J Youth Adolescence (2007) 361048–1057DOI 10.1007s10964-00.docx

J Youth Adolescence (2007) 36:1048–1057
DOI 10.1007/s10964-006-9163-1
O R I G I N A L PA P E R
Construct Validity of Adolescent Antisocial Personality
Disorder
Jeanette Taylor · Irene J. Elkins · Lisa Legrand ·
Dawn Peuschold · William G. Iacono
Received: 4 October 2006 / Accepted: 14 December 2006 /
Published online: 17 January 2007
C© Springer Science+Business Media, LLC 2007
Abstract This study examined the construct validity of
antisocial personality disorder (ASPD) diagnosed in adoles-
Jeanette Taylor is an Assistant Professor at Florida State
University.
She received her Ph.D. in 1999 from the University of
Minnesota. Her
research examines biological, cognitive, and environmental
influences
on personality disorders and substance use disorders.
Irene J. Elkins is a Co-investigator at the Minnesota Center for
Twin
and Family Research. She received her Ph.D. in 1993 from the
University of Kansas and is a member of the Society for
Research on
Psychopathology. Her major research interests include the
relationship

of personality, ADHD, and antisocial behavior to substance
abuse, as
well as gene-environment interrelationships.
Lisa Legrand is a Research Associate with the Minnesota Center
for
Twin and Family Research at the University of Minnesota. She
received her Ph.D. in 2003 from the University of Minnesota.
Her
major research interests include gene-environment interplay in
the
development of externalizing psychopathology.
Dawn Peuschold is a Senior Clinical Forensic Psychologist at
the
Hennepin Country District Court and an instructor at the
University of
Minnesota. She completed a post-doctoral fellowship in forensic
psychology at the University of Massachusetts Medical School
after
receiving her Ph.D. in clinical psychology from the University
of
Minnesota. She is interested in risk factors for reoffense in
juveniles.
William G. Iacono is a Distinguished McKnight University
Professor
at the University of Minnesota, the institution from which he
received
his Ph.D. in psychology. He conducts longitudinal studies of
adolescent twins and adoptees aimed at understanding the
development of adult adjustment and mental health.
J. Taylor (�)
Department of Psychology, Florida State University,
Tallahassee, Florida 32306-1270, USA

e-mail: [email protected]
I. J. Elkins · L. Legrand · D. Peuschold · W. G. Iacono
Department of Psychology, University of Minnesota – Twin
Cities,
Minnesota, USA
cence. Boys and girls were grouped by history of DSM-III-R
conduct disorder (CD) and ASPD: Controls (n = 340) had
neither diagnosis; CD Only (n = 77) had CD by age 17 but
no ASPD through age 20; Adolescent ASPD (n = 64) had
ASPD by age 17. The Adolescent ASPD group was then
compared to 20 young adult men who met criteria for ASPD
(ASPD group). As expected, the Adolescent ASPD group
had significantly more depression and substance use disor-
ders, a greater performance >verbal IQ discrepancy, more
deviant peers, and poorer academic functioning than the CD
Only group and Controls. The Adolescent ASPD and ASPD
groups did not differ on most variables. Results support the
construct validity of Adolescent ASPD and suggest that such
a diagnosis could help identify adolescents at risk for persis-
tent antisocial behavior.
Keywords Antisocial personality . Adolescent antisocial
behavior . Co-morbidity
Antisocial behavior is so common among adolescents
that some have called it normative (Moffitt, 1993). Non-
normative manifestations of antisocial behavior can be iden-
tified in children and adolescents by applying diagnostic cri-
teria for conduct disorder (CD) as outlined in the Diagnostic
and Statistical Manual of Mental Disorders, 4th Ed. (DSM-
IV: American Psychiatric Association, 1994). Another DSM
diagnosis that describes manifestations of antisocial behavior
is antisocial personality disorder (ASPD), which is reserved
for adults aged 18 and older because research shows that
while most adult antisocial behavior has its roots in child-

hood, most antisocial children do not grow up to be antisocial
adults (Robins, 1966). As such, the age requirement for the
ASPD diagnosis helps avoid placing a lasting label on chil-
dren who might very well desist in their antisocial behavior.
Unfortunately, this age restriction on the application of the
Springer
J Youth Adolescence (2007) 36:1048–1057 1049
ASPD diagnosis may result in overlooking a clinically im-
portant group of antisocial adolescents.
Research suggests that earlier onset of CD and other types
of conduct problems are associated with a poor prognosis in
terms of academic functioning and trajectory of antisocial
behavior (e.g., Moffitt & Caspi, 2001). A similar set of poor
outcomes might hold for early presentations of ASPD, in
which case the identification of those individuals would be
useful in order to hasten intervention. Moreover, the type of
intervention that is applied with adolescents exhibiting CD
only might differ from that applied to those with ASPD in
terms of duration needed to see improvement, client moti-
vation for treatment, adherence to treatment, and need for
treatment of more numerous and/or severe co-morbid disor-
ders (e.g., substance dependence). At present there are no
studies in the literature that speak directly to the construct
validity of an ASPD diagnosis assigned before age 18 (here-
after referred to as “Adolescent ASPD”). However, there is
some precedent for examining “adult” constructs in child-
hood or adolescence.
Psychopathy is a manifestation of antisocial behavior
comprised of an antisocial lifestyle combined with a cal-

lous and emotionally detached interpersonal style, and early
descriptions of this clinical phenomenon informed the cre-
ation of the ASPD diagnosis. However, psychopathy is not
a DSM-based diagnosis and has no age restriction for its
application, and a large body of research exists on the
downward extension of the psychopathy construct to chil-
dren. Findings indicate that children with CD who also
show callous and unemotional traits have similar charac-
teristics of adults with psychopathy in terms of response
styles (Barry et al., 2000; O’Brien & Frick, 1996), an-
tisocial behavior pattern (Christian et al., 1997; Frick &
Ellis, 1999), and processing of emotional stimuli (Blair,
1999; Frick et al., 1999), suggesting that the psychopa-
thy construct is likely valid for children. To the extent that
psychopathy and ASPD have overlapping features and tap
a similar construct (i.e., a persistent antisocial personal-
ity/behavioral style), it stands to reason that the ASPD diag-
nosis might similarly be valid when applied to adolescents
who are not yet 18, but this notion has not yet been tested.
Although the bulk of research appears to support the
downward extension of psychopathy to children, it is im-
portant to note that this extension has occurred with some
controversy and debate over the possible dangers of applying
such a label to children (Frick, 2002; Lynam, 2002; Seagrave
& Grisso, 2002). One of the primary concerns is that adult
psychopathy is thought to be treatment resistant and, there-
fore, identifying juveniles as psychopathic might lead the
criminal justice system to view those individuals as simi-
larly untreatable perhaps leading to a premature dismissal
of the possibility of rehabilitation that could result in more
harsh sentences (Seagrave & Grisso, 2002). Another concern
is that characteristics of adults such as psychopathic traits
might appear transiently in juveniles (Seagrave & Grisso,
2002) thereby leading to the potential for false labels that

could carry serious consequences. However, Frick (2002)
persuasively argues that psychopathy is not that different
from other disorders found in adults that have been applied
to children and adolescents (e.g., depression), and a similar
argument could be made for ASPD.
The goal of the present study was to examine the validity
of Adolescent ASPD. To this end, we compared adolescents
who met the criteria for ASPD to those with CD only
on robust correlates of antisocial behavior to show that
CD and Adolescent ASPD are distinct in important ways
and may therefore convey different information about an
adolescent’s antisocial behavior problem. Alcohol and drug
dependence (Cottler et al., 1995; Helzer & Pryzbeck, 1988;
Hesselbrock, 1991; Taylor & Carey, 1998), depression
(Moffitt et al., 2001; Pager, 1998; Schuckit, 1986), and
attention-deficit/hyperactivity disorder (ADHD) (Disney
et al., 1999; Lahey et al., 1995) are positively associated
with antisocial behavior across gender. A differential rate of
association of these disorders with ASPD versus CD in ado-
lescents would indicate potential differences with regard to
psychological impairment that could impact the severity of
antisocial behavior (e.g., alcohol or drug dependence serving
to increase antisocial behavior) and/or the complexity of the
treatment plan (e.g., more issues to address in treatment).
Cognitive functioning has also been linked to antisocial
behavior. Specifically, verbal intelligence is inversely related
to antisocial behavior in children and adolescents (White
et al., 1994), and low verbal IQ predicts persistence of antiso-
cial behavior (Farrington & Hawkins, 1991; Simonoff et al.,
2004). Furthermore, antisocial behavior is often associated
with a marked performance IQ > verbal IQ (PIQ > VIQ)
discrepancy (Cornell & Wilson, 1992; Snow & Thurber,
1997). Given these findings, it may be the case that ASPD
(characterized by persistence of antisocial behavior from

adolescence into adulthood) is underlied, in part, by greater
VIQ deficits than CD that does not develop into ASPD.
Antisocial behavior has also been robustly linked to de-
viance in peer groups (e.g., Jessor et al., 1995; Simons et al.,
1994) and to poor academic functioning (Patterson, 1986).
Antisocial peers might serve both as models for and facil-
itators of antisocial behavior, which suggests that they can
impact the likelihood that antisocial behavior will persist (as
in the case of ASPD) or desist (as in the case of CD only). The
persistence associated with ASPD might also translate into
a particularly poor level of academic functioning given that
the effects of early school problems likely compound over
time. The potential differential association of deviant peers
and academic functioning with ASPD versus CD only might
have important impacts on family functioning (e.g., greater
family conflict surrounding highly deviant friends and bad
Springer
1050 J Youth Adolescence (2007) 36:1048–1057
grades among adolescents with ASPD) as well as treatment
plans (e.g., need to focus attention on academic improve-
ments and perhaps remedial education for adolescents with
ASPD).
Finally, the influence of genes on CD and ASPD is well
documented and a recent study shows that a similar set of ge-
netic factors contributes to both disorders in men and women
but adult symptoms of ASPD have an overall greater genetic
influence than CD (Kendler et al., 2003). Thus, ASPD and
CD that does not progress into ASPD might have different
etiological structures that could help explain the difference

in their course (i.e., greater genetic influence on ASPD might
contribute to persistence of antisocial behavior).
Hypotheses
Our hypothesis was that Adolescent ASPD is a valid con-
struct, and this was examined in two ways. First, we com-
pared an Adolescent ASPD group, a CD Only group (who
had a CD diagnosis, but not ASPD), and a Control group
(with no CD or ASPD diagnosis) on robust correlates of an-
tisocial behavior. We had the following specific predictions
for these comparisons:
1. Boys and girls in the Adolescent ASPD group will show
significantly higher rates of alcohol and drug dependence,
ADHD, and depression than adolescents in the CD Only
and Control groups.
significantly lower verbal IQ and a greater PIQ > VIQ dis-
crepancy than those in the CD Only and Control groups.
significantly greater peer group deviance and poorer aca-
demic functioning (lower grades, more suspensions and
absences) than those in the CD Only and Control groups.
a higher rate of paternal ASPD than the other groups.
Second, we compared the Adolescent ASPD group to a
young adult group who had ASPD using all DSM criteria
(including the age of onset criterion) in order to show that
the ASPD diagnosis conveys similar clinical characteristics
whether it is assigned before or after age 18. This is important
in showing that the ASPD diagnosis could be used before

age 18 to help identify adolescents at risk for following a
more persistent antisocial behavior trajectory.
Method
Participants
Participants were drawn from a sample of 578 16- to 18-
year-old male (M = 17; SD = .45) and 674 female (M = 17;
SD = .65) same-sex monozygotic (MZ) and dizygotic (DZ)
twin pairs participating in the Minnesota Twin Family Study
(MTFS). Twin pairs were identified through Minnesota state
birth records for the years 1972 through 1977 (male ado-
lescent cohort) and 1975 through 1979 (female adolescent
cohort). Exclusion criteria were minimal (twins could not
be adopted nor could they have a physical or intellectual
disability as determined through a detailed phone interview
with the twins’ mother). After complete description of the
study to participants, parents provided informed written con-
sent for their own and their minor child’s participation in the
MTFS. Children under 18 provided written assent to partic-
ipate. Families were paid for their participation. The racial
composition of the entire MTFS sample (98% white) is con-
sistent with the demographics of Minnesota in the birth years
sampled. Twins were recruited for the first follow-up assess-
ment at an average age of 20 (i.e., approximately 3 years after
the intake assessment). Most (88%) of the 1,252 twins com-
pleted diagnostic measures for that follow-up assessment.
Participants were grouped based on diagnoses of CD and
ASPD from clinical interviews conducted at ages 17 (intake)
and 20 (first follow-up). At both ages, symptoms of ASPD
(including symptoms of CD present prior to age 15) were
assessed independently in each twin using the Structured
Clinical Interview for DSM-III-R Personality Disorders

