consist of two facets (which can be
interpreted as nested factors). The Interpersonal/Affective factor

The Hare Psychopathy Checklist: Screening Version (PCL:SV;


Hart, Cox, & Hare, 1995) was developed as a relatively quick and
inexpensive way of assessing psychopathic traits in offenders and
forensic patients (Hart et al., 1995, p. 1). The PCL:SV conceptualizes psychopathy analogously to its longer counterpart, the
Hare Psychopathy ChecklistRevised (PCLR; Hare, 1991,
2003). Both the PCL:SV and PCLR (1991 version) divided
psychopathy into two components: affective/interpersonal (e.g.,
callousness, remorselessness, grandiosity, deceitfulness) and socially deviant/impulsive lifestyle (e.g., criminal behavior, lack of
goals, juvenile delinquency, poor behavior controls). The PCL:SV
termed these Part 1 and Part 2, intending them to be the conceptual analogues of the PCLRs empirically derived Factor 1 and
Factor 2. Initial validation studies revealed high correspondence
between the PCL:SV and PCLR, with rs of .80, .68, and .81
between Total, Part/Factor 1, and Part/Factor 2 scores, respectively, across five samples (Hart et al., 1995).
A fair amount of research suggests that the PCL:SV relates to
external constructs in much the same way as the PCLR. For
instance, it correlates with adverse outcomes such as drug and

Laura S. Guy and Kevin S. Douglas, Department of Psychology, Simon


Fraser University, British Columbia, Canada.
Laura S. Guy is funded by the Social Sciences and Humanities Research
Council of Canada. We gratefully acknowledge the Correctional Service of
Canada and the BC Forensic Psychiatric Services Commission for providing the resources necessary to complete this research. However, the views
expressed are those of the authors alone. We are also grateful to Christopher D. Webster, Derek Eaves, and Stephen Hart, who were in receipt of
a grant from the British Columbia Health Research Foundation that funded
some of the data collection in the forensic sample.
Correspondence concerning this article should be addressed to Laura S.
Guy, Department of Psychology, Simon Fraser University, RCB 5246,
8888 University Drive, Burnaby, British Columbia V5A 1S6, Canada.
E-mail: lguy@sfu.ca
225

BRIEF REPORTS

226

comprises the Interpersonal and Affective facets, and the Social


Deviance factor comprises the Lifestyle and Antisocial facets. The
first of these three facets are identical in item content to Cooke and
Michies (2001) three factors. The Antisocial facet reintegrates
into the model five items (poor behavioral controls, early behavioral problems, juvenile delinquency, revocation of conditional
release, criminal versatility) that relate to antisocial behavior that
are not part of Cooke and Michies three-factor model.
Cooke and Michie (2001) concluded that a 9-item, three-factor
PCL:SV analogue of their three-factor PCLR model best fit both
data and theory. Skeem, Mulvey, and Grisso (2003) also reported
that a three-factor PCL:SV model fit the MacArthur data (Monahan et al., 2001) better than the original two-factor model, and
other researchers also have provided support for the three-factor
model using the PCL:SV (Cooke, Kosson, & Michie, 2001; Johansson, Andershed, Kerr, & Levander, 2002). Conversely, additional research with the PCL:SV has concluded that a four-factor
model fits as well as or better than the three-factor model (Hill,
Neumann, & Rogers, 2004; Vitacco, Neumann, & Jackson, 2005).
In light of these recently developed and competing three- and
four-factor models of psychopathy, it is relevant to test whether the
PCL:SV remains strongly related to the PCLR at the factor level of
the different models. In two samples totaling 363 offenders and
forensic patients, we investigated the overall and factor correspondence (using the two-, three- and four-factor models) between the
measures and, for one sample, how well the PCL:SV and PCLR (and
their new factors) each predicted violent criminal recidivism.
Further, in an attempt to increase the relevance of our analyses
to the broader psychopathy research literature and to clinical
assessment, we provided estimates of PCLR/PCL:SV correspondence using both file-only ratings (a common research rating
procedure) and file interview coding procedures on a subset of
cases. Previous research that has compared these coding procedures has focused on the PCLR alone (Alterman, Cacciola, &
Rutherford, 1993; Bolt, Hare, Vitale, & Newman, 2004; Grann,
Lngstrom, Tengstrom, & Stlenheim, 1998; Serin, 1993; Wong,
1988). Though indicating reasonable correspondence between
methods (with lower scores typically emerging from file-only
ratings; cf. Serin, 1993),1 to our knowledge, this research has not
included an analysis of the scoring method issue with respect to the
PCL:SV, or the correspondence between the PCLR and PCL:SV.
Further, for a subset of cases we were able to include ratings made
by independent raters as well as the same raters on both the PCLR
and PCL:SV. This feature permits an estimation of the range of
correlations between measures depending on whether the raters are
the same or different. It is relevant to research in which either the
same or different raters might be used for the two measures, and to
applied assessment contexts in which the full PCLR may be
completed either by the same or different assessors after the
PCL:SV has been completed.

