Development and Validation of the Subtypes of Antisocial Behavior Questionnaire

Abstract

There is converging evidence that physical aggression, rule-breaking, and social aggression constitute meaningfully distinct, if somewhat overlapping, components of the broader construct of antisocial behavior. Indeed, these subtypes appear to have different developmental trajectories, demographic correlates, and personological underpinnings. They also demonstrate important etiological distinctions. One potential limitation to accumulating additional scientific insights into the correlates and origins of these three types of antisocial behavior is the lack of an efficient self-report assessment in the public domain. We developed the 32-item Subtypes of Antisocial Behavior Questionnaire (STAB) to fill this gap. Our goal was to develop a brief measure that could reliably and validly assess each of the three major subtypes of antisocial behavior and that would be freely available for other researchers. The present series of studies provides initial evidence of the factorial validity, internal consistency, and criterion-related validity of the STAB scales. In short, it appears that the STAB is a brief and useful measure that can be used to differentiate and assess physically aggressive, rule-breaking, and socially aggressive forms of antisocial behavior.

Full text

AGGRESSIVE BEHAVIOR
Volume 35, pages 376–398 (2009)
Development and Validation of the Subtypes
of Antisocial Behavior Questionnaire
S. Alexandra Burt

and M. Brent Donnellan
Department of Psychology, Michigan State University, East Lansing, Michigan
:::::::::::::::::::::::::::::::::::::::::
There is converging evidence that physical aggression, rule-breaking, and social aggression constitute meaningfully distinct, if
somewhat overlapping, components of the broader construct of antisocial behavior. Indeed, these subtypes appear to have different
developmental trajectories, demographic correlates, and personological underpinnings. They also demonstrate important
etiological distinctions. One potential limitation to accumulating additional scientific insights into the correlates and origins of
these three types of antisocial behavior is the lack of an efficient self-report assessment in the public domain. We developed the
32-item Subtypes of Antisocial Behavior Questionnaire (STAB) to fill this gap. Our goal was to develop a brief measure that could
reliably and validly assess each of the three major subtypes of antisocial behavior and that would be freely available for other
researchers. The present series of studies provides initial evidence of the factorial validity, internal consistency, and criterion-
related validity of the STAB scales. In short, it appears that the STAB is a brief and useful measure that can be used to
differentiate and assess physically aggressive, rule-breaking, and socially aggressive forms of antisocial behavior. Aggr. Behav.
35:376–398, 2009.
r
2009 Wiley-Liss, Inc.
:::::::::::::::::::::::::::::::::::::::::
Keywords: physical aggression; social aggression; rule-breaking; antisocial behavior instrument; Subtypes of Antisocial Behavior
Questionnaire
INTRODUCTION
Antisocial behaviors are actions that harm others,
violate societal norms, and/or infringe on the
personal or property rights of others. Typical
examples include illegal actions such as vandalism,
theft, and assault as well as interpersonally harmful
behaviors such as the use of racial slurs and the
spreading of damaging rumors. Even so, the specific
manifestation of antisocial behavior varies markedly
from individual to individual [Lahey and Waldman,
2003; Loeber and Stouthamer-Loeber, 1998; Offord
and Bennett, 1994; White et al., 2001]. Consistent
with this observation, factor analytic studies have
indicated that there are at least two moderately
correlated antisocial factors: an ‘‘overt’’ or physi-
cally aggressive/oppositional factor and a ‘‘covert’’
or nonaggressive/rule-breaking factor [Frick et al.,
1993; Loeber and Schmaling, 1985]. This distinction
is evident in both empirically derived behavioral
rating scales, such as Achenbach’s Child Behavior
Checklist (CBCL), and in factor analyses of conduct
disorder and oppositional defiant disorder symp-
toms [Tackett et al., 2003, 2005]. Physical aggression
(e.g., physically attacking others and bullying) and
nonaggressive rule-breaking (e.g., lying, stealing
without confrontation, and vandalism) also appear
to have different developmental trajectories. Physi-
cal aggression is most prevalent during the toddler
years [Tremblay, 2003], after which mean levels of
these behaviors steadily decrease [Stranger et al.,
1997; Tremblay, 2003]. Rule-breaking, by contrast,
is relatively infrequent during childhood, and
increases dramatically over the course of adoles-
cence, only to fall off again during the transition
into adulthood [Stranger et al., 1997]. Moreover,
physical aggression exhibits high levels of
rank-order stability across development, such
that those young children with the highest levels of
these behaviors continue to be particularly
aggressive as adults [Tremblay, 2003], whereas
rule-breaking does not exhibit this high level of
stability.
Published online 17 July 2009 in Wiley InterScience (www.
interscience.wiley.com). DOI: 10.1002/ab.20314
Received 22 January 2009; Revised 11 June 2009; Accepted 15 June
2009

Correspondence to: Alex Burt, Department of Psychology,
Michigan State University, 107D Psychology Building, East Lansing,
MI 48824. E-mail: burts@msu.edu
r
2009 Wiley-Liss, Inc.