(SCID-II; Spitzer et al., 1987). (The DSM-III-R criteria
were the diagnostic standard at the time of the assessments.)
Highly trained interviewers who had at least a B.A. degree in
psychology conducted all interviews. At age 17 only, symp-
toms of CD were also assessed via the Diagnostic Interview
for Children and Adolescents – Revised (DICA-R-P) (Her-
janic & Reich, 1982; Reich & Welner, 1988) administered to
the twins’ mother in an independent interview. A case con-
ference team of two advanced clinical psychology graduate
students assigned symptoms of disorders after reviewing the
interview data from a particular informant (i.e., symptoms
reported by the mother were assigned independently from
those reported by the twin for any diagnosis with both
informant reports). Separate case conferences were used to
assign symptoms for members of each twin pair and teams
were blind to the co-twin’s diagnostic status and to the pair’s
zygosity during the case conferences. Symptoms assigned
during case conferences were entered into a computer
and algorithms (to implement DSM-III-R criteria) were
employed to produce study diagnoses. When reports from
both the mother and the twin were available, a symptom was
counted toward the diagnosis if either informant endorsed
it, as is typical in a best-estimate diagnostic strategy (Bird,
Gould, & Staghezza, 1992). At age 20, the twin was the sole
informant for all interviews and therefore diagnoses at that
assessment were based solely on the twin’s report.
Seventy-seven participants (25 girls; 32.5%) comprised
the CD Only group, which had a probable or definite (2 or
Springer

more symptoms plus duration) CD diagnosis at age 17 and
no more than one adult antisocial behavior (AAB) symptom
from the ASPD criteria through age 20. Sixty-four partic-
ipants (17 girls; 26.6%) comprised the Adolescent ASPD
group, which had a probable or definite ASPD diagnosis (3
or more AAB symptoms and a probable or definite CD diag-
nosis) at age 17. Twenty men (but no women) were identified
as having probable or definite ASPD when all DSM-III-R
criteria were applied (including age at onset). These men,
who comprised the ASPD group, had CD at age 15 but did
not meet (probable or definite) ASPD criteria until age 20.
The inclusion of probable cases helped maximize the num-
ber of participants classified, but did not greatly inflate the
rates of CD (6.2%) or ASPD (6.7%) in the overall sam-
ple. Table 1 presents the rates of endorsement of CD and
AAB symptoms for each group. Three hundred forty Con-
trols (274 girls; 81%) were identified who had no CD or AAB
symptoms through age 20.
Note that eight of the Adolescent ASPD group members
were 18 at the time of the “age 17” assessment because there
was not enough time in a year for all pairs to visit at age 17.
Analyses were rerun with the eight 18-year-olds removed
with similar findings, therefore, results for the full Adoles-
cent ASPD group are reported. Also, there was a potential
concern about comparing a disproportionately female con-
trol group to two largely male antisocial behavior disorder
groups. Thus, a control group with a similar proportion of
girls to that found in the disordered groups was created by
randomly selecting 50 girls from the 274 control girls and
combining them with the 66 control boys (N = 116; 43%
girls). All analyses were rerun with this control group and
results were quite similar, therefore, the results using the full
Control group (N = 341) are reported.
Approximately 30–60% of each group was comprised of

MZ or DZ twin pairs representing related individuals within
the group: Control (102 pairs or 204 related individuals), CD
Table 1 Symptom characteristics of the conduct disorder only
(CD only), adolescent antisocial personality disorder
(adolescent ASPD), and
antisocial personality disorder (ASPD) groups
CD Only (n = 77) Adolescent ASPD (n = 64) ASPD (n = 20)
Symptom (DSM-III-R) N % N % N %
Conduct disorder (CD)
Stolen without confrontation 14 18.2 37 57.8 12 60.0
Ran away overnight 3 3.9 9 14.1 1 5.0
Often lies 21 27.3 35 54.7 11 55.0
Set fires deliberately 1 1.3 5 7.8 1 5.0
Often truant 6 7.8 29 45.3 6 30.0
Broken into home/car 4 5.2 23 35.9 5 25.0
Destroyed property 26 33.8 32 50.0 14 70.0
Cruel to animals 31 40.3 17 26.6 7 35.0
Used weapon in a fight 16 20.8 13 20.3 6 30.0
Often initiates physical fights 35 45.5 31 48.4 9 45.0
Stolen with confrontation 1 1.3 5 7.8 1 5.0
Cruel to people 4 5.2 17 26.6 5 25.0
Adult Antisocial Behavior (AAB)
Inconsistent work behavior 2 2.6 41 64.1 10 50.0
Fails to conform to social norms 1 1.3 46 71.9 17 35.0
Irritable and aggressive 4 5.2 40 62.5 13 65.0
Fails to honor financial debts 0 0 5 7.8 4 20.0
Fails to plan ahead/impulsive 0 0 3 4.7 1 5.0
No regard for the truth 2 2.6 38 59.4 9 45.0
Reckless disregard for safety 28 36.4 53 82.8 19 95.0
Irresponsible parent 0 0 0 0 0 0
Non-monogamous 0 0 1 1.6 0 0
Lacks remorse 1 1.3 28 43.8 4 20.0

Mean SD Mean SD Mean SD
No. of CD Symptoms 2.56 1.21 4.39 2.09 4.25 1.45
No. of AAB Symptoms 0.49 0.50 3.98 1.02 3.85 0.93
Note. CD Only group members were allowed to have one AAB
symptom at 17 or 20, and the table reflects AAB symptoms
endorsed at either age.
For the Adolescent ASPD group, the table reflects symptoms
reported at age 17, when that group met criteria for ASPD. For
the ASPD group, the
rates for CD are from age 17 and rates for AAB are from age 20
given that they had a CD diagnosis at 17 but did not meet the
AAB component of
ASPD until age 20.
Springer
Only (10 pairs), Adolescent ASPD (11 pairs), and ASPD (3
pairs). As indicated below, steps were taken to account for
this non-independence of observations in the analyses.
Measures
Co-morbid disorders
At ages 17 and 20, twins were independently interviewed
with the expanded Substance Abuse Module (SAM) of the
Composite International Diagnostic Interview (Robins et al.,
1988) to assess lifetime criteria of substance use disorders
including nicotine dependence and abuse and dependence

on alcohol, cannabis, amphetamines, opioids, sedatives, co-
caine, PCP, hallucinogens, and inhalants. Lifetime criteria of
major depression were assessed using the Structured Clinical
Interview for DSM-III-R (SCID; Spitzer et ., 1987) at ages 17
and 20. At age 17, twins reported their symptoms of ADHD
via the DICA-R and mothers reported on the twins’ ADHD,
depression, and substance use disorders via the DICA-R-P.
The case conference procedure described above was used to
assign symptoms of ADHD, depression, and substance use
disorders, and computer algorithms implementing DSM-III-
R criteria were used to determine diagnoses. At age 17, diag-
noses were derived using the aforementioned best estimate
strategy. At age 20, diagnoses were based solely on the twin’s
report. Independent case conference teams (blind to the orig-
inal team’s symptom ratings and to the study diagnoses) rated
a random sample of cases and produced diagnostic reliability
coefficients (kappa) of .75 or greater.
Co-morbid diagnoses were considered present if criteria
were met at a definite diagnostic level. Alcohol, nicotine,
and cannabis use disorders were examined separately from
all other drug classes because they were the most common;
the remaining drugs were combined into an illicit abuse (i.e.,
criteria met for abuse of any illicit drug other than cannabis)
and an illicit dependence (criteria met for dependence on any
illicit drug other than cannabis) variable.
Cognitive functioning
Verbal IQ (VIQ) was estimated from the Information and Vo-
cabulary subtests and performance IQ (PIQ) was estimated
from the Picture Arrangement and Block Design subtests
of the Wechsler Adult Intelligence Scale – Revised (WAIS-
R; Wechsler, 1981) administered at age 17. The PIQ > VIQ
discrepancy reflected the simple difference between PIQ and
VIQ.

Peer group
At age 17, each twin nominated up to four teachers to comp-
lete a teacher rating form modeled after the Connors Teacher
Rating Scale (Connors, 1969; Pelham et al., 1989) and the
Rutter Child Scale B (Rutter, 1967). The Bad Peers scale
included four descriptors (tough, dangerous, rebellious, and
involved with alcohol and drugs) rated on a 5-point scale.
Ratings were averaged across teachers and high scores indi-
cated greater levels of the four descriptors in the twin’s peer
group (Alpha = .77).
Academic functioning
Official school records are not obtained by the MTFS given
the variability in grading procedures (e.g., assignment of
grades versus some other evaluative mark) and standards
(e.g., using something other than a 4-point academic scale)
used across school districts from which twins were drawn.
As such, at age 17, teachers completing the teacher rating
form (described above) also rated twins’ academic achieve-
ment in four areas: English, Math, Science, and Social Stud-
ies. Achievement in each area was rated on an academic
grade scale of A (coded 4) to F (coded 0). Ratings were
averaged across teachers with high scores reflecting better
grades. Teachers also reported the number of school suspen-
sions and unexcused absences, and ratings were averaged
across teachers within each of these variables.
Paternal history of ASPD
At age 17, twins participated in the MTFS with their parents.
Biological fathers were independently interviewed regard-
ing symptoms of ASPD using the SCID-II. The consensus

case conference procedure described above was applied to
diagnostic data from fathers to arrive at study diagnoses of
ASPD. For each twin, paternal history of ASPD was denoted
as either positive (lifetime probable or definite diagnosis of
ASPD present in biological father) or negative (no lifetime
diagnosis of ASPD in biological father). (Our decision to
focus on paternal ASPD was a practical one based on initial
examination of the data that showed that only four biologi-
cal mothers of adolescents assigned to any of the groups met
criteria for ASPD thus precluding an analysis of maternal
ASPD effects.)
Analyses
In the first test of the study hypothesis, the Adolescent ASPD,
CD Only, and Control groups were compared on rates of
co-morbid diagnoses at age 17 using chi-square tests with
one twin from each twin pair within a group randomly se-
lected and removed prior to analysis to meet the requirement
of independence of observations. To correct for the number
of tests conducted, alpha was set to .01 for the omnibus
chi-square tests. Alpha was set to .05 (1-tailed) for follow-up
chi-square contrasts following significant omnibus tests.
Springer
The three groups were also compared on VIQ score and
PIQ > VIQ discrepancy, Bad Peers scale scores, and aca-
demic achievement in analyses using the Mixed procedure
in SPSS 12.0 which provided a multilevel framework that
allowed twins to be nested within families and thus allowed
all cases to be included in the analyses. A mixed linear model

