The Revised Psychopathy Checklist: Reliability and Factor Structure

Abstract

The revised Psychopathy Checklist (PCL) is a 20-item scale scored from interview and file information. Analyses of data from 5 prison samples (
N = 925) and 3 forensic psychiatric samples (
N = 356) indicate that the revised PCL resembles its 22-item predecessor in all important respects. It has excellent psychometric properties, and it measures 2 correlated factors that were cross-validated both within and between samples. Correlations between the original PCL and the revised version approached unity for both the factors and the full scale. We conclude that the revised PCL measures the same construct as the original and that the PCL is a reliable and valid instrument for the assessment of psychopathy in male forensic populations. (PsycINFO Database Record (c) 2012 APA, all rights reserved)

Full text

Psychological
Assessment:
A
Journal
of
Consulting
and
Clinical Psychology
1990,
Vol.
2, No.
3,338-341
Copyright
1990
by the
American Psychological Association,
Inc.
1040-3590/90/S00.75
The
Revised Psychopathy Checklist: Reliability
and
Factor Structure
Robert
D.
Hare, Timothy
J.
Harpur,
A. R.
Hakstian,
Adelle
E.
Forth
and
Stephen
D.
Hart
University
of
British Columbia
Vancouver,
British Columbia, Canada
Joseph
P.
Newman
University
of
Wisconsin
The
revised Psychopathy Checklist (PCL)
is a
20-item
scale scored
from
interview
and file
informa-
tion. Analyses
of
data
from
5
prison
samples
(N=
92
5)
and 3
forensic
psychiatric
samples
(N=
356)
indicate
that
the
revised
PCL
resembles
its
22-item
predecessor
in all
important respects.
It has
excellent
psychometric properties,
and it
measures
2
correlated factors that were cross-validated
both within
and
between samples. Correlations between
the
original
PCL and the
revised version
approached unity
for
both
the
factors
and the
full
scale.
We
conclude that
the
revised
PCL
mea-
sures
the
same construct
as the
original
and
that
the PCL is a
reliable
and
valid instrument
for the
assessment
of
psychopathy
in
male
forensic
populations.
During
the
last decade,
we
have devoted considerable
effort
to the
development
of an
assessment procedure
for
criminal
psychopathy
that
has
acceptable psychometric properties
and
that
is
closely tied
to
traditional clinical conceptions
of
psy-
chopathy.
The
result
is the
Psychopathy Checklist (PCL).
The
original
PCL
consists
of 22
personality
and
behavioral
items,
which
are
completed
on the
basis
of
interview
and file
information
(Hare, 1980). There
is a
considerable amount
of
published
and
unpublished evidence attesting
to the
reliability
and
validity
of the
PCL.1
The PCL has
been shown
to
measure
two
correlated factors (Harpur, Hakstian,
&
Hare,
1988).
There
are
differences
in
external correlates
of the
factors,
and we
have
argued that measurement
of
both factors
is
important
for a
comprehensive assessment
of
psychopathy (Harpur
&
Hare,
1989;
Harpur, Hare,
&
Hakstian, 1989).
The
revised Psychopathy Checklist
(PCL-R;
Hare, 1985)
differs
from
the PCL in
several ways.
Two PCL
items, previous
diagnosis
as a
psychopath
or
similar
and
drug
or
alcohol
not
direct
cause
of
antisocial
behavior, have been omitted because
they
had
relatively
low
correlations with
the
total score
and
were
difficult
to
score.
The
scoring
of
some items
was
made
more stringent
by
requiring more extreme instances
of
deviant
behavior
for a
score
of 2. For a few
items,
the
scoring criteria
were
modified somewhat.
It
is
important
for the
continuity
of
research carried
out
using
the
PCL to
demonstrate empirically that
the
revised version
Preparation
of
this report
was
supported
by
Grant
MT-4511
from
the
Medical
Research Council
of
Canada
to
Robert
D.
Hare
and by
Grant
MH37711
from
the
National Institute
of
Mental Health
to Jo-
seph
P.
Newman.
Correspondence concerning this article
and
requests
for an ex-
tended report should
be
addressed
to
Robert
D.