Method
Participants
Sample 1 comprised 188 male offenders sentenced to at least 2 years
imprisonment in Western Canada. To ensure a reasonable base rate of
violence, we selected participants for inclusion on the basis of known
outcome status. Participants in the violent reoffender group (n 93) were
selected randomly from offenders known to have been reconvicted for a

violent offense (including sexual) that resulted either in a federal or a


provincial sentence following their release dates. Participants in the group
without violent reconvictions (n 95) were selected randomly from those
who were released during the same time period but who did not have a new
violent offense following release. The average age was 38.25 years (SD
10.84). Almost all had a history of violent offenses (98%), and most were
Caucasian (73%; 19% Aboriginal, 8% Other).
Sample 2 comprised 175 consecutive persons coming before a criminal
Review Board for release from dispositions of Not Criminally Responsible
on Account of Mental Disorder in a maximum-security forensic hospital in
Western Canada. The average age at admission was 33 years (SD 9.4).
Most had a violent index offense (86%), were male (88%), and had
schizophrenia (55%).

Measures
The two measures used were the PCLR (Hare, 1991, 2003) and
PCL:SV (Hart et al., 1995). The PCLR has 20 items that are rated 0 (not
present), 1 ( possibly present), or 2 (definitely present) and can range from
0 to 40. The PCL:SV has 12 items with the same item scoring and a score
range of 0 to 24. Some PCL:SV items are conceptual integrations of more
than one PCLR item, and others focus on a similar feature (i.e., antisocial
behavior) but have different coding procedures. The commonly used score
cutoffs for classifications of psychopathy are 30 and 18 on the PCLR and
PCL:SV, respectively.

Measure Reliabilities
For Sample 1, interrater reliability (ICC1), based on 28 cases, was good:
PCLR Total (.94); PCL:SV Total (.83). For Sample 2, interrater reliability
(n 30) also was good: PCLR Total (.91); PCL:SV Total (.86). Coefficient alpha in Sample 1 was acceptable for the PCLR (.84) and PCL:SV
(.83) Total scores, as it was in Sample 2 for both the PCLR (.82) and
PCL:SV (.84) Total scores.

Procedure
For participants in the offender sample (Sample 1), scores on the PCLR
and PCL:SV were based on file information only. Raters (n 3) were
graduate-level research assistants in the employ of the Correctional Service
of Canada (CSC) who completed the measures for research purposes. They
were trained by CSC personnel. For 108 (61.7%) participants in the
forensic psychiatric sample (Sample 2), scores on both measures were
based on file review and interview, whereas the remaining participants
scores (n 67; 38.3%) were based on file information only. Raters (n
4) in the forensic sample were graduate-level research assistants who
completed the PCLR and PCL:SV for research purposes. They were
trained by Stephen Hart. In both samples, file information used to complete
the measures included all documents contained in participants clinical and
legal files, which are very comprehensive and detailed. There were no
instructions given regarding the order in which the PCLR or PCL:SV
were to be completed. Raters were blind to participant recidivism status.
In both samples, most PCLR and PCL:SV ratings were made by the
same raters, which has two implications. First, error due to raters is
eliminated from the PCLR/PCL:SV comparisons. Second, correlations
between instruments should be considered upper estimates of their degree
of relationship, and, in conjunction with independent ratings, can provide
a range of relationship between measures. In order to provide an estimate

Alterman et al. (1993) found that increasing amounts of rater information resulted in higher PCLR Total and factors scores. However, these
researchers examined this issue by comparing interview only versus interview increasing amounts of file information.