was fit for each cognitive variable using restricted maximum
likelihood estimation. Models included fixed effects in a full
factorial model with tests for group, gender, and the group ×
gender interaction. The effect of twins nested within families
was modeled as a random effect in each model. Alpha was
set to .01 for the omnibus F-tests, and follow-up contrasts to
significant effects were conducted using the Bonferroni test
with alpha set to .05 (1-tailed). Initial examination of the sus-
pensions and unexcused absences variables revealed highly
positively skewed distributions that were not amenable
to transformations to improve normality. Thus, these two
variables were examined using non-parametric Kruskal
Wallis omnibus and follow-up tests. Such tests do not lend
themselves to nesting and, therefore, one twin was removed
from each pair within a group for the analyses. Alpha was
set to .025 (.05/2) for the omnibus tests and .05 for follow-up
tests. Finally, groups were compared on paternal ASPD
history using a chi-square test with one twin from each twin
pair within a group removed and alpha was set to .05.
The second test of the study hypothesis entailed a compar-
ison of the Adolescent ASPD and the ASPD groups. Groups
were first compared on their rates of endorsement of each
CD and AAB symptom using chi-square tests with co-twins
removed as described above. Groups were then compared on
rates of co-morbid disorders at age 17 (when the Adolescent
ASPD group first met criteria for ASPD) and at age 20 (when
the ASPD group first met criteria for ASPD) using chi-square
tests with co-twins removed as indicated previously. This
strategy afforded an examination of the clinical problems
that precede, occur with, and follow an ASPD diagnosis.
The two groups were then compared on VIQ and PIQ > VIQ
discrepancy, Bad Peers scores, and academic achievement
using the Mixed procedure in SPSS 12.0. Models included
a fixed effect for group and the effect of twins nested within
families as a random effect (there were no gender effects in

the models because the ASPD group was all male). Next, dif-
ferences in school suspensions and unexcused absences were
examined with Mann-Whitney non-parametric tests with co-
twins removed. Finally, groups were compared on paternal
ASPD history using a chi-square test with co-twins removed.
The alpha for all tests was set to .05 (2-tailed).
Results
The Adolescent ASPD group was distinct from the CD
Only group in most analyses, largely supporting our hy-
potheses. As expected, the omnibus chi-square test was
significant (p < .001) for each disorder: ADHD (χ 2 [2]
= 24.13), depression (χ 2 [2] = 38.81), alcohol abuse (χ 2
[2] = 220.06), alcohol dependence (χ 2 [2] = 153.41),
nicotine dependence (χ 2 [2] = 164.73), cannabis abuse (χ 2
[2] = 125.48), cannabis dependence (χ 2 [2] = 105.75),
illicit drug abuse (χ 2 [2] = 46.15), and illicit drug depen-
dence (χ 2 [2] = 28.51). Fig. 1 presents the percent of each
group with a co-morbid diagnosis at age 17. As expected,
the Adolescent ASPD group had a significantly (p < .001)
greater rate of depression and all substance use disorders
than the CD Only and Control groups (which did not dif-
fer significantly). Both the CD Only and Adolescent ASPD
group differed significantly (p < .007) from Controls (but
not from each other) on rate of ADHD. (To investigate pos-
sible gender differences, analyses were rerun separately by
gender and results were unchanged although a few effects
were significant at p < .05 instead of p < .01 likely due to
the reduced power of the within-gender analyses.)
Table 2 presents a summary of the group comparisons
on the cognitive, peer, academic, and paternal ASPD vari-
ables. Some participants were missing data on some of the
variables resulting in the variable df for tests reported in the

table. The group × gender interaction was non-significant
for VIQ, PIQ > VIQ, Bad Peers, and all academic achieve-
ment variables. The main effect for group was significant
(p < .001) for all variables except VIQ, English, and Math
(the latter two approached significance at p < .02, which
was just short of the adopted alpha of .01). As expected,
the Adolescent ASPD group was significantly (p < .05) dif-
ferent from the Control and CD Only groups (which were
not significantly different) on PIQ > VIQ, Bad Peers, and
Science grades. The Adolescent ASPD group differed sig-
nificantly only from Controls on Social Studies grades and
paternal history of ASPD. The Kruskal Wallis test was sig-
nificant at p < .02 for school suspensions (χ 2 [2] = 8.05)
and for unexcused absences (χ 2 [2] = 35.01). As expected,
the Adolescent ASPD group was significantly higher in mean
rank than the CD Only and Control groups, who differed sig-
nificantly (p = .048) only on unexcused absences (with the
CD Only group having a higher mean rank than the Control
group).
The second test also supported our hypothesis in that the
Adolescent ASPD group was not significantly different from
the ASPD group in most analyses. As expected, groups did
not differ significantly on their rates of endorsement of the
individual CD or AAB symptoms (see Table 1), suggesting
that adolescents and young adults diagnosed with ASPD
endorse similar kinds of symptoms. Also as expected, the
groups did not differ significantly on any diagnosis at age
20 and differed at age 17 on alcohol abuse (χ 2 [1] = 8.90,
p = .003), nicotine dependence (χ 2 [1] = 9.37, p = .002),
and cannabis (χ 2 [1] = 5.49, p = .02) where the Adolescent
Springer

0
10
20
30
40
50
60
70
80
90
100
AD
HD
De
pr
es
sio
n
Al

co
ho
l A
bu
se
Al
co
ho
l D
ep
en
de
nc
e
Ni
co
tin
e
De
pe
nd
en
ce

Ca
nn
ab
is
Ab
us
e
Ca
nn
ab
is
De
pe
nd
en
ce
Illi
cit
A
bu
se
Illi
cit

D
ep
en
de
nc
e
%
o
f
G
ro
u
p
w
ith
D
ia
g
n
o
si
s
Control (n = 237-8) CD Only (n = 76-7) Adolescent ASPD (n =
51-3)
Fig. 1 Rate of co-morbid

diagnoses in the control,
conduct disorder (CD) only, and
adolescent antisocial personality
disorder (Adolescent ASPD)
groups at age 17. Data were
missing for a few cases on one
or more disorders resulting in
slightly variable group Ns.
ADHD = attention-deficit/
hyperactivity disorder. Illicit
drug abuse and dependence
refer to abuse and dependence
on illicit drugs other than
cannabis. The vertical line with
crossbars represents the 95% CI
for the percent estimate
ASPD group had the higher rates of co-morbidity. Figure 2
presents the rate of co-morbid diagnoses for each group at
ages 17 and 20.
As expected, the two ASPD groups did not differ signifi-
cantly on any of the cognitive, peer, academic achievement,
or paternal history variables. Table 2 presents a summary
of those analyses. Note that there were two members of the
Adolescent ASPD group with co-twins in the ASPD group,
which might bias the result toward our null expectation given
that members of a twin pair have identical data for paternal
history. The analysis for the paternal ASPD history vari-
able was re-run after removing the two Adolescent ASPD
twins who had ASPD co-twins and the results were nearly
identical (the rate of paternal ASPD went from 21.7% in
the Adolescent ASPD group to 20.5% and the χ 2 remained
non-significant). Finally, the Adolescent ASPD and ASPD
groups did not differ significantly on school suspensions or

unexcused absences.
Discussion
Adolescent antisocial behavior is an important topic of re-
search given its negative consequences to both families and
Table 2 Summary of group comparisons on cognitive
functioning, peer group, academic achievement, and paternal
ASPD history
Control vs. CD Only vs. Adolescent ASPD Adolescent ASPD
vs. ASPD
Control CD Only Adolescent ASPD ASPD
F (df) Mean (SE) Mean (SE) Mean (SE) F (df) Mean (SE)
VIQ 2.67 (2, 472.8) 98.18 (0.99) 95.80 (1.59) 93.47 (1.96) 0.16
(1, 66.6) 92.62 (2.74)
PIQ > VIQ 73.36 (2, 439.9)∗ ∗ ∗ 7.64 (1.25)a 8.79 (2.13)a
18.44 (2.55)b 0.30 (1, 67.1) 21.10 (3.64)
Bad Peers 51.51 (2, 313.2)∗ ∗ ∗ 0.14 (0.10)a 0.15 (0.02)a 0.43
(0.03)b 0.56 (1, 52.0) 0.39 (0.05)
Achievement
English 4.19 (2, 270.3) 3.15 (0.06) 3.09 (0.10) 2.54 (0.19) 1.56
(1, 38.8) 1.92 (0.25)
Math 3.94 (2, 226.3) 2.98 (0.08) 2.96 (0.15) 2.32 (0.23) 0.41 (1,
38.0) 2.15 (0.23)
Science 10.33 (2, 241.0)∗ ∗ ∗ 3.07 (0.08)a 2.85 (0.13)a 2.07
(0.21)b 0.52 (1, 2.5) 2.09 (0.20)
Social
Studies
6.69 (2, 261.3)∗ ∗ ∗ 3.19 (0.07)a 2.91 (0.12)a,b 2.55 (0.18)b
2.61 (1, 40.8) 1.94 (0.23)

χ 2 (df) % (95% CI) % (95% CI) % (95% CI) χ 2 (df) % (95%
CI)
Paternal
ASPD
12.79 (2)∗ ∗ 5.5a(2.89–9.46) 9.5a,b(3.58–19.59) 21.7b(10.95–
36.36) 0.15 (1) 16.7(2.09–48.41)
Note. Groups within a row that do not share a superscript
differed at p < .05. For the Adolescent ASPD vs. ASPD
comparisons, none of the effects
were statistically significant (all ps > .11). The Bad Peers
scores were log-transformed (log10 [x + 1]) for analyses and
means are presented in
log-transformed units. The academic achievement variables
were rated on a grading scale of A (4) to F (0). CD = conduct
disorder; ASPD =
antisocial personality disorder; VIQ = verbal IQ; PIQ =
performance IQ; Paternal ASPD refers to presence vs. absence
of a lifetime diagnosis
of ASPD in the twin’s biological father. The entries in the table
reflect the percent of each group with ASPD present in the
biological father.
∗ ∗ p < .002; ∗ ∗ ∗ p < .001.
Springer
0
10
20
30

40
50
60
70
80
90
100
AD
HD
De
pr
es
sio
n
Al
co
ho
l A
bu
se
Al
co
ho
l D
ep

en
de
nc
e
Ni
co
tin
e
De
pe
nd
en
ce
Ca
nn
ab
is
Ab
us
e
Ca
nn
ab

is
De
pe
nd
en
ce
Illi
cit
A
bu
se
Illi
cit
D
ep
en
de
nc
e
%
o
f
G

ro
u
p
w
ith
D
ia
g
n
o
si
s
Adolescent ASPD (n = 51-3) ASPD (n = 17)
B) Co-morbid Diagnoses at Age 20
A) Co-morbid Diagnoses at Age 17
0
10
20
30
40
50