Hare, Department
of
Psychology,
University
of
British
Columbia, Vancouver, British Colum-
bia,
Canada
V6T
1Y7.
does
not
differ
substantially
from
the
original. Even minor
changes
in
scoring criteria
can
result
in
substantive changes
in
the
meaning
of an
item.
The
purpose
of
this article
is to
demon-
strate that
the
changes introduced into
the
PCL-R
have
not
altered
its
psychometric properties
and
that
the
body
of re-
search
accumulated using
the PCL can be
generalized
to the
revised
measure.
Method
We
obtained
PCL-R
item scores
for five
samples
of
male prison
inmates
(Samples
1-5;
n
=
925)
and
three samples
of
male
forensic
psychiatric
patients (Samples
6-8;
n =
356).
These samples were
com-
pletely
independent
of
those
reported
by
Harpur
et
al.
(1988).
All of the
inmates
in
Sample
3 and at
least
85% of
those
in the
other samples were
White; most
of the
other inmates were Native American. Subjects' ages
ranged
from
16
to 69
years,
but the
majority were under
age 40
(see
Harpur
&
Hare, 1989,
for a
detailed breakdown
of the
ages
of
four
of
the
current samples). Because inmates
suffering
from serious mental
disorders
are
usually
transferred
to
forensic hospitals,
the
prison
sam-
ples generally
did not
include
any
psychotic patients. Subjects
in the
forensic
psychiatric samples
suffered
from
a
variety
of
mental
dis-
orders, including psychotic disorders
(see
Hart
&
Hare, 1989).
1.
Sample
1
consisted
of 241
inmates
of a
federal medium-security
institution
in
British Columbia, Canada,
all of
whom were serving
sentences
of 2
years
or
longer. These subjects
had
volunteered
to
take
part
in
several research projects.
Two
ratings were available
for
176
of
the
inmates.
2.
Subjects
in
Sample
2
were
122
inmates
at a
provincial prison
in
British
Columbia,
all
serving sentences
of
less than
2
years. They
had
volunteered
to
participate
in
several laboratory
and
paper-and-pencil
tests.
Two
ratings were available
for 37 of the
inmates.
3.
Sample
3
consisted
of 369
inmates
at a
minimum-security institu-
1
A
mimeographed manual
for
this revised
PCL is
available
on re-
quest,
and a
more
formal
manual
is
in
preparation.
A
current bibliogra-
phy
of
research using
the PCL
will
be
included
in the
manual.
338

BRIEF
REPORTS
339
tion
in
Wisconsin. They
had
volunteered
to
participate
in
several
ex-
periments conducted
by
Joseph Newman
and his
colleagues.
Two
rat-
ings were available
for 72 of the
inmates.
4.
Sample
4
consisted
of 106
inmates
who
were
assessed,
using
a
French translation
of the PCL
(Cote
&
Hodgins,
1989),
just
prior
to
conditional release from
minimum-,
medium-,
or
maximum-security
federal
prisons
in
Quebec.
Two
ratings were available
for 70 of the
inmates.
5.
Sample
5
consisted
of 87
inmates
in a
medium-security prison
in
Kingston, Ontario. They
had
volunteered
to
participate
in
research
projects conducted
by
Ralph Serin. Only
one
rating
was
available
for
each inmate.
6.
Sample
6
consisted
of 80
patients
who had
been
remanded
to a
forensic
hospital
in
British Columbia
for
inpatient assessment
of fit-
ness
to
stand trial (see Hart
&
Hare,
1989).
Two
raters completed
the
PCL-R
for
each patient.
7.
Sample
7
consisted
of
165
patients
in an
intensive therapeutic
program
for
mentally
disordered
offenders
in a
forensic psychiatric
unit
of the
Mental
Health
Center
at
Penetanguishene,
Ontario,
Canada
(Rice
&
Harris,
1989).
Only
one
rating
was
available
for
each patient.
8.
Sample
8
consisted
of
111
patients
at the
Regional Psychiatric
Center (RPC)
in
Saskatchewan, Canada.
The RPC is a
forensic psychi-
atric
hospital that accepts inmates
from
federal prisons.
The
ratings
were collected
by
Stephen Wong. Only
one
rating
was
available
for
each
inmate.
We did not
have
access
to
item scores
for
this
sample.
Each
PCL-R
item
was
scored
on a
3-point scale
(0,1,2)
according
to
the
degree
to
which
it
applied
to the
individual.