BRIEF REPORTS

227

using independent ratings were as follows: Total score (.85 and


.89), Factor 1 (.87 and .74), Factor 2 (.85 and .82), Cooke and
Michies Arrogant and Deceitful Interpersonal Style (.89 and .60),
Cooke and Michies Deficient Affective Experience (.86 and .70),
Cooke and Michies Impulsive and Irresponsible Behavioral Style
(.87 and .82), and Hares Antisocial facet (.75 and .79).

of measure correspondence based on independent ratings, we used the


interrater reliability cases from each sample. This provided 58 pairs of
independent PCLR/PCL:SV ratings. In fact, it provided two sets of
PCLR/PCL:SV correlations for each of the 58 cases (i.e., for any given
case, Rater 1 PCLR with Rater 2 PCL:SV, and Rater 1 PCL:SV with
Rater 2 PCLR). We present the average of these two correlations in the
findings reported below.

Diagnostic Efficiency Analyses

Outcomes

The samples were divided into dichotomous groups (psychopathy vs. nonpsychopathy) on the PCLR and PCL:SV using Total
scores of 30 and 18, respectively, as cutoff scores. For the forensic
sample, this strategy yielded an overall classification hit rate of
94%. Sensitivity was perfect (i.e., all 9 participants with PCLR
scores of 30 and above had PCL:SV scores of 18 and above), and
specificity was 94%. Positive predictive power (PPP) and negative
predictive power (NPP) were .47 and 1, respectively. The PCL:SV
also performed well in the correctional sample, with a hit rate of 89%,
sensitivity of 90%, specificity of 89%, and PPP of .53 and NPP of .99.
Relatively low base rates of psychopathy were observed in both
samples using a cutoff score of 30 on the PCLR. Consequently,
and in light of Hares (2003) suggestion of some descriptive cutoff
scores, where 25 would be considered high on the PCLR, we
recomputed the analyses using a PCLR Total score of 25 as being
indicative of psychopathy. This strategy resulted in a slight improvement in overall hit rates. Utility estimates in the forensic
sample were as follows: hit rate (95%), sensitivity (72%), specificity (99%), PPP (.95), and NPP (.96). Values in the correctional
sample were as follows: hit rate (92%), sensitivity (73%), specificity (98%), PPP (.92), and NPP (.92).

Recidivism data are available for Sample 1 over the course of 7.7 years
(on average; range 6 11 years). Recidivism was defined as reconviction
and reincarceration for a violent offense.

Results
Intercorrelations of Measures
Table 1 presents the intercorrelations between the PCL indices
for both samples using the same raters for both the PCLR and
PCL:SV. Correlations between PCL:SV and PCLR Total scores
were very high (r .95 forensic; r .94 correctional). Correlations between Factors 1 4 on the PCLR with their derived
counterparts on the PCL:SV also were very strong, ranging from
.87 to .92 (forensic) and .88 to .95 (correctional).
Expectedly, correlations based on independent ratings were
lower, in some cases fairly substantially. This may be accounted
for in part by the smaller samples on which the independent ratings
were based (i.e., ns 28 and 30, respectively, for the correctional
and forensic samples). For the total and original factors, the
intermeasure correlations were reduced generally from the .90s to
the .80s. For Cooke and Michies (2001) first two factors (interpersonal and affective deficits), the reduction in correlations for
independent ratings was greater (to .60 .70) in one of the two
subsamples. For the correctional and forensic samples, respectively, the correlation between the PCLR and PCL:SV indices

Receiver Operating Characteristic (ROC) Analyses


ROC analyses were completed for each sample using PCLR
Total scores of 25 and 30 as cutoffs to form dichotomous group-

Table 1
Intercorrelations Between the Hare Psychopathy ChecklistRevised (PCLR) Indices in the
Correctional and Forensic Samples
Measure
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.