60
70
80
90
100
De
pr
es
sio
n
Al
co
ho
l A
bu
se
Al
co
ho
l D
ep
en

de
nc
e
Ni
co
tin
e
De
pe
nd
en
ce
Ca
nn
ab
is
Ab
us
e
Ca
nn
ab
is

De
pe
nd
en
ce
Illi
cit
A
bu
se
Illi
cit
D
ep
en
de
nc
e
%
o
f
G
ro
u

p
w
ith
D
ia
g
n
o
si
s
Adolescent ASPD (n = 35-9) ASPD (n = 16-7)
Fig. 2 Rate of co-morbid diagnoses in the adolescent antisocial
per-
sonality disorder (Adolescent ASPD) and antisocial personality
disor-
der (ASPD) groups at age 17 and age 20. Data were missing for
a few
cases on one or more disorders resulting in slightly variable
group Ns.
ADHD = attention-deficit/hyperactivity disorder. Illicit drug
abuse
and dependence refer to abuse and dependence on illicit drugs
other
than cannabis. The vertical line with crossbars represents the
95% CI
for the percent estimate
society. The results of the present study suggest that Adoles-

cent ASPD is a valid construct that could identify adolescents
with non-normative antisocial behavior. This might better
enable parents, clinicians, and teachers to reach out early to
adolescents who might be on a more persistent pathway for
antisocial behavior.
Adolescent ASPD was found to be distinct from CD in
terms of co-morbid psychiatric diagnoses, cognitive func-
tioning, peer deviance, and academic performance. More-
over, because no evidence of an interaction between gender
and group was found, Adolescent ASPD appears to be a
valid construct for both genders. Thus, loosening the DSM
criteria for ASPD to allow diagnosis among adolescents
could provide a means to identify adolescents at risk for
persistent antisocial behavior.
Consistent with the large body of literature linking CD
to other childhood behavioral disorders, both CD and Ado-
lescent ASPD were significantly associated with ADHD in
Springer
this study. But, as predicted, boys and girls with Adolescent
ASPD had significantly higher rates of depression and sub-
stance use disorders as compared to those with CD Only.
However, the present findings are not inconsistent with pre-
vious studies relating CD to substance use disorders (e.g.,
Disney et al., 1999) as members of the Adolescent ASPD
group also had CD. If groups had been selected simply on
CD status, then the “CD group” would have had high rates
of substance use disorders, but our results suggest that the

elevated rates would have been attributable to the members
with Adolescent ASPD. This suggests that the assessment of
ASPD in adolescents could identify those at greatest risk for
substance use disorders or, stated another way, restricting the
ASPD diagnosis to those over age 18 may result in a missed
opportunity at identifying antisocial adolescents that are at
high risk for early-onset substance use disorders.
An important finding was that, as expected, Adolescent
ASPD and CD were distinct in many ways. Equally impor-
tant was the finding that an ASPD diagnosis—regardless
of when it was assigned—was associated with substantial
rates of co-morbid psychiatric disorder in adolescence and
early adulthood. In addition, the Adolescent ASPD and ASPD
groups showed a similar PIQ > VIQ discrepancy, the mag-
nitude of which ( >18 points) is considered not only statis-
tically significant but also abnormal (Hsu et al., 2000) and
corroborates the idea that cognitive deficits are associated
with persistent forms of antisocial behavior (Moffitt, 1993).
In addition, the Adolescent ASPD and ASPD groups showed
similarly deviant peer groups and similarly poorer academic
functioning (in both achievement and discipline) at age 17.
These results suggest that cognitive deficits in verbal rela-
tive to performance IQ, affiliation with deviant peers, and
poorer academic functioning may contribute to the devel-
opment and/or maintenance of persistent antisocial behavior
that could be identified in adolescents by applying ASPD
criteria.
Our expectation for a statistically significant difference
between the CD Only and Adolescent ASPD groups on pater-
nal ASPD history was not supported. However, the data were
ordered in the expected direction, with the rate of paternal
ASPD for the adolescent ASPD group more than double that
seen in the CD Only group. Only the Adolescent ASPD group
showed a high rate of substance use disorders—another form

of externalizing disorder. Thus, the biological risk transmit-
ted in families may primarily manifest as an antisocial be-
havior disorder in children (CD or ASPD), but other factors
(e.g., deviant peer group, cognitive functioning) then con-
tribute to the continuity of the antisocial behavior (CD vs.
ASPD) and perhaps also to the likelihood that other external-
izing disorders manifest (e.g., substance use disorders; see
Krueger et al., 2002).
The major strengths of this study included the use of a
mixed gender sample with groups carefully selected on CD
and ASPD diagnostic status. Analyses of gender were con-
ducted wherever possible and showed similar results for boys
and girls. Structured clinical interviews provided excellent
measurement of CD, ASPD, and co-morbid disorders. The
use of teachers as raters of peer group and academic func-
tioning served to avoid the complete reliance on self-reports
that plagues many studies. It also likely resulted in less biased
information than might have been obtained through self or
even parent reports on those measures. Finally, the use of lon-
gitudinal data allowed us to compare individuals diagnosed
with ASPD in adolescence to those who met the criteria in
early adulthood. The longitudinal data also allowed us to
examine prospectively the co-morbidity between Adolescent
ASPD, depression, and substance use disorders.
Limitations of the study also existed and included the use
of a predominantly white sample, indicating that caution be
used in generalizing our results to non-white populations.
Certainly, additional work is needed to confirm and extend
the present findings. In addition, though the rates of CD and
ASPD were consistent with those found in the general popu-
lation, our ASPD group size was modest. Thus, confidence in
our results would be enhanced with replications using larger
samples (e.g., obtained from clinics where the disorders are

found at higher rates). Such replications would also bolster
confidence in the predicted null findings given that null re-
sults could arise from actual similarity across groups but
also from methodological or statistical anomalies. Finally,
we acknowledge that teachers might not be able to provide
the most comprehensive rating of peer group given their lim-
ited context of interaction with the adolescents. Moreover, it
is possible that the teacher’s perceptions of the peer group
could affect their rating of academic performance (e.g., per-
ception of a negative peer group might lead to more negative
academic evaluations).
In summary, this study suggests that Adolescent ASPD is
a valid construct that appears to provide incremental clinical
utility beyond what is gleaned from a CD diagnosis alone.
Given the high prevalence of antisocial behavior during ado-
lescence, it may be difficult for parents and professionals
alike to determine whether an adolescent is exhibiting nor-
mative, transitory antisocial behavior or whether he/she is
at risk for a more persistent course. The application of the
ASPD criteria to adolescents under the age of 18 appears to
identify a clinically important subgroup of antisocial ado-
lescents who might benefit most from interventions to help
deter them from a continued course of antisocial behav-
ior. Such applications, however, should be made cautiously
until further research confirms the present findings given the
potential dangers that diagnostic labels can bring.
Acknowledgements The present study was supported in part by
National Institute of Drug Abuse grant DA 05147 and National
In-
stitute on Alcohol Abuse and Alcoholism grant AA09367.
Springer

References
Barry CT, Frick PJ, DeShazo TM, McCoy MG, Ellis M, Loney
BR
(2000) The importance of callous-unemotional traits for extrnd-
ing the concept of psychopathy to children. J Abnorm Psychol
109:335–340
Bird HR, Gould MS, Staghezza B (1992) Aggregating data from
multi-
ple informants in child psychiatry epidemiological research. J
Am
Acad Child Adolesc Psychiatry 31:78–85
Blair RJR (1999) Responsiveness to distress cues in the child
with psychopathic tendencies. Pers Individ Differences 27:135–
145
Christian RE, Frick PJ, Hill NL, Tyler L, Frazer DR (1997)
Psychopathy
and conduct problems: II. Implications for subtyping children
with
conduct problems. J Am Acad Child Adolesc Psychiatry
36:233–
241
Conners CK (1969) A teacher rating scale for use in drug
studies with
children. Am J Psychiatry 126:152–156
Cornell D, Wilson LA (1992) The PIQ > VIQ discrepancy in
violent
and nonviolent delinquents. J Clin Psychol 48:256–261

Cottler LB, Price RK, Compton WM, Mager DE (1995)
Subtypes of
adult antisocial Behavior among drug abusers. J Nerv Ment Dis
183:154–161
Disney ER, Elkins IJ, McGue M, Iacono WG (1999) Effects of
ADHD,
conduct disorder, and gender on substance use and abuse in ado-
lescence. Am J Psychiatry 156:1515–1521
Farrington DP, Hawkins JD (1991) Predicting participation,
early onset
and later persistence in officially recorded offending. Criminal
Behav Ment Health 1:1–33
Frick PJ (2002) Juvenile psychopathy from a developmental
perspec-
tive: Implications for construct development and use in forensic
assessments. LawHum Behav 26:247–253
Frick PJ, Ellis M (1999) Callous-unemotional traits and
subtypes of
conduct disorder. Clin Child Fam Psychol Rev 2:149–168
Frick PJ, Lilienfeld SO, Ellis M, Loney B, Silverthorn P (1999)
The
association between anxiety and psychopathy dimensions in
chil-
dren. J Abnorm Child Psychol 27:383–392
Helzer JE, Pryzbeck TR (1988) The co-occurrence of alcoholism
with
other psychiatric disorders in the general population and its
impact
on treatment. J Stud Alcohol 49:219–224

Herjanic B, Reich W (1982) Development of a structured
psychiatric
interview for children: Agreement between child and parent on
individual symptoms. J Abnorm Child Psychol 10:307–324
Hesselbrock M (1991) Gender comparison of antisocial
personality
disorder and depression in alcoholism. J Subst Abuse 3:205–
220
Hsu LM, Hayman J, Koch J, Mandell D (2000) Relation of
statistically
significant, abnormal, and typical WAIS-R VIQ-PIQ
discrepancies
to full scale IQs. Eur J Psychol Assess 16:107–114
Jessor R, Van Den Bos J, Vanderryn J, Costa FM, Turbin MS
(1995)
Protective factors in adolescent problem behavior: Moderator
ef-
fects and developmental change. Dev Psychol 31:923–933
Kendler KS, Prescott CA, Myers J, Neale MC (2003) The
structure of
genetic and environmental risk factors for common psychiatric
and
substance use disorders in men and women. Arch Gen
Psychiatry
60:929–937
Krueger RF, Hicks BM, Patrick CJ, Carlson SR, Iacono WG,
McGue
M (2002) Etiologic connections among substance dependence,
antisocial behavior, and personality: Modeling the externalizing
spectrum. J Abnorm Psychol 111:411–424

Lahey BB, Loeber R, Hart EL, Frick PJ, Applegate B, Zhang Q,
Green
SM, Russo MF (1995) Four-year longitudinal study of conduct
disorder in boys: Patterns and predictors of persistence. J
Abnorm
Psychol 104:83–93
Lynam DR (2002) Fledgling psychopathy: A view from
personality
theory. Law Hum Behav 26:255–259
Moffitt TE (1993) Adolescence-limited and life-course-
persistent an-
tisocial behavior: A developmental taxonomy. Psychol Bull
100:674–701
Moffitt TE, Caspi A (2001) Childhood predictors differentiate
life-
course persistent and adolescence-limited antisocial pathways in
males and females. Dev Psychopathol 13:355–376
Moffitt TE, Caspi A, Rutter M, Silva PA (2001) Sex differences
in
antisocial behavior (pp. 135–150). Cambridge University Press,
Cambridge, UK
O’Brien BS, & Frick PJ (1996) Reward dominance:
Associations with
anxiety, conduct problems, and psychopathy in children. J
Abnorm
Child Psychol 24:223–240
Pajer KA (1998) What happens to “bad” girls? A review of the
adult
outcomes of antisocial adolescent girls. Am J Psychiatry

155:862–
870
Patterson GR (1986) Performance models for antisocial boys.
Am Psy-
cholozist 41:432–444
Pelham WE, Milich R, Murphy DA, Murphy HA (1989)
Normative
data on the IOWA Conners Teacher Rating Scale. J Clin Child
Psychol 18:259–262
Reich W, Welner Z (1988) Diagnostic Interview for Children
and
Adolescents—Revised: DSM-III-R Version (DICA-R).
Washing-
ton University, St. Louis.
Robins LN (1966) Deviant Children Grown Up. Williams &
Wilkins:
Baltimore, MD
Robins LN, Wing J, Wittchen HU, Helzer JE, Babor TF, Burke
J,
Farmer A, Jablenski A, Pickens R, Regier DA, Sartorius N,
Towle
LH (1988) The Composite International Diagnostic Interview:
An epidemiologic instrument suitable for use in conjunction
with
different diagnostic systems and different cultures. Arch Gen
Psy-
chiatry 45:1068–1077
Rutter M (1967) A children’s behaviour questionnaire for
completion
by teachers: Preliminary findings. J Child Psychol Psychiatry