Occasionally,
it was
not
possible
to
complete
an
item with confidence,
in
which
case
the
item
was
omitted,
and the
total
score
was
prorated
to a
20-item
scale.
For
computation
of
coefficient
alpha,
the
value
1 was
assigned
to
omit-
ted
items.
Different
raters
were involved
in
assessments made
at
differ-
ent
sites. They included clinical psychologists, graduate students,
and
research
assistants.
Where possible,
a
second rater completed
the
PCL-R
independently after observing
the
original interview
(which
was
either live
or
recorded
on
videotape)
and
reviewing
the file
infor-
mation. When available,
these
ratings were
averaged
together.
For
the five
prison samples
and for
Sample
8,
items
in the PCL
were
scored
by
integrating interview
and file
information.
However, some
of
the
patients
in
Sample
6 and all of
those
in
Sample
7
could
not be
interviewed because they were
no
longer
at the
institution;
for
these
patients,
PCL-R
ratings
were made
on the
basis
of
extensive correc-
tional
and
hospital
files
(see Wong,
1988).
We
performed factor analyses
on
Samples
1,2,3,4,
and 7.
Samples
5
and
6
were omitted because
of
small sample sizes,
and
item
scores
for
Sample
8
were
not
available.
Two
subjects from Sample
7, for
whom
more than
four
PCL
items could
not be
scored because
of
missing
data, were dropped
for
these analyses. Correlation matrices were cal-
culated after replacing missing
data
by
item means
for
that sample.
We
followed
the
procedures
used
by
Harpur
et
al.
(1988),
to
which
the
reader
is
referred
for
details.
We
used
split-half
cross-validation
(Everett, 1983)
on
Samples
1 and 3. We
next
compared
congruence
coefficients
for the
best
two-,
three-,
and
four-factor
oblique solutions
from
the five
data sets independently.
A
single invariant factor pattern
was
subsequently obtained
for
Samples
1,3,4,
and 7,
using
the
Method
1
procedure
described
by
Meredith
(1964).
Finally,
the
psychometric
properties
of
scales derived from
the
factors were examined.
Results
and
Discussion
Full-Scale
PCL
The
items
in the
PCL-R
are
listed
in
Table
1,
along
with
the
mean
item-total
correlation
for
Samples
1
-7 and the
common
primary-factor
pattern
for the
four
samples
included
in the
Table
1
Revised
Psychopathy
Checklist
Items,
Their
Mean
Item-Total
Correlations,
and the
Common
Primary-Factor
Pattern
Item
1
.
Glibness/superficial
charm
2.
Grandiose sense
of
self-worth
3.
Need
for
stimulation
4.
Pathological lying
5.
Conning/manipulative
6.
Lack
of
remorse
or
guilt
7.
Shallow
affect
8.
Callous/lack
of
empathy
9.
Parasitic
lifestyle
10.
Poor behavioral controls
1
1
.
Promiscuous sexual behavior
12.
Early behavior
problems
1
3.
Lack
of
realistic
goals
14.
Impulsivity
15.
Irresponsibility
16.
Failure
to
accept responsibility
17.
Many short-term relationships
18.
Juvenile
delinquency
19.
Revocation
of
conditional
release
20.
Criminal versatility
Mean
item-total
correlation*
.50
.52
.58
,53
.58
.53
.58
.64
.44
.50
.37
.46
.50
.52
.53
.38
.27
.32
.35
.43
Factor*
1
.86
.76
.09
.62
.59
.53
.57
.53
-.00
.14
.35
-.01
.10
.01
.16
.47
.18
-.18
-.00
.15
2
-.25
-.16
.56
.03
.10
.11
.10
.22
.56
.44
.08
.56
.56
.66
.51
.02
.18
.59
.44
.33
*
Mean
of
Fisher-Z transformed,
corrected
item-total
correlations
for
Samples
1-7,
weighted
by
sample size.
b
Coefficients
> 0.4
have
been
underlined. Coefficients
for the
com-
mon
solution
are
scaled
so
that
the
column sums
of
squares
are
equal
to
those
of the
average
of the
four
sample solutions.
pooled-factor
solution.
The means,
medians,
and
standard
de-
viations
of the
PCL-R
total
scores
were
highly
similar
for
both
the
prison
and
forensic
psychiatric
samples.