PCLR
Factor 1
Factor 2
C&M 1
C&M 2
C&M 3
Hare Facet 4
PCL:SV
Part 1
Part 2
C&M 1
C&M 2
C&M 3
Hare Facet 4

10

11

12

13

14

.80
.89
.71
.70
.86
.81
.94
.78
.87
.68
.71
.83
.78

.82

.46
.89
.88
.47
.39
.83
.96
.49
.86
.85
.49
.43

.90
.55

.41
.40
.95
.90
.80
.45
.94
.38
.44
.90
.85

.65
.77
.41

.59
.44
.33
.73
.85
.41
.95
.56
.42
.32

.69
.85
.48
.33

.41
.37
.74
.85
.46
.58
.93
.48
.42

.82
.53
.89
.45
.42

.74
.78
.46
.90
.40
.43
.94
.74

.75
.40
.82
.24
.40
.49

.72
.39
.85
.30
.39
.71
.88

.95
.86
.82
.67
.73
.76
.69

.86
.85
.74
.79
.80
.77

.81
.94
.56
.78
.75
.54
.42
.89

.49
.89
.89
.49
.43

.89
.60
.91
.42
.55
.81
.79
.90
.61

.38
.51
.93
.92

.62
.74
.40
.92
.33
.43
.26
.68
.82
.41

.58
.39
.30

.71
.81
.52
.38
.90
.47
.43
.78
.83
.58
.38

.48
.47

.78
.54
.81
.41
.47
.89
.49
.77
.53
.86
.38
.50

.73

.78
.55
.80
.36
.52
.54
.87
.81
.56
.89
.36
.54
.50

Note. Values for the correctional and forensic samples are presented to the left and the right, respectively, of
the center diagonal. Correlations between each PCLR index and its PCL:SV counterpart are in boldface. C&M
Cooke and Michie, 2001; C&M 1 Arrogant and Deceitful Interpersonal Style; C&M 2 Deficient
Affective Experience; C&M 3 Impulsive and Irresponsible Behavioral Style; PCL:SV Hare Psychopathy
Checklist: Screening Version. C&M 1, 2, and 3 are identical in item content to Hares (2003) first three factors.
All correlations were significant at p .01.

BRIEF REPORTS

228

ings of psychopathy and nonpsychopathy, and the PCL:SV


Total score as the predictor of these groups. ROC analysis can be
used to estimate predictive accuracy when there is a continuous
predictor and a dichotomous outcome. This analysis does not
consider base rate information and therefore is useful when the
outcome of interest has a relatively low base rate (Amenta, Guy, &
Edens, 2003; Mossman & Somoza, 1991). ROC analysis yields an
effect size called the area under the curve (AUC). The AUC is
the probability that a randomly chosen person who scores positive
on the dependent measure (in this case, is classified as having
psychopathy according to the PCLR) will score higher than a
person who does not have psychopathy on the predictor measure
(Mossman & Somoza, 1991). AUC values of approximately .71
are equal to a Cohens d of 0.80 (Dunlap, 1999), indicative of a
large statistical effect size. Using 30 on the PCLR as the cutoff
score in the forensic sample, the AUC value for the PCL:SV was
.90 ( p .001; 95% confidence interval [CI] .971.0). Considering 25 as the cutoff score, AUC .98 ( p .001; 95% CI
.96 1.0). Similarly high predictive accuracy was noted in the
correctional sample. Using 30 as the cutoff score resulted in an
AUC value of .95 ( p .001; 95% CI .92.98). When the
cutoff score was set at 25, the AUC value increased to .98 ( p
.001; 95% CI .96 1.0).