8:1–
11
Schuckit MA (1986) Genetic and clinical implications of
alcoholism
and affective disorder. Am J Psychiatry 143(2):140–147
Seagrave D, Grisso T (2002) Adolescent development and the
mea-
surement of juvenile psychopathy. Law Hum Behav 26:219–239
Simonoff E, Elander J, Holmshaw J, Pickles A, Murray R,
Rutter M (2004) Predictors of antisocial personality:
Continuities
from childhood to adult life. Br J Psychiatry 184:118–127
Simons RL, Wu C-I, Conger RD, Lorenz FO (1994) Two routes
to
delinquency: Differences between early and late starters in the
impact of parenting and deviant peers. Criminology 32:247–275
Snow M, Thurber S (1997) Cognitive imbalance and antisocial
person-
ality characteristics. J Clin Psychol 53:351–354
Spitzer RL, Williams JB, Gibbon M (1987) Structured clinical
interview
for DSM-III-R. Biometrics Research Department, New York
State
Psychiatric Institute, New York
Spitzer RL, Williams JB, Gibbon M, First MB (1987) Structured
clinical
interview for DSM-III-R personality disorders (SCID-II).
Biomet-
rics Research Department, New York State Psychiatric Institute,

New York
Taylor J, Carey G (1998) Antisocial behavior, substance abuse,
and
somatization in families of adolescent drug abusers and
adolescent
controls. Am J Drug Alcohol Abuse 24:635–646
Wechsler D (1981) Manual for the Wechsler Adult Intelligence
Scale –
Revised (WAIS-R). The Psychological Corporation, San
Antonio,
TX
White JL, Moffitt TE, Caspi A, Jeglum-Bartusch D, Needles DJ,
Stouthamer-Loeber M (1994) Measuring impulsivity and exam-
ining its relationship to delinquency. J Abnorm Psychol
103:192–
205
Springer
Reproduced with permission of the copyright owner. Further
reproduction prohibited without permission.
<<
/ASCII85EncodePages false
/AllowTransparency false
/AutoPositionEPSFiles true
/AutoRotatePages /None
/Binding /Left
/CalGrayProfile (None)
/CalRGBProfile (sRGB IEC61966-2.1)
/CalCMYKProfile (ISO Coated)

/sRGBProfile (sRGB IEC61966-2.1)
/CannotEmbedFontPolicy /Error
/CompatibilityLevel 1.3
/CompressObjects /Off
/CompressPages true
/ConvertImagesToIndexed true
/PassThroughJPEGImages true
/CreateJDFFile false
/CreateJobTicket false
/DefaultRenderingIntent /Perceptual
/DetectBlends true
/ColorConversionStrategy /sRGB
/DoThumbnails true
/EmbedAllFonts true
/EmbedJobOptions true
/DSCReportingLevel 0
/SyntheticBoldness 1.00
/EmitDSCWarnings false
/EndPage -1
/ImageMemory 524288
/LockDistillerParams true
/MaxSubsetPct 100
/Optimize true
/OPM 1
/ParseDSCComments true
/ParseDSCCommentsForDocInfo true
/PreserveCopyPage true
/PreserveEPSInfo true
/PreserveHalftoneInfo false
/PreserveOPIComments false
/PreserveOverprintSettings true
/StartPage 1
/SubsetFonts false
/TransferFunctionInfo /Apply
/UCRandBGInfo /Preserve
/UsePrologue false

/ColorSettingsFile ()
/AlwaysEmbed [ true
]
/NeverEmbed [ true
]
/AntiAliasColorImages false
/DownsampleColorImages true
/ColorImageDownsampleType /Bicubic
/ColorImageResolution 150
/ColorImageDepth -1
/ColorImageDownsampleThreshold 1.50000
/EncodeColorImages true
/ColorImageFilter /DCTEncode
/AutoFilterColorImages false
/ColorImageAutoFilterStrategy /JPEG
/ColorACSImageDict <<
/QFactor 0.76
/HSamples [2 1 1 2] /VSamples [2 1 1 2]
>>
/ColorImageDict <<
/QFactor 0.76
>>
/JPEG2000ColorACSImageDict <<
/TileWidth 256
/TileHeight 256
/Quality 30
>>
/JPEG2000ColorImageDict <<
/TileWidth 256
/TileHeight 256
/Quality 30
>>
/AntiAliasGrayImages false
/DownsampleGrayImages true
/GrayImageDownsampleType /Bicubic

/GrayImageResolution 150
/GrayImageDepth -1
/GrayImageDownsampleThreshold 1.50000
/EncodeGrayImages true
/GrayImageFilter /DCTEncode
/AutoFilterGrayImages true
/GrayImageAutoFilterStrategy /JPEG
/GrayACSImageDict <<
/QFactor 0.76
>>
/GrayImageDict <<
/QFactor 0.15
>>
/JPEG2000GrayACSImageDict <<
/TileWidth 256
/TileHeight 256
/Quality 30
>>
/JPEG2000GrayImageDict <<
/TileWidth 256
/TileHeight 256
/Quality 30
>>
/AntiAliasMonoImages false
/DownsampleMonoImages true
/MonoImageDownsampleType /Bicubic
/MonoImageResolution 600
/MonoImageDepth -1
/MonoImageDownsampleThreshold 1.50000
/EncodeMonoImages true
/MonoImageFilter /CCITTFaxEncode
/MonoImageDict <<
/K -1
>>

/AllowPSXObjects false
/PDFX1aCheck false
/PDFX3Check false
/PDFXCompliantPDFOnly false
/PDFXNoTrimBoxError true
/PDFXTrimBoxToMediaBoxOffset [
0.00000
0.00000
0.00000
0.00000
]
/PDFXSetBleedBoxToMediaBox true
/PDFXBleedBoxToTrimBoxOffset [
0.00000
0.00000
0.00000
0.00000
]
/PDFXOutputIntentProfile (None)
/PDFXOutputCondition ()
/PDFXRegistryName (http://www.color.org?)
/PDFXTrapped /False
/Description <<
/DEU
<FEFF004a006f0062006f007000740069006f006e007300200066
006f00720020004100630072006f0062006100740020004400690
07300740069006c006c0065007200200036002e000d0050007200
6f006400750063006500730020005000440046002000660069006
c0065007300200077006800690063006800200061007200650020
007500730065006400200066006f0072002000640069006700690
0740061006c0020007000720069006e00740069006e0067002000
61006e00640020006f006e006c0069006e0065002000750073006
100670065002e000d002800630029002000320030003000340020
0053007000720069006e006700650072002d005600650072006c0
061006700200047006d0062004800200061006e00640020004900

6d007000720065007300730065006400200047006d00620048000
d000d0054006800650020006c0061007400650073007400200076
0065007200730069006f006e002000630061006e0020006200650
0200064006f0077006e006c006f00610064006500640020006100
7400200068007400740070003a002f002f00700072006f00640075
006300740069006f006e002e0073007000720069006e006700650
072002e00640065002f007000640066002f000d00540068006500
72006500200079006f0075002000630061006e00200061006c007
3006f002000660069006e00640020006100200073007500690074
00610062006c006500200045006e0066006f00630075007300200
05000440046002000500072006f00660069006c00650020006600
6f0072002000500069007400530074006f0070002000500072006
f00660065007300730069006f006e0061006c0020003600200061
006e0064002000500069007400530074006f00700020005300650
0720076006500720020003300200066006f007200200070007200
650066006c00690067006800740069006e006700200079006f007
500720020005000440046002000660069006c0065007300200062
00650066006f007200650020006a006f006200200073007500620
06d0069007300730069006f006e002e>
/ENU
<FEFF004a006f0062006f007000740069006f006e007300200066
006f00720020004100630072006f0062006100740020004400690
07300740069006c006c0065007200200036002e000d0050007200
6f006400750063006500730020005000440046002000660069006
c0065007300200077006800690063006800200061007200650020
007500730065006400200066006f0072002000640069006700690
0740061006c0020007000720069006e00740069006e0067002000
61006e00640020006f006e006c0069006e0065002000750073006
100670065002e000d002800630029002000320030003000340020
0053007000720069006e00670065007200200061006e006400200
049006d007000720065007300730065006400200047006d006200
48>
>>
>> setdistillerparams
<<
/HWResolution [2400 2400]

/PageSize [2834.646 2834.646]
>> setpagedevice
Document 1 of 1
The construct validity of the extraversion subscales of the
Myers-Briggs type indicator
Author: Zumbo, Bruno D; Taylor, Shannon V
ProQuest document link
Abstract (Abstract): It seems that the MBTI scales formed a
factor of their own and that the EPQ combined with
Sociability and Impulsiveness to create the third factor. This
comes as no surprise since the EPQ seems
intuitively to define extraversion in terms of these two
variables. The fact that the MBTI and the EPQ occurred
as separate factors suggests that these two measures do not hold
identical views on extraversion. However, as
evidenced by the factor correlation matrix, Factor 1 (MBTI) and
Factor 3 (SOC/EPQ/IMP) had a fairly high
correlation. From this one could conclude that although each
measure is unique, each contains a high degree of
sociability and impulsiveness in its definition. The interfactor
correlation among factors one and three was in
support of Campbell &Heller (1987). The substantial correlation
among factors one and three, and the factor
loadings of the SOC, EPQ, and IMP suggested that the MBTI
was primarily a sociability measure with a minimal
association with impulsivity. This was also in support of
Campbell &Heller. The second factor (INF/ANX) yielded
only low negative correlations with the other factors and it
seems therefore that although it took up 14.5% of the
variance, it bears little relation to the other factors. Finally,

Factor 4, Dominance, which was responsible for only
4.2% of the variance, obtained only a low loading on the MBTI
factor in the three - factor solution and was
relegated to its own factor in the final four - factor solution.
From this first phase of the results it is clear that
extraversion cannot be described as entirely unitary. It would
appear to be defined in these measures by
sociability, impulsiveness, a lack of inferiority and anxiety and
the presence of dominance. Part 2 of the results
made apparent the possibility of a single higher - order factor
capable of bridging the gap between the original
four factors. The results did, in fact, indicate that these can be
brought together into one interpretable, though
less convincing higher - order factor. We may ask at this point
whether the results were out of the ordinary and
this question is difficult to answer since the study was unique in
its approach. To the authors' knowledge, there
have been no scale - level analyses of extraversion, though
many item - level analyses exist. Such item
analyses have revealed that extraversion is composed of many
factors. For example, Sipps &Alexander (1987)
reported finding three factors of extraversion which were
sociability, impulsivity/non - planning and risk - taking /
jocularity while [Browne, J.A.] (1971) reported twelve.
Certainly, it would be safe to say that sociability has never
been far away from any set of results. Apart from being present
in most factor analyses of extraversion, it has
been shown that the extraversion scale of the MBTI is largely a
measure of sociability (Sipps &DiCaudo, 1988).
Given that the present study declared MBTI extraversion as the
most salient factor, one must wonder about the
strength of the relationship between extraversion and
sociability.
Links: Obtain full text from Shapiro Library
Full text: Abstract
This paper examined the construct validity of the extraversion