Means
ranged
from
20.1
to
23.9
(M=
22.1),
medians
from 20.0
to
25.0
(M =
22.8),
and
standard
deviations
ranged
from
6.7 to 9.0 (M =
7.95).
The
distribution
of
scores
within
each
sample
was
ap-
proximately
normal.
Although
several
different
types
of
insti-
tutions
and
many
different
raters
were
involved,
the
character-
istics
of
each
sample
within
each
type
of
setting
were
very
simi-
lar
to one
another.
To
assess
interrater
reliability,
intraclass
correlation
coeffi-
cients
(ICCs;
Shrout
&
Fleiss,
1979)
were
calculated
for
subsam-
ples
of
subjects
rated
by
more
than
one
rater
(see
sample
de-
scriptions
for
№).
They
ranged
from
.78 to .94
(M
=
.86)
for a
single
rating
and
from
.87 to .97
(M
=
.93)
for the
average
of two
ratings
(computations
based
on the
Spearman-Brown
Prophecy
Formula).2
Internal
consistency
was
assessed
by
Cronbach's
co-
efficient
alpha,
which
ranged
from
.85 to .89
(M=
.88),
and by
the
mean
interitem
correlation,
which
ranged
from
.23 to .30
(M
=
.27).
These
results
indicate
that
the
PCL-R,
like
the
PCL,
can
be
considered
a
homogeneous,
unidimensional
scale
(Har-
pur
et
al.,
1989).
There
were
no
apparent
differences
in
reliabil-
ity
coefficients
obtained
in
prison
versus
forensic
psychiatric
samples.
2
Whenever
possible,
assessments
should
be
based
on the
average
of
at
least
two
independent
ratings.

340
BRIEF
REPORTS
Factor
Structure
Factor comparability
coefficients
were calculated
for
Sam-
ples
1 and 3,
with
from
two to five
factors extracted. Only
the
two-factor
solutions
produced
factors
that
cross-validated
within
each sample.
The
results
for the
PCL-R
were very simi-
lar
to
those presented
by
Harpur
et
al.
(1988)
for the PCL and
suggest
that only
two
factors
should
be
extracted.
Comparability
coefficients
for
two-factor solutions
of in-
creasing obliquity were also calculated.
The
value
of the
coeffi-
cients increased steadily with
the
increasing obliquity
of the
solutions.
The
most oblique Harris-Kaiser independent cluster
solution
resulted
in the
highest
coefficients,
suggesting
that
an
oblique
transformation
is
most appropriate
for
these data
(Everett,
1983).
The
independent two-factor solutions
for
Samples
1,3,4,
and
7
were
in
close agreement;
all of the
congruence
coefficients
between
corresponding factors equaled
or
surpassed
.83, close
to the
criterion
of .85
used
by
Harpur
et al.
(1988)
to
indicate
the
equivalence
of two
factors. There
was
less agreement among
these samples
for the
three-
or
four-factor
solutions. However,
Sample
2 did not
produce
a
solution congruent with
any of the
other
four
samples.
For
this reason,
it was not
included
in the
pooled-factor
solution described below.
The
primary-factor pattern matrices
for
Samples
1,3,4,
and 7
were
used
to
obtain
the
pooled solution, shown
in
Table
1,
which
represents
a
best
fit to the
four
sample correlation matri-
ces. Examination
of
both
the
pooled
and
sample solutions
re-
vealed
a
clear two-factor structure
very
similar
to
that
de-
scribed
for the
PCL.
As was the
case
for the
PCL,
the
item
promiscuous
sexual behavior
failed
to
load above
.4 on
either
factor.
The
loadings
for two
items
differed
for the
PCL-R
solu-
tion: many short-term relationships
and
criminal versatility
failed
to
load above
.4 on
Factor
2, as
they
had for the
PCL.
If
anything,
the
former
loaded
on
Factor
1.
All
three items demon-
strated inconsistent loadings across
the
four
sample solutions.
Correlations between
the two
factors ranged
from
.39 to .54
(M
=
.48)
in the
separate-sample solutions.
The
congruence
coefficients
between corresponding factors
for the
common
pattern
and
sample patterns transformed into congruence with
the
common pattern were
uniformly
high (range
=
.94-98).