Categorical Classification Analyses


Participants were assigned to low, moderate, or high PCLR and
PCL:SV groups based on conventionally used values (for PCLR,
low 0 19; moderate 20 29; high 30; for PCL:SV,
low 0 12; moderate 1317; high 18). Table 2 presents
the proportion of each sample classified as low, moderate, and
high on each measure in a 3 3 table with exact agreement on the
diagonal. The proportions in each category were significantly
different than what would be expected based on chance. For the
correctional sample: 2(4, N 177) 145.72, p .001, and
Spearmans r (correlation between low, moderate, and high clas-

Table 2
Classification Values for Low, Moderate, and High Hare
Psychopathy ChecklistRevised (PCLR) and Hare
Psychopathy Checklist: Screening Version (PCL:SV) Groups
PCLR categorical group
PCL:SV
categorical
group

Low
n

Moderate
%

High

5.4
71.1
52.6

0
0
9

47.4

11.9
75.0
47.2

0
2
19

3.3
52.8

Forensic sample
Low
Moderate
High

105
13
0

94.6
28.9

6
32
10

Correctional sample
Low
Moderate
High

74
13
0

88.1
21.7

10
45
17

Note. Values reported are the number of participants and percentages


within each categorical group for a particular sample.

sifications across measures) .79, p .001. For the forensic


sample: 2(4, N 175) 165.10, p .001, and Spearmans r
.82, p .001. Discrepancies were observed in both samples most
often between PCL-R/moderate and PCL:SV/high classifications.
That is, while the PCL:SV categorized 19 (forensic sample) people
in the high category, the PCLR categorized 9 of these same
people as high, and 10 as moderate. In the correctional sample,
36 people were categorized as high on the PCL:SV, compared
with 21 by the PCLR (and 15 as moderate).

Predictive Accuracy
The predictive validity (violent recidivism vs. not violent recidivism) of the PCL:SV relative to the PCLR was examined in the
correctional sample. In general, values for PCL:SV indices were
slightly lower (although not in all cases) than values for PCLR
indices. Point-biserial correlations for PCLR and PCL:SV Total
scores were moderately high: .42 and .37, respectively, both ps
.001. Values for PCLR Factors 1 and 2 and PCL:SV Parts 1 and
2 were .19 and .50 on the one hand, and .21 and .45, on the other
(all at least p .05). Point-biserial correlations for Cooke and
Michies factors for the PCLR and PCL:SV, respectively, were as
follows: Arrogant and Deceitful Interpersonal Style (.16, .17),
Deficient Affective Experience (.16, .20), and Impulsive and Irresponsible Behavioral Style (.54, .51). Correlations for Hares Facet
4 for the PCLR and PCL:SV, respectively, were .43 and .36. All
values were significant at least at the p .05 level. Using tests for
the difference between dependent correlations, we obtained only
one significant difference, t(130) 2.36, p .02, between the
PCLR and PCL:SV in terms of their relationship to violence
(rs .43 and .36 for the PCLR and PCL:SV fourth factors,
respectively). The possibility that this difference stemmed from
chance must be entertained as seven comparisons were tested.

Impact of Type of Scoring Method


Point-biserial correlations were computed between the PCLR
and PCL:SV indices for the forensic psychiatric patients whose
scores were based on file information only (n 67) and for
patients whose scores were based on both interview and file
information (n 108). Correlations were high and ranged from
.90 to .94 in the file-information-only group and from .85 to .95 in
the interview file group (all ps .001). In most cases, correlations for the two groups were very close to one another. Values
for the file-information-only and interview file groups, respectively, were as follows: Total (.94, .95); Hares (1991, 2003)
Factor 1 (.93, .94); Hares Factor 2 (.92, .90); Cooke and Michies
(2001) Factor 1 (.92, .93); Cooke and Michies Factor 2 (.90, .91);
Cooke and Michies Factor 3 (.90, .89); Hares Facet 4 (.91, .85).
Significantly higher mean scores were obtained when an interview was used for the PCLR Total (17.45 vs. 14.58), t(173)
2.58, p .01; d 0.41; PCLR Factor 2 (9.19 vs. 7.88), t(173)
2.12, p .04; d 0.33; and PCLR Facet 4 indices (4.91 vs.
4.02), t(160) 2.05, p .04; d 0.33. There were significantly
different mean scores on seven PCLR items;2 all means were
higher in the interview group. The magnitudes of the differences
for each relevant item were as follows: 1 (d 0.40), 7 (d 0.35),
8 (d 0.35), 11 (d 0.47), 17 (d 0.40), 18 (d 0.42), and 20
(d 0.35). Differences between the file-only versus file inter-