and introversion subscales of the Myers - Briggs
Type Indicator (MBTI) with the Eysenck Personality
Questionnaire (EPQ) and the Howarth Personality
Questionnaire (HPQ). We further explored the accompanying
issue of the unitary nature of extraversion. A
conjoint factor analysis undertaken at the scale level produced
four first - order factors: Factor 1 - MBTI
extraversion/introversion, Factor 2 - Inferiority/ Anxiety, Factor
3 - Sociability/EPQ/Impulsiveness and Factor 4 -
Dominance. A subsequent higher - order factor analysis
revealed one factor. The results suggested that the
MBTI subscales were largely a measure of sociability and that
extraversion as represented in these three
measures is not, at first glance, unitary although interfactor
correlations do exist in varying degrees.
Resume
http://ezproxy.snhu.edu/login?url=http://search.proquest.com/do
cview/220473870?accountid=3783
http://pn8vx3lh2h.search.serialssolution.com/?ctx_ver=Z39.88-
2004&ctx_enc=info:ofi/enc:UTF-
8&rfr_id=info:sid/ProQ:psychologyshell&rft_val_fmt=info:ofi/f
mt:kev:mtx:journal&rft.genre=article&rft.jtitle=Canadian%20Jo
urnal%20of%20Behavioural%20Science&rft.atitle=The%20cons
truct%20validity%20of%20the%20extraversion%20subscales%2
0of%20the%20Myers-
Briggs%20type%20indicator&rft.au=Zumbo,%20Bruno%20D;Ta
ylor,%20Shannon%20V&rft.aulast=Zumbo&rft.aufirst=Bruno&r
ft.date=1993-10-
01&rft.volume=25&rft.issue=4&rft.spage=590&rft.isbn=&rft.bt
itle=&rft.title=Canadian%20Journal%20of%20Behavioural%20S
cience&rft.issn=0008400X&rft_id=info:doi/
Le present document porte sur la validite conceptuelle des sous
- echelles d'extraversion et d'introversion de

l'indicateur de type de Myers - Briggs (MBTI), ainsi que sur le
questionnaire de personnalite d'Eysenck (EPQ) et
sur le questionnaire de personnalite de Howarth (HPQ). La
question connexe de la nature unitaire de
l'extraversion y est egalement examinee. Il est ressorti d'une
analyse factorielle mixte de differentes echelles
quatre facteurs de rang un: facteur 1 - extraversion/introversion
MBTI, facteur 2 - inferiorite/anxiete, facteur 3 -
sociabilite/EPQ/impulsivite et facteur 4 - dominance. Un facteur
de rang superieur est ressorti d'une analyse
subsequente. Les resultats semblent indiquer que les sous -
echelles du MBTI constituent, en grande partie,
une mesure de la sociabilite et que l'extraversion, representee
dans les trois mesures susmentionnees, n'est
pas unitaire de prime abord, meme si des correlations existent
entre les facteurs a divers degres. There are
now several measures of extraversion available to the researcher
and practitioner. Of these the Myers - Briggs
Type Indicator (MBTI) and its corresponding extraversion -
introversion subscales are widely used in Canada
and the United States. Workshops and a very lucrative
enterprise around the MBTI has evolved in the Canadian
and American workplace. However, the construct validity of
this measure has not yet been clearly established.
There are a few studies which have explored the construct
validity of the MBTI. Sipps &Alexander(1987)
administered the MBTI and the Eysenck Personality Inventory
(EPI) scales and factor analyzed the combined
results. (For a comparison of the EPI and the later EPQ, please
see Campbell &Reynolds, 1984). In their
conjoint item analysis three factors of extraversion emerged: a
sociability component, an impulsivity/non -
planning component and a liveliness/risk - taking/jocularity
component. A synthesis of the reliability and validity
studies of the MBTI can be found in Carlyn (1977), Carskadon
(1979) or Carlson (1985). Briefly, various studies

investigated the split - half reliability coefficients for the MBTI
scales. They found that the reliability varied from
.80 to .90. However, the test - retest tend to be slightly lower
but acceptable. Furthermore, Thompson and
Borrello (1986) and Tzeng, Outcalt, Boyer, Ware, and Landis
(1984) have demonstrated item - level construct
validity of the MBTI. That is, the results of the assessments of
the MBTI tend to substantiate that it is a reliable
instrument that has four distinct psychometric dimensions
(validated by item - level factor analysis) that are
unidimensional and consistent with the theoretical constructs of
the MBTI. Interwoven in the issue of the
construct validity of the MBTI extraversion - introversion
subscales is the questionable unitary nature of
extraversion. As Zumbo and Hubley (1993) state, establishing
validity is difficult because it is generally
intertwined with the construction and verification of scientific
theories. That is, trying to verify a measure is often
difficult to separate from trying to verify a theory. Specifically,
it is important to confront the question of whether
the criterion scales concur in their conception of extraversion.
That is, to establish the construct validity of the
MBTI, the authors see a need to examine empirically the
correlational framework in which these subscales
exist. Necessarily, such a task must deal with the innate tension
which exists between the conceptual and the
concrete forms of a term. Campbell &Reynolds (1984, p. 318)
warn that the "actual labels attached to concepts
and scales is not at issue; the point is that semantic similarity
bears no clear relationship to - and cannot
substitute for - empirical similarity". The authors support this
statement and seek to clarify, empirically, what it is
that is measured in certain extraversion subscales.
The purpose of the present paper is directed at the above
question although in a somewhat novel fashion. The
study is limited to three measures of extraversion: the MBTI,

the Eysenck Personality Questionnaire (EPQ), and
the Howarth Personality Questionnaire (HPQ). To the authors'
knowledge, scale - level construct validity
analysis of extraversion has not been reported in the literature.
Appropriate subscales were combined into one
test which enabled us to carry out a factor analysis to
investigate the overall factor structure. In doing this, it was
possible to examine the construct validity of the MBTI
extraversion - introversion subscale and whether
extraversion as represented in these tests is fairly unified.
Before presenting the current study we will briefly review the
literature on the unitary nature of extraversion.
Confusion as to what it is that tests of extraversion are
measuring stems not only from the area of test
construction but from basic problems involved in the definition
of extraversion. Taking a brief look at the history
of extraversion we see that it has roots as far back as the late
seventeenth century when extraversion was used
in reference to the mind as a turning outward of one's thoughts
toward objects and was at the same time used
in chemistry meaning "to render visible or sensible the latent
constituents of a substance" (Browne, 1971, p. 7).
The word "extraversion" is taken from "extra" meaning
"outward" and "vert - ere" meaning "to turn". By contrast,
introversion is derived from a different prefix, "intro", meaning
"inward" (Browne, 1971, p. 7). Although this
definition may provide some insight into the literal meaning of
extraversion - introversion, it is not clear what is
directed outward or inward.
A discussion of the history of extraversion can be found in
Eysenck (1970). Early in the present century,
research began to emerge which focussed on the biological side
of extraversion - introversion (for a recent

discussion, see Stelmack, 1990). As well, later, ideas were
beginning to surface concerning the adaptive
qualities of extraversion. Speculation about extraversion
occurred on what is now considered a subjective level
until the 1930's when factor analysis of test scores enabled a
quantitative definition by researchers such as the
Guilfords (Guilford &Guilford, 1934). The Guilfords published
several articles on their attempts to identify various
factors of personality and of extraversion, using factor analysis.
Factor analysis of personality dimensions
probably attained its height in the 1960's when its use in
personality theory was widely popular.
FACTORS OF EXTRAVERSION
H.J. Eysenck and S.B.G. Eysenck have been important
contributors to the study of extraversion. Eysenck and
Eysenck (1963) discussed the possibility of sociability and
impulsiveness being aspects of extraversion that
were not independent. In a subsequent publication (Eysenck
&Eysenck, 1967) their factor analysis involved
scores on the lemon test (a physiological measure of
extraversion) as well as scores on extraversion items and
neuroticism items. Their results indicated two uniform factors
which were extraversion and neuroticism. Eysenck
&Eysenck concluded that for the purposes of the experiment,
extraversion was unitary in nature due to its
appearance as a unitary factor.
Guilford (1977) published a response to a long series of
disagreements between himself and the Eysencks
concerning the nature of extraversion. Guilford disagreed with
Eysenck's notion of extraversion being a second
- order factor resting on impulsiveness (factor R in Guilford's
scheme) and sociability (factor s). The dispute was
based on the fact that in factor analysis, a certain hierarchy of
factors exists in which first - order factors are
more molecular than second order factors. Having discovered
many different first - order factors, the task is to

assess how they may combine to form second - order factors.
In the case of J.P. Guilford and H.J. Eysenck, they both
believed that extraversion could exist as a second order
factor, but whereas Eysenck believed it to be connected to R
and S, Guilford believed it to be more closely tied
to R and T (Thoughtfulness). Guilford suggested to Eysenck
that his options were either to adopt extraversion
as resting on R and T since they were more highly correlated
than R and S, or to adopt extraversion as being
equivalent to the first order factor R. Guilford believed that R
and S were independent and that Eysenck's
extraversion factor became "rotated out of existence" (Guilford,
1977, p. 415). EXTRAVERSION AND SOCIAL
DESIRABILITY
Having looked at extraversion from a theoretical point of view,
let us examine some popular conceptions of
extraversion. Farley and Goh (1976) instructed students
whencompleting a measure of psychoticism,
neuroticism and extraversion to answer by relaying either a
"good impression" (reflecting social desirability), a
"worst impression" (reflecting psychiatric illness) or to answer
"normally". Results showed increased scores on
psychoticism, neuroticism and introversion for the "worst
impression" set. Here we see evidence of a popular
belief that introversion reflects mental instability. Furnham and
Henderson (1982) conducted a similar study in
which subjects who were instructed to give a bad impression on
various personality tests demonstrated
significantly lower extraversion scores while subjects instructed
to give a good impression scored significantly
higher on extraversion. These two pieces of research raise the
disconcerting possibility of test response bias on
measures of extraversion.

This leads us now to the method used in the present study which
is a conjoint factor analysis, a procedure
enabling various items or various scales to be combined and
analyzed together in the same matrix. For
example, Sells, Demaree &Will (1970) used 600 items for their
conjoint analysis while Browne and Howarth
(1977) combined four hundred items. CONJOINT FACTOR
ANALYSIS OF EXTRAVERSION Conjoint factor
analytic studies have been used for measures of more specific
scales than personality inventories. They have
proven themselves useful in the analysis of measures of
extraversion although very few actually exist to date.
Guilford and Guilford (1934) chose thirty - five items from
current introversion - extraversion tests which they
combined into one questionnaire. The results produced eighteen
factors of which four were considered central:
a) a tendency to fear the environment, b) an emotional
sensitivity to the environment, c) impulsiveness, and d)
interest in self. It was concluded that "the usual scale bearing
the name [extraversion] does not refer to any real
dimension of personality any more than the usual intelligence
test measures a single real variable of mental
ability" (Guilford &Guilford, 1934, p. 399).
Finally, the study which most resembles the present one was
reported in 1971 (Browne, 1971). Browne
combined 400 items taken from twenty different item sources.
An item factor analysis yielded twelve factors:
sociability, adjustment - emotionality, social shyness, trust vs.
suspicion, impulsivity, persistence, sex
&superego, Freudian introversion, dominance, unidentified,
cooperativeness - considerateness and finally,
inferiority. This broad spectrum of factors led Browne to
conclude that extraversion is too broad to be
considered a unitary concept.
THE PRESENT STUDY
This leads us to the question, now, of whether tests which claim

to be measuring the same thing, such as
extraversion, are actually measuring the same thing. It may be
that two tests claiming to measure the exact
same trait, for example, may not be measuring the same one,
and this may be due to either a disparity between
operational definitions or to inadequacies in the construction of
the tests. The present study is focussed on
these questions and has chosen to examine the concept of
extraversion as reflected by various measures.
METHOD
Subjects
The sample consisted of 210 introductory psychology students
who completed their questionnaires in full. The
sample was constructed of 109 females and 92 males(f.1). The
age range of the 171 subjects who indicated
their age was 19 to 52. The overall mean age was 22.4 (SD =
5.4). The average age for females was 23.2 (SD
= 6.2) and for males 21.4 (SD = 3.9). Our sample is
representative, in both age and gender composition, of
those previously used in investigating the psychometric
properties of the MBTI, EPQ, and HPQ.
Psychometric Instruments
A 98 - item questionnaire was developed as a measure of
extraversion and was constructed from three
measures of extraversion. These three scales were as follows:
1. The extraversion and introversion subscales of the Myers
Briggs Type Indicator (MBTI) Form G.
2. The extraversion subscale of the Eysenck Personality
Questionnaire (EPQ).
3. The five subscales of the Howarth Personality Questionnaire
(HPQ) which comprise extraversion. The HPQ
(Howarth, 1980) has been constructed from an item pool based
on twenty personality factors of which ten traits
are represented in the HPQ. For a more detailed history of the
HPQ and subsequent research see Howarth
&Zumbo (1988). The HPQ and EPQ were chosen because of