The
congruence
coefficients
between
the
sample patterns
themselves
ranged
from
.82 to .94
(M
=
.89).
On
the
basis
of the
results reported above,
we
judged that
Factors
1 and 2
could
be
defined
by
eight
and
nine
PCL-R
items,
respectively. Scores
on
Factor
1
were derived
by
summing
the
scores
on
Items
1,
2, 4, 5, 6, 7, 8, and
16.
Scores
on
Factor
2
were derived
by
summing
the
scores
on
Items
3,9,10,12,13,14,
15,18,
and
19.
The
interrater reliability
and
internal consistency
of the
fac-
tors remained high
despite
the
small number
of
items included
in
each scale.
The
mean
ICCs
for
Factors
1 and 2
were
.77 and
.85, respectively,
for
a
single rating,
and .87 and
.92, respectively,
for
the
average
of two
ratings.
The
mean
coefficient
alphas were
.84 and .79 for
Factors
1 and 2,
respectively.
The
corresponding
mean
interitem correlations were
.39 and
.31. Although
the
ICCs
were slightly higher
for
Factor
2,
which
is
scored more
on
the
basis
of
objectively
defined
criteria than
for
Factor
1,
which
relies
more
on
judgments
and
inferences about personality,
Factor
1
generally demonstrated higher internal consistency
than Factor
2, as
indicated
by
both alpha
coefficients
and
mean
interitem
correlations.
Correlation
Between
PCL and
PCL-R
The
empirical factor structure
and the
nature
of the
scoring
changes
introduced suggest that
the
PCL-R
measures
the
same
construct that
is
measured
by the
PCL. However, more direct
evidence
may be
obtained
by
comparing
PCL
total
scores
and
factor-scale
scores directly.
So
far,
we
have been able
to
obtain
independent ratings
on the PCL and
PCL-R
only with Sample
2.
The
correlation between
the PCL and the
PCL-R
total
scores
was
.88; correlations
of .82 and .81
were obtained
for
Factors
1 and 2,
respectively. However, these values
are
limited
by
the
reliability
of the
scales themselves. True estimates
of the
correlations between
the
constructs
measured
by the PCL and
the
PCL-R
are
obtained when
the
correlations
are
disatten-
uated,
or
corrected
for
unreliability
of the
scales (Hakstian,
Schroeder,
&
Rogers,
1988).
Using
ICC
estimates
of
reliability,
the
true correlation between
the PCL and the
PCL-R
total
scores lies between
.95 and
l.O.3
Correlations
for
Factor
1 lie
between
.94 and
1.0;
those
for
Factor
2 lie
between
.93 and
1.0.
Discussion
The PCL is a
reliable
and
valid
measure
of
psychopathy
in
male criminal populations.
The
current results demonstrate
the
reliability
and
concurrent validity
of the
PCL-R.
In
addi-
tion,
the
results
of a
number
of
recent studies (described
in the
forthcoming
manual)
confirm
the
construct
and
predictive
va-
lidity
of the
revised version.
We
have
no
hesitation
in
conclud-
ing
that both versions measure
the
same construct
and
that
findings
obtained with
one
version
are
generalizable
to the
other. Both versions provide continuous scores
reflecting
the
extent
to
which
an
individual matches
the
prototypical psycho-
path,
as
well
as
factor
scores
that permit
the
social deviance
component
of
psychopathy
to be
separated
from
the
cluster
of
personality
traits that
are
fundamental
to the
construct.
In
addition
to
validating
the
PCL-R,
these results also pro-
vide
a
strong replication
of an
earlier study
on the
factorial
structure
of the PCL
(Harpur
et al,
1988).
Of the
11
samples
analyzed
to
date
(N =
2,119),
the
optimal two-factor solution
obtained
by
independent analysis
of
each sample
has
been
highly
consistent
for 10 of the
samples.
We
feel
confident that
the
eight-
and
nine-item scales described here
are
adequate
measures
of the two
factors
and
that results reported
for the
PCL
factors
will
be
applicable
to the
PCL-R
factors.
Interpre-
tation
of the
factors
as
measuring
a
selfish, callous,
and re-
morseless
use of
others (Factor
1) and a
chronically unstable,
antisocial,
and
socially deviant
lifestyle
(Factor
2)
remains
un-
changed.