BRIEF REPORTS

view groups were similar for the PCL:SV counterparts (details are
available upon request), although there was a trend toward significance for the PCL:SV Total score to be higher in the file
interview group (11.61 vs. 10.30); t(173) 1.71, p .09, d
0.27. There were significantly different mean scores on three
PCL:SV items,3 all of which also were higher in the interview
group. The magnitudes of the differences for each relevant item
were as follows: 1 (d 0.42), 5 (d 0.45), and 9 (d 0.37).
Coefficient alpha in the file-information-only group was acceptable for the PCLR (.81) and PCL:SV (.86) Total scores, as it was
in the interview group for both the PCLR (.84) and PCL:SV (.84)
Total scores.

Discussion
In the present study we investigated the overall and factor
correspondence between the PCL:SV and PCLR in forensic and
correctional samples in light of the newly proposed three- and
four-factor models of the PCLR. These models represent efforts
to refine the PCL measures and to advance theory pertaining to
psychopathy. Both models support the parsing out of the original
Factor 1 into the affective and interpersonal features of the construct. The inclusion of five items that reflect criminal behavior in
the four-factor model, but not in the three-factor model, has
impelled research investigating the relation between antisocial
behavior and psychopathy. Based on results of structural equation
modeling analyses, Cooke, Michie, Hart, and Clark (2004) concluded that antisocial behavior is viewed best as a secondary
symptom or consequence of psychopathy. The emergence of competing factor models and the subsequent evaluative research that
has ensued are important contributions to the field of clinical
forensic psychology, as substantial repercussions for clinical practice (in terms of risk assessment) and research approaches could be
expected to occur if the newly proposed factors have differential
relationships with external criteria.
Our findings indicated a high degree of convergence in performance between the two measures as evidenced by high intermeasure correlations between Total and factor/part/facet scores; high
AUC values; and, for the most part, similar classification accuracy
for PCL categories. The same raters scored the PCLR and PCL:
SV, a procedure that, methodologically, is distinct from that of
other studies that have compared the measures (Cooke et al., 1999;
Hart et al., 1995). Although this aspect of the procedure is a
strength of the current study in that (as noted earlier) error due to
raters is eliminated from the PCLR/PCL:SV comparisons, it also
is a limitation in that ratings on both measures that are made by the
same person may yield inflated intermeasure correlations. This
overestimate occurs because any random variance in ratings of the
shorter form of a measure (here, the PCL:SV) is at least partially
reproduced in ratings of the longer form (here, the PCLR; see
Smith, McCarthy, & Anderson, 2000). In an effort to address this
potential inflation, we provided estimates of intermeasure correspondence using a subset of independent ratings, a feature of the
research that extends past PCL:SV-PCLR studies (i.e., Cooke et
al., 1999; Hart et al., 1995) by providing a range of intermeasure
correspondence depending on whether the same or different raters
were used. This feature of the research also has applied assessment
implications because the full PCLR could be completed either by
the same or different evaluators as the PCL:SV in actual practice.

229

When independent ratings were made, the correspondence between measures was expectedly lower (although weight given to
these ratings should be tempered with the fact that they were based
on a small subset of cases). In general, however, correlations
between measures remained high.
Predictive validity of the measures was comparable. Clearly, the
predictive validity of Part/Factor 2 far surpassed that of Part/Factor
1 in this sample. A similar pattern was observed for Cooke and
Michies (2001) and Hares (2003) factors/facets, and by Skeem
and Mulvey (2001) using the MacArthur (Monahan et al., 2001)
data set of civil psychiatric patients. It should be noted that the
present study did not address the validity of the PCL:SV either
unto itself or as a screen in circumstances in which there is
insufficient information to complete the PCLR. As such, interpretation of our results is confined to contexts in which the full
PCLR could be scored.
Finally, results also provide support for the use of the PCL:SV
as a screening measure, where false positive prediction errors are
favored compared with false negative prediction errors. The
present study contributes to the body of literature on the assessment of psychopathy by providing estimates of scores based on file
information alone versus file information in conjunction with an
interview. Previous studies have indicated reasonable correspondence between PCLR file-only versus file interview ratings
(Bolt et al., 2004; Grann et al., 1998; Wong, 1988), although they
have not been able to assess whether this coding difference matters
for the relationship between the PCLR and PCL:SV (a feature
with relevance to the many file-only PCL:SV studies in the literature) or how the PCL:SV rather than the PCLR is affected by
file-only ratings. In sum, these findings indicate the PCL:SV has a
robust relation to the PCLR at both the global and factor levels,
and that this relationship holds across coding methods and rater
(in)dependence.