their popularity and because they were soundly
constructed on the basis of factor analysis.
The questionnaire was constructed by simply amalgamating the
subscales beginning with the MBTI followed by
the EPQ and the HPQ. There was no particular reason for this
ordering. The item order of each subscale was
unchanged from the order in which they originally appeared and
instructions for each section were retained from
the original. The MBTI extraversion subscale contains 21 items,
the EPQ also contains 21 items and the HPQ
contains 12 items for each subscale by which extraversion is
defined: Sociability, Anxiety, Dominance,
Impulsiveness and Inferiority (Howarth, 1985) for a total of 60
items from the HPQ. From these 102 items four
were eliminated due to replication (EPQ #70 "Can you get a
party going?", HPQ #21 "Do you like going out a
lot?", #71 "I am a good social mixer" and #101 "I generally
keep in the background on social occasions").
Scoring
A. Item Scores. Each item of the questionnaire was scored
according to the test manuals. A score of one was
given to a response endorsing extraversion and zero elsewhere.
This resulted in 98 dichotomously scored
items.
B. Scale Scores. To provide a score with which to validate the
extraversion scales, we computed a scale score
for each subject for each extraversion scale. This was
accomplished by assigning a unit weight to each item
which comprised the HPQ and EPQ scales. Therefore, the
subjects received one point for each extraversion
item they endorsed (i.e., answered in favour of extraversion).
This procedure is identical to that given in the test
manuals of the EPQ and HPQ.

It should be noted that for the EPQ and HPQ introversion is, in
theory, lack of extraversion. That is, they are
envisioned as two ends of a continuum. However, the MBTI
differentiates between extraversion and introversion
in that subjects receive scores for extraversion as well as for
introversion. For the most part, extraversion items
are identical to introversion items except that each item is
weighted differently depending on whether the
subject responds in favour of extraversion or introversion. The
weighting for each item is given in the test
manual. The scale scores for the MBTI were, therefore, formed
in accordance with these weightings.
When data analysis of extraversion scales was executed, those
items which had been removed from any given
scale due to replication elsewhere in the questionnaire were
reinserted into their appropriate scale calculations
so that all scales contained their complete set of items.
Reliabilities
Reports of the MBTI have generally yielded satisfactory
reliabilities. Carskadon (1977) reported test - retest
reliabilities on the extraversion/introversion scale as being .79
for males and .83 for females. The manual for the
EPQ reports a test - retest reliability for the extraversion
subscale as being high at .89 and being equally high
for university students. The reliabilities for the five subscales of
the HPQ range from .72 to .84 in value
(Howarth, 1985).
RESULTS
The means, standard deviations and reliabilities were calculated
for the items comprising each scale (see Table
1). These statistics were computed from the dichotomous data
resulting from the item scoring. It is important to
note that the two MBTI scales and the EPQ scale were each
scored out of 21 and each of the HPQ scales,
(Sociability (SOC), Anxiety (ANX), Dominance (DOM),
Impulsiveness (IMP) and Inferiority (INF), were scored

out of 12.
PART 1
Factorability of the Correlation Matrix
Table 2 contains the correlation matrix of the eight scales.
These eight scales were obtained via the scale
scoring description given above. The Kaiser - Meyer Olkin
(KMO) measure of sampling adequacy (Kaiser,
1970) was used yielding a value of .79 which is more than
adequate. Bartlett's Sphericity test was used in order
to determine whether the correlation matrix differed from the
identity matrix. The test yielded a value of 1322.27,
p <.001. Given the magnitude of the correlations these results
were not surprising and allowed us to
appropriately use the factor model. A maximum likelihood
factor analysis followed by oblique rotation (Direct
Quartimin, that is, Direct Oblimin with a delta value set at zero)
was used for all factor solutions. A delta value of
zero was chosen because when it is set at a large negative value
the factor solutions are increasingly
orthogonal, a value of zero allows for factors to be fairly highly
correlated, while values approaching one
produce factors that are highly correlated. We chose the
moderate value zero. It should be noted, however, that
even though there is a relation between the value of delta and
the size of the correlation among the factors, the
maximum value of the correlation among the factors for a given
value of delta depends on the data. This is
particularly why the Direct Quartimin was chosen, it allows the
factors to be highly correlated (a solution very
likely in a domain like extraversion) while not forcing a
positive manifold (i.e., a one factor second - order
solution; that is, something akin to a g - factor of extraversion)
on the factor space. Furthermore, maximum

likelihood analysis allows for a statistical test of the goodness -
of - fit between the data and the factor model
while oblique rotation which is recommended by Hakstian
(1971) and Hakstian &Abel (1974), does not constrain
the resulting factors to being orthogonal and allows us to
investigate any possible second - order factors.
Various factor solutions were tested in which one - through to
five - factor solutions were examined. Criteria
used in determining the suitability of the model were twofold.
The Chi - squared goodness - of - fit test and the
residuals between the observed and the reproduced correlations
each served to test how well the model fit the
actual data. All of the residuals should be less than 0.10 in
order for the solution to be appropriate for the data
(McDonald, 1985). Neither the one - factor solution nor the two
- factor solution met any of the criteria. The three
- factor solution did not meet the criterion of the goodness - of -
fit although the residuals were deemed
adequate. A glance at the pattern matrix for the 3 - factor
solution relayed reasonably high loadings of variables
on factors except for the variable Dominance which had a low
loading on all factors. It is the authors' opinion
that the 3 - factor model is an under extraction. Without going
into much detail, Factor 1 included MBTI -
introversion, MBTI - extraversion, and Dominance. Factor 2
included Inferiority and Anxiety while Factor 3
included Sociability, EPQ, and Impulsiveness. The four - factor
solution was decided upon as being most
suitable. The goodness - of - fit test, x(Fe 2) (2) = 0.7463, p
<.65, suggested an acceptable fit and all the
residuals were acceptable at less than .10. Furthermore, all the
communalities were less than 1.0 (see Table 2)
which is an indicator of an adequate solution.
The factors obtained were as follows: (see Table 3)
Factor 1. This first factor was indeed responsible for a large
portion of the total variance (46.0%) and was made

up of the MBTI scales. It makes theoretical sense for both
MBTI scales to be in the same factor since, as we
have seen, they are almost perfectly negatively correlated in the
original correlation matrix.
Factor 2. Howarth's Inferiority and Anxiety scales created this
factor. Howarth's own definition describes
extraversion as "lacking either anxiety or feelings of
inferiority" (Howarth, 1985) and it would appear that a
lacking in both of these is somehow related. Factor 2 accounted
for 14.5% of the total variance.
Factor 3. Sociability, EPQ extraversion and Impulsiveness
formed the third factor which accounted for 6.6% of
the total variance. Factor 4. The final factor responsible for
only 4.2% of the total variance included only one
variable, Dominance.
In order to obtain a general idea of how the factors related to
each other, let us now look at the factor correlation
matrix. Table 4 indicates negative correlations between Factor 2
and all factors, which was expected
considering Howarth's belief that his scales measure something
which other extraversion scales lack. One
cannot help noticing the large value (.787) which has resulted
between Factor 1 (MBTI) and Factor 3
(SOC/EPQ/IMP). However, weaker correlations seemed to exist
between Factors 4 and 1 and between Factors
4 and 3.
Before proceeding, it should be noted that the results of this
section were obtained from scale scores. These
are the scores that an individual would obtain when scoring the
measures according to their respective test
manuals.
This is important to note because there exists an alternative
manner of analysing this data. That is, under
statistically optimal conditions (i.e., given that we have 98
items we need well over 1500 subjects) the
alternative approach would be to simultaneously fit all of the

scales via confirmatory factor analysis, then the
resulting inter - factor correlation matrix would serve as the
input for the exploratory factor analysis. However,
the alternative analysis was not conducted for the following
reasons:
1. In administering these scales in an applied setting factor
score coefficients are not used. Rather, unit weights
are utilized. We were interested in examining the construct
validity of these scales within the context that they
are being used. See Zumbo and Hubley (1993) for a discussion
of the importance of considering the intended
use of tests when investigating their validity.
2. The sample size in the present study is too small to allow for
item - level analysis. This study was planned for
scale - level analysis and therefore we planned for a factor
analysis of eight variables. Furthermore, with the
simultaneous fit method, our sample size would result in poor
estimation of factor coefficients and factor scores.
This would have a cascading disastrous effect on the joint
distribution of the variables and the correlation
coefficients among the factors (Tucker, 1971). Therefore, we
have followed the recommendations of Bartlett
(1937) and Overall and Klett (1983) to use unit weights rather
than factor scoring coefficients(f.2)
3. It can be seen in the literature review that the item - level
factorial validity of the EPQ, HPQ, and MBTI are
well supported. In fact, these measures were chosen because of
their psychometric properties. Another item -
level analysis is not necessarily needed to substantiate the
factorial validity of these measures.
In summary, the simultaneous method was not utilized because
of the issue of establishing validity within the
context of test use, the inappropriate sample size, and that the

item - level factorial validity has been established
in previous papers. Therefore, the four - factor solution based
on the factor analysis of the scale scores can now
be submitted to higher - order factor analysis.
PART 2
Having extracted four factors from the data, the next step was to
examine the structure of possible second -
order factors. The four - factor correlation matrix achieved an
adequate KMO - value of .61 and a significant
value of 263.82, p <.05 on the Bartlett test of sphericity. These
results allowed us to proceed with a higher -
order factor analysis. The four - factor solution which was
found to be appropriate in Part 1 of the results was
subjected to a factor analysis in order to explore the possibility
of extracting a single interpretable higher - order
factor. Factor analysis proceeded as before, although rotation
was not necessary for a one - factor solution.
Using the same criteria as in Part 1, it was found that the Chi -
squared goodness - of - fit was not statistically
significant with X'Symbol not transcribed'2 (2)= 5.74, p >.05
with all residuals less than .10. The one - factor
solution was, therefore, found to be appropriate.
This second - order factor was labelled a general extraversion -
introversion factor. The highest loading on this
one general factor was Factor 1 with a value of .979 followed
by factors 3, 4, and 2, with loadings of .804, .441,
and -.273, respectively. This general extraversion - introversion
factor is a coalescence of the MBTI
extraversion/lack of introversion as well as Sociability,
Impulsiveness and Eysenck's extraversion. Again, the
small negative loading on Factor 2 lends credence to Howarth's
statement that INF and ANX measure
something that Eysenck's extraversion scale lacks.
Out of curiosity the previously rejected three - factor solution
was subjected to higher - order analysis. The
criteria for the Kaiser - Meyer - Olkin and Bartlett tests were