Sample
2 did not
possess
the
same factor structure
as the
other samples
in
this study. There
are
several possible explana-
3
The
precise value depends
on
whether
the
estimate
of
reliability
for
one
rater
or for two is
used
for
disattenuation. Because some
of the
subjects
in
each sample were
rated
by two
raters,
these
values represent
upper
and
lower
bounds
for the
reliabilities.

BRIEF
REPORTS
341
tions
for
this.
The
relatively small sample size
(n
=
122)
may
have
resulted
in a
discrepant
and
unreliable factor structure result-
ing
from
sampling error.
In
addition,
the
data
for
this sample
were
gathered
on
provincial prison inmates,
for
whom
far
less
extensive
or
complete
file
information
was
available. Finally,
because this
was the first
sample with which
we
used
the
PCL-
R, it is
possible that minor changes
in
details
of the
scoring
procedure were introduced
as the
data were
gathered—changes
that could have
led to
somewhat variable ratings
for
some items.
Despite
the
fact
that
the
factor structure
of
Sample
2 was
anomalous,
the
psychometric properties
of the
factor-scale
scores derived
from
this sample were similar
to
those
for the
other samples. These scores were essentially identical
to
factor-
scale scores independently derived
from
the
PCL.
It
is
clear that
the
PCL,
in
both
of its
versions,
is a
robust
measure
of
psychopathy
in
White, male criminals. Research
is
currently
underway
to
assess
the
applicability
of the PCL to
Black,
Native American,
and
female
criminals,
as
well
as to
noncriminals.
References
Cote,
G., &
Hodgins,
S.
(1989).
Une
mesure
de
psychopathic
validation
de
la
version
fran(ais
(The
Psychopathy Checklist-French transla-
tion). Manuscript
in
preparation.
Everett,
J. E.
(1983).
Factor
comparability
as a
means
of
determining
the
number
of
factors
and
their rotation.
Mullivariate
Behavioral
Research,
18,
197-218.
Hakstian,
A. R.,
Schroeder,
M. L., &
Rogers,
W
T.
(1988).
Inferential
procedures
for
correlation
coefficients
corrected
for
attenuation.
Psychometrika,
53,
27-43.
Hare,
R. D.
(1980).
A
research scale
for
the
assessment
of
psychopathy
in
criminal populations. Personality
and
Individual
Differences,
1,
111-119.
Hare,
R. D.
(1985).
The
Psychopathy Checklist. Unpublished manu-
script, University
of
British Columbia, Vancouver, Canada.
Harpur,
T.
J.,
Hakstian,
A.
R.,
&
Hare,
R. D.
(1988).
Factor structure
of
the
Psychopathy Checklist. Journal
of
Consulting
and
Clinical Psy-
chology,
56,
741-747.
Harpur,
T.
J.,
&
Hare,
R. D.
(1989).
The
assessment
of
psychopathy
as a
function
of
age.
Manuscript submitted
for
publication.
Harpur,
T.
J.,
Hare,
R. D, &
Hakstian,
A. R.
(1989).
Two-factor
concep-
tualization
of
psychopathy: Construct validity
and
assessment
im-
plications. Psychological Assessment:
A
Journal
of
Consulting
and
Clinical
Psychology,
1,6-11.
Hart,
S.
D.,
&
Hare,
R. D.
(1989). Discriminant
validity
of the
Psy-
chopathy Checklist
in a
forensic psychiatric population. Psychologi-
cal
Assessment:
A
Journal
of
Consulting
and
Clinical
Psychology,
1,
211-218.
Meredith,
W
(1964). Rotation
to
achieve factorial invariance.
Psycho-
metrika,
29,
187-206.
Rice,
M.
E.,
&
Harris,
G. T.
(1989).
A
longterm study
of
psychopaths
treated
in a
therapeutic community. Manuscript submitted
for
publi-
cation.
Shrout,
P.
E.,
&
Fleiss,
J. L.
(1979). Intraclass correlations: Uses
in
assessing rater reliability.
Psychological
Bulletin,
86,
420-428.
Wong,
S.
(1988).
Is
Hare's Psychopathy Checklist reliable without
an
interview?
Psychological
Reports,
62,
931-934.
Received
June
22,1989
Revision
received January
4,1990
Accepted January
23,1990
•