2
For PCLR Item 13, there was a trend toward significance for the mean
in the file interview group to be higher than in the file-only group ( p
.07; d .30).
3
A trend toward significance was observed for the mean in the file
interview group to be higher than in the file-only group for PCL:SV Item
7 ( p .06; d .29) and for PCL:SV Item 12 ( p .10; d .26).

References
Alterman, A. I., Cacciola, J. S., & Rutherford, M. J. (1993). Reliability of
the Revised Psychopathy Checklist in substance abuse patients. Psychological Assessment, 5, 442 448.
Amenta, A. E., Guy, L. S., & Edens, J. F. (2003). Sex offender risk
assessment: A cautionary note regarding measures attempting to quantify risk. Journal of Forensic Psychology Practice, 3, 39 50.
Belfrage, H., Fransson, G., & Strand, S. (2000). Prediction of violence
using the HCR20: A prospective study in two maximum-security
correctional institutions. Journal of Forensic Psychiatry, 11, 67175.
Belfrage, H., & Rying, M. (2004). Characteristics of spousal homicide
perpetrators: A study of all cases of spousal homicide in Sweden
1990 1999. Criminal Behaviour & Mental Health, 14, 121133.
Bolt, D. M., Hare, R. D., Vitale, J. E., & Newman, J. P. (2004). A
multigroup item response theory analysis of the Psychopathy ChecklistRevised. Psychological Assessment, 16, 155168.
Cooke, D. J., Kosson, D. S., & Michie, C. (2001). Psychopathy and

230

BRIEF REPORTS

ethnicity: Structural, item and test generalizability of the Psychopathy


Checklist Revised (PCLR) in Caucasian and African-American participants. Psychological Assessment, 13, 531542.
Cooke, D. J., & Michie, C. (2001). Refining the construct of psychopathy:
Towards a hierarchical model. Psychological Assessment, 13, 171188.
Cooke, D. J., Michie, C., Hart, S. D., & Clark, D. A. (2004). Reconstructing psychopathy: Clarifying the significance of antisocial and socially
deviant behavior in the diagnosis of psychopathic personality disorder.
Journal of Personality Disorders, 18, 337357.
Cooke, D. J., Michie, C., Hart, S. D., & Hare, R. D. (1999). Evaluating the
Screening Version of the Hare Psychopathy ChecklistRevised (PCL:
SV): An item response theory analysis. Psychological Assessment, 11,
313.
Douglas, K. S., Ogloff, J. R. P., Nicholls, T. L., & Grant, I. (1999).
Assessing risk for violence among psychiatric patients: The HCR20
violence risk assessment scheme and the Psychopathy Checklist: Screening Version. Journal of Consulting and Clinical Psychology, 67, 917
930.
Doyle, M., Dolan, M., & McGovern, J. (2002). The validity of North
American risk assessment tools in predicting in-patient violent behaviour in England. Legal and Criminological Psychology, 7, 141154.
Dunlap, W. P. (1999). A program to compute McGraw and Wongs
common language effect size indicator. Behavior Research Methods,
Instruments, & Computers, 31, 706 709.
Grann, M., Lngstrom, N., Tengstrom, A., & Stlenheim, E. G. (1998).
Reliability of file-based retrospective ratings of psychopathy with the
PCLR. Journal of Personality Assessment, 70, 416 426.
Hare, R. D. (1991). Manual for the Hare Psychopathy ChecklistRevised.
Toronto, Ontario, Canada: Multi-Health Systems.
Hare, R. D. (2003). Manual for the Hare Psychopathy ChecklistRevised
(2nd ed.). Toronto, Ontario, Canada: Multi-Health Systems.
Hart, S. D., Cox, D. N., & Hare, R. D. (1995). The Hare Psychopathy
Checklist: Screening Version (PCL:SV). Toronto, Ontario, Canada:
Multi-Health Systems.
Hart, S. D., & Hare, R. D. (1994). Psychopathy and the Big 5: Correlations
between observers ratings of normal and pathological personality. Journal of Personality Disorders, 8, 32 40.
Hart, S. D., Hare, R. D., & Forth, A. E. (1994). Psychopathy as a risk
marker for violence: Development and validation of a screening version
of the Revised Psychopathy Checklist. In J. Monahan & H. Steadman
(Eds.), Violence and mental disorder: Developments in risk assessment
(pp. 8198). Chicago: University of Chicago Press.
Hill, C. D., Neumann, C. S., & Rogers, R. (2004). Confirmatory factor
analysis of the Psychopathy Checklist: Screening Version in offenders
with Axis I disorders. Psychological Assessment, 16, 90 95.
Hill, C. D., Rogers, R., & Bickford, M. E. (1996). Predicting aggressive
and socially disruptive behavior in a maximum security forensic hospital. Journal of Forensic Sciences, 41, 56 59.
Howard, R., Payamal, L. T., & Neo, L. H. (1997). Response modulation