met. The resulting three - factor matrix was similar
to the four - factor matrix with Factor 1 (which now includes
Dominance) loading almost perfectly followed by
Factor 3 and Factor 2. The ordering of factors was similar to the
above - mentioned solution and the factors
themselves, described earlier, were almost identical in the
content of their variables. Again, Factor 2 made up of
Inferiority and Anxiety received a negative loading. This
solution lends to the robustness of the unifactorial
extraversion model. DISCUSSION
In this paper we were interested in the construct validity of the
MBTI subscales as a vehicle for exploring the
potential unitary nature of extraversion. A conjoint factor
analysis of certain extraversion scales seemed to be
an interesting gateway to these issues.
Though some studies have carried out item analyses of
extraversion (Sipps &Alexander, 1987; Browne, 1971),
the present study is unique to the authors' knowledge in its
factor analytic calculations of extraversion at the
scale level. What this means is that each measure of
extraversion was represented in its entirety somewhere
within the overall factor structure of the combined scales, as
opposed to individual items being represented
somewhere within the overall structure of the combined items.
In other words, conjoint analysis at the scale
level was more qualified to address the question of how each
scale may differ from the other scales in terms of
factor membership and the extent to which the factors are
correlated. Conversely, analysis at the item level is
limited to the question, "How well are each of the factors
reproduced by the items?". So it is clear that conjoint
scale analysis is better equipped to deal with our present
concern which is that of discrepancies between scales

in what is being measured, i.e. factorial validity. See Howarth
and Zumbo (1989; 1988) for further discussion.
It seems that the MBTI scales formed a factor of their own and
that the EPQ combined with Sociability and
Impulsiveness to create the third factor. This comes as no
surprise since the EPQ seems intuitively to define
extraversion in terms of these two variables. The fact that the
MBTI and the EPQ occurred as separate factors
suggests that these two measures do not hold identical views on
extraversion. However, as evidenced by the
factor correlation matrix, Factor 1 (MBTI) and Factor 3
(SOC/EPQ/IMP) had a fairly high correlation. From this
one could conclude that although each measure is unique, each
contains a high degree of sociability and
impulsiveness in its definition. The interfactor correlation
among factors one and three was in support of
Campbell &Heller (1987). The substantial correlation among
factors one and three, and the factor loadings of
the SOC, EPQ, and IMP suggested that the MBTI was primarily
a sociability measure with a minimal
association with impulsivity. This was also in support of
Campbell &Heller. The second factor (INF/ANX) yielded
only low negative correlations with the other factors and it
seems therefore that although it took up 14.5% of the
variance, it bears little relation to the other factors. Finally,
Factor 4, Dominance, which was responsible for only
4.2% of the variance, obtained only a low loading on the MBTI
factor in the three - factor solution and was
relegated to its own factor in the final four - factor solution.
From this first phase of the results it is clear that
extraversion cannot be described as entirely unitary. It would
appear to be defined in these measures by
sociability, impulsiveness, a lack of inferiority and anxiety and
the presence of dominance. Part 2 of the results
made apparent the possibility of a single higher - order factor
capable of bridging the gap between the original

four factors. The results did, in fact, indicate that these can be
brought together into one interpretable, though
less convincing higher - order factor. We may ask at this point
whether the results were out of the ordinary and
this question is difficult to answer since the study was unique in
its approach. To the authors' knowledge, there
have been no scale - level analyses of extraversion, though
many item - level analyses exist. Such item
analyses have revealed that extraversion is composed of many
factors. For example, Sipps &Alexander (1987)
reported finding three factors of extraversion which were
sociability, impulsivity/non - planning and risk - taking /
jocularity while Browne (1971) reported twelve. Certainly, it
would be safe to say that sociability has never been
far away from any set of results. Apart from being present in
most factor analyses of extraversion, it has been
shown that the extraversion scale of the MBTI is largely a
measure of sociability (Sipps &DiCaudo, 1988). Given
that the present study declared MBTI extraversion as the most
salient factor, one must wonder about the
strength of the relationship between extraversion and
sociability.
We would like to emphasize that factorial construct validity
should be conducted not only at the item level
whereby one establishes the reproducibility of the postulated
scales but also at the scale level whereby several
similarly labelled scales can be explored within a common
factor space. We concur with Campbell &Reynolds
(1984) in the following: In summary, we suggest that the most
useful question to ask is not "Which is the real
factor?" but rather "How do the available measures operate?"
Researchers should continue to examine the
utility of primary and higher - order traits... As a final caveat,
researchers should not be misled by semantic
similarity, either between models or across successive
instruments for a given model; a dimension of

personality is best understood by its operating characteristics,
not by its label (p. 319). One may wonder
whether research in the area of extraversion is still necessary.
Although the literature is less prevalent than in
the past, the concept of extraversion has infiltrated personality
measures and common language and remains
an integral part of many personality inventories. As long as the
term is still in use, research in this domain
cannot be futile, nor can it be futile to examine and question the
construct validity of representative scales. The
authors believe there is a need for further scale level analyses
as a process for expanding our knowledge of
extraversion and its psychometric homologues.
Thanks to Prof. Bob Stelmack of the University of Ottawa and
three anonymous reviewers for comments on an
earlier draft. Parts of this paper were reported by the second
author for her B.A. Honours thesis at Carleton
University under the supervision of BDZ.
Send correspondence to: Bruno D. Zumbo, Measurement and
Evaluation, Faculty of Education, University of
Ottawa, Ottawa, Ontario CANADA K1N 6N5. E - mail: ZUMBO
AT ACADJM1. UOTTAWA.CA.
References
Bartlett, M.S. (1937). The statistical conception of mental
factors. British Journal of Psychology, 22, 97 - 104.
Browne, J.A. (1971). Extraversion: In search of a personality
dimension. Doctoral thesis, University of Alberta,
1971.
Browne, J.A. &Howarth, E. (1977). A comprehensive factor
analysis of personality questionnaire items.
Multivariate Behavioural Research, 12, 399 - 427.
Campbell, J.B. &Heller, J.F. (1987). Correlations of
extraversion, impulsivity and sociability with sensation

seeking and MBTI - introversion. Personality and Individual
Differences, 8, 133 - 136. Campbell, J.B.
&Reynolds, J.H. (1984). A comparison of the Guilford and
Eysenck Factors of Personality. Journal of Research
in Personality, 18, 305 - 320.
Carlson, J.G. (1985). Recent assessments of the Myers - Briggs
Type Indicator. Journal of Personality
Assessment, 49, 356 - 365. Carlyn, M. (1977). An assessment of
the Myers - Briggs Type Indicator. Journal of
Personality Assessment, 41, 461 - 473.
Carskadon, T.G. (1977). Test - retest reliabilities of continuous
scores on the Myers - Briggs Type Indicator.
Psychological Reports, 41, 1011 - 1012.
Eysenck, H.J. (1970). Historical Introduction. In H.J. Eysenck
(Ed.), Readings in Extraversion - Introversion, Vol.
1, Theoretical and Methodological Issues. John Wiley &Sons:
New York.
Eysenck, H.J. &Eysenck, S.B.G. (1967). On the unitary nature
of extraversion. Acta Psychologica, 26, 383 -
390.
Eysenck, S.B.G. &Eysenck, H.J. (1963). On the dual nature of
extraversion. British Journal of Social and
Clinical Psychology, 2, 46 - 55.
Farley, F. &Goh, D. (1976). PENmanship: Faking the P - E - N.
British Journal of Social and Clinical
Psychology, 15, 139 - 148.
Furnham, A. &Henderson, M. (1982). The good, the bad and the
mad: Response bias in self - report measures.
Personality and Individual Differences, 3(3), 311 - 320.
Guilford, J.P. (1977). Will the real factor of extraversion -
introversion please stand up? A reply to Eysenck.
Psychological Bulletin, 84, 412 - 416.
Guilford, J.P. &Guilford, R.B. (1934). An analysis of the
factors in a typical test of introversion - extraversion.
Journal of Abnormal and Social Psychology, 28, 377 - 399.
Hakstian, A.R. (1971). A comparative evaluation of several

prominent methods of oblique factor transformation.
Psychometrika, 36, 175 - 193. Hakstian, A.R. &Abel, R.A.
(1974). A further comparison of oblique factor
transformation methods. Psychometrika, 39, 429 - 444.
Howarth, E. (1980). Major factors of personality. Journal
of Psychology, 104, 171 - 183.
Howarth, E. (1985). Scale descriptions, norm tables and
bibliography for the HPQ, APF and HMACL.
Laboratory for Research in Personality, University of Alberta.
Howarth, E. &Zumbo, B.D. (1989). An empirical investigation
of Eysenck's typology. Journal of Research in
Personality, 23, 343 - 353. Howarth, E. &Zumbo, B.D. (1988).
Exploring new personality dimensions. Laboratory
for Research in Personality, University of Alberta. Kaiser, H.F.
(1970). A second generation Little Jiffy.
Psychometrika, 35, 401 - 415.
McDonald, R.P. (1985). Factor analysis and related methods.
Hillsdale, N.J.: Lawrence Erlbaum.
Overall, J.E., &Klett, C.J. (1983). Applied multivariate
analysis. Malabar, Florida: Robert E. Kreiger Publishing
Co..
Sells, S.B., Demaree, R.G., &Will, D.P. Jr. (1970). Dimensions
of personality: A conjoint factor structure of
Guilford and Cattell trait markers. Multivariate Behavioural
Research, 5, 391 - 422. Sipps, G.J. and Alexander,
R.A., (1987). The multifactorial nature of extraversion -
introversion in the Myers - Briggs Type Indicator and
Eysenck Personality Inventory. Educational and Psychological
Measurement, 47, 543 - 552.
Sipps, G.J. and DiCaudo, J., (1988). Convergent and
discriminant validity of the Myers - Briggs Type Indicator
as a measure of sociability and impulsivity. Educational and
Psychological Measurement, 48, 445 - 451.

Stelmack, R.M. (1990). Biological Bases of Extraversion:
Psychophysiological Evidence. Journal of Personality,
58, 293 - 311. Thompson, B., &Borrello, G.M. (1986).
Construct validity of the Myers - Briggs Type Indicator.
Educational and Psychological Measurement, 46, 745 - 752.
Tucker, L.R. (1971). Relations of factor score estimates to their
use. Psychometrika, 36, 427 - 436.
Tzeng, O.C.S., Outcalt, D., Boyer, S.L., Ware, R., &Landis, D.
(1984). Item validity of the Myers - Briggs Type
Indicator. Journal of Personality Assessment, 48, 255 - 256.
Zumbo, B.D., &Hubley, A.M. (1993). The role of validity
theory in the resolution of observational indeterminacy.
Carleton Papers in Applied Language Studies, 10, 18 - 37.
Submitted June 28, 1991 Revision July 6, 1992 Accepted
October 14, 1992
Footnote
(f.1) To comply with ethical standards at our institution,
subjects were given the option of submitting age and
gender on the questionnaire. Nine subjects did not indicate their
gender.
(f.2) Although a simultaneous factor analysis of all items was
not conducted, a scale - by - scale unifactorial
solution was obtained. Factor scores were computed for each
scale separately and these factor scores were
used to create an 8 x 8 correlation matrix. The results of this
analysis were a replication of the scale - level
results in Table 4. We would like to thank a reviewer for this
suggestion and prompting our consideration or this
issue.
TABLE 1
Mean Scores, Standard Deviations and Reliabilities for Each
Scale
------------------------------------------------------------ Scale Mean
Standard Reliability Deviation -------------------------------
----------------------------- MBTI-Ext 13.90 6.30 .83 MBTI-Int
14.14 6.34 .83 EPQ 14.51 4.79 .86 SOC 8.36 3.34 .85

J Youth Adolescence (2007) 361048–1057DOI 10.1007s10964-00.docx

J Youth Adolescence (2007) 361048–1057DOI 10.1007s10964-00.docx

Recommended

Recommended

More Related Content

Similar to J Youth Adolescence (2007) 361048–1057DOI 10.1007s10964-00.docx

Similar to J Youth Adolescence (2007) 361048–1057DOI 10.1007s10964-00.docx (20)

More from priestmanmable

More from priestmanmable (20)

Recently uploaded

Recently uploaded (20)

J Youth Adolescence (2007) 361048–1057DOI 10.1007s10964-00.docx