deficits in psychopaths: A failure to confirm and a reconsideration of the


PattersonNewman model. Personality and Individual Differences, 22,
707717.
Johansson, P., Andershed, H., Kerr, M., & Levander, S. (2002). On the
operationalization of psychopathy: Further support for a three-faceted
personality oriented model. Acta Psychiatrica Scandanavica, 106, 81
83.
Kropp, P. R., & Hart, S. D. (2000). The Spousal Assault Risk Assessment
(SARA) Guide: Reliability and validity in adult male offenders. Law and
Human Behavior, 24, 101118.
Monahan, J., Steadman, H. J., Silver, E., Appelbaum, P. S., Robbins, P. C.,
Mulvey, E. P., et al. (2001). Rethinking risk assessment: The MacArthur
study of mental disorder and violence. New York: Oxford University
Press.
Mossman, D., & Somoza, E. (1991). ROC curves, test accuracy, and the
description of diagnostic tests. Journal of Neuropsychiatry and Clinical
Neurosciences, 3, 330 333.
Serin, R. C. (1993). Diagnosis of psychopathology with and without an
interview. Journal of Clinical Psychology, 49, 367372.
Skeem, J., & Mulvey, E. P. (2001). Psychopathy and community violence
among civil psychiatric patients: Results from the MacArthur Violence
Risk Assessment Study. Journal of Consulting and Clinical Psychology,
69, 358 374.
Skeem, J., Mulvey, E. P., & Grisso, T. (2003). Applicability of traditional
and revised models of psychopathy to the Psychopathy Checklist:
Screening Version. Psychological Assessment, 15, 4155.
Smith, G. T., McCarthy, D. M., & Anderson, K. G. (2000). On the sins of
short-form development. Psychological Assessment, 12, 102111.
Steadman, H. J., Silver, E., Monahan, J., Appelbaum, P. S., Robbins, P. C.,
Mulvey, E. P., et al. (2000). A classification tree approach to the
development of actuarial violence risk assessment tools. Law and Human Behavior, 24, 83100.
Strand, S., Belfrage, H., Fransson, G., & Levander, S. (1999). Clinical and
risk management factors in risk prediction of mentally disordered offendersMore important than historical data? A retrospective study of
40 mentally disordered offenders assessed with the HCR20 violence
risk assessment scheme. Legal and Criminological Psychology, 4, 67
76.
Vitacco, M. J., Neumann, C. S., & Jackson, R. L. (2005). Testing a
four-factor model of psychopathy and its association with ethnicity,
gender, intelligence, and violence. Journal of Consulting and Clinical
Psychology, 73, 466 476.
Wong, S. (1988). Is Hares Psychopathy Checklist reliable without the
interview? Psychological Reports, 62, 931934.

Received January 21, 2005


Revision received September 28, 2005
Accepted October 4, 2005