Research has also supported distinctions
between the correlates of physically aggressive and
rule-breaking antisocial behavior, such that deficits
in affective regulation are particularly characteristic
of physical aggression [Burt and Donnellan, 2008;
Burt and Larson, 2007; Cohen and Strayer, 1996;
DeMarte, 2008; Pardini et al., 2003], whereas
impulsivity appears to be more strongly associated
with rule-breaking [Burt and Donnellan, 2008;
DeMarte, 2008]. Physical aggression and rule-break-
ing also demonstrate etiological distinctions. Speci-
fically, physical aggression appears to be more
heritable than rule-breaking (i.e., genetic influences
account for 65 and 48% of the variance, respec-
tively), whereas rule-breaking is influenced more
by the shared environment than is aggression
(i.e., shared environmental influences account for 5
and 18% of the variance, respectively) [Burt, 2009;
Tackett et al., 2005]. Recent work has also suggested
that associations with particular candidate genes
(namely, 5HT
2A
His452Tyr and DAT1) vary across
physical aggression and rule-breaking [Burt and
Mikolajewski, 2008], such that these particular genes
are uniquely associated with rule-breaking. In short,
there is converging evidence that physical aggression
and nonaggressive, rule-breaking constitute two
separable though correlated subtypes of antisocial
behavior.
Social aggression (also described as indirect or
relational aggression) constitutes yet another form of
antisocial behavior, one that uses social relationships
as a means of harming others. It encompasses
behaviors such as gossiping, ostracism, and ‘‘steal-
ing’’ friends, behaviors that can be expressed either
overtly (e.g., threatening to end a friendship) or
covertly (e.g., spreading rumors). Researchers have
suggested that social aggression should be distin-
guished from other types of antisocial or aggressive
behaviors [Bjorkqvist et al., 1992; Crick et al., 1997;
Crick and Grotpeter, 1995; Vaillancourt et al., 2003],
in part because of purported gender differences in
base rates. It is proposed that girls are far less likely
to engage in physical aggression but are equally or
even more likely to engage in social aggression. The
exclusion of social aggression from assessment
batteries may thus lead researchers to overlook girls
with antisocial proclivities [Crick and Zahn-Waxler,
2003]. Nonetheless, social aggression appears to be
antisocial to the extent that victims of social
aggression report psychological outcomes (e.g., lone-
liness and depressive symptoms) quite similar to
those experienced by victims of physically aggressive
behaviors [Crick and Bigbee, 1998; Crick et al., 2002].
In short, although the socially aggressive behaviors in
question are not necessarily illegal, they do constitute
a form of antisocial behavior.
Recent evidence also suggests that social aggres-
sion may be somewhat distinct from both rule-
breaking and physical aggression. For example,
DeMarte [2008] found that physical aggression,
rule-breaking, and social aggression comprised three
oblique dimensions of antisocial behavior, findings
that replicated in both normative and criminally
active samples. Moreover, social aggression predicts
psychological maladjustment over and above the
effects of physical aggression, indicating that the
harmful effects of social aggression are not redun-
dant with other forms of antisocial behavior [Crick
and Bigbee, 1998]. Social aggression also differs
from the other two subtypes demographically.
Social aggression may be more pronounced in
females than in males [though results have been
inconsistent, as some studies have reported equiva-
lent levels of social aggression across gender; Forrest
et al., 2005], a demographic pattern that does not
extend to either physical aggression or rule-break-
ing, both of which are more common in males as
compared with females [Moffitt, 2003]. Moreover,
research has demonstrated that the developmental
trajectory of social aggression is distinct from those
of both physical aggression and rule-breaking (as
described above). Social aggression is consistently
observed as early as the preschool years [e.g., Crick
et al., 1997; McNeilly-Choque et al., 1996], but is
most common during adolescence [e.g., Cairns et al.,
1989; Osterman et al., 1998]. It then remains
relatively frequent until early adulthood when it
decreases [e.g., Xie et al., 2005].
Finally, physical and social aggression evidence
different associations with comorbid psychopathol-
ogy, peer relations, and neuroendocrine functioning.
Whereas physical aggression is highly comorbid
with externalizing disorders, social aggression more
often coexists with internalizing disorders [Crick,
1997]. Similarly, physical aggression has been
associated with increased peer rejection, whereas
social aggression has been associated with higher
levels of peer acceptance (at least among males)
[Crick et al., 1997]. Finally, social aggression
appears to have unique biological correlates.
Physical aggression is linked to heightened cortisol
in the morning and steep declines over the course of
the day. By contrast, social aggression is associated
with low levels of cortisol in the morning and
blunted diurnal changes during the day [Murray-
Close et al., 2008]. Similarly, DeMarte [2008] found
that social aggression is associated with delayed
pubertal onset and more feminine finger length
377STAB Questionnaire
Aggr. Behav.

ratios (which are thought to index low levels of
prenatal exposure to testosterone). In sum, physical
aggression, rule-breaking, and social aggression
appear to be major subtypes of antisocial behavior,
each with different developmental trajectories,
demographic patterns, correlates, and etiological
underpinnings.
Need for a New Instrument
The emergence of the three constructs of physical
aggression, rule-breaking, and social aggression
creates a need for explicit measures of these facets
of antisocial behavior. The existing literature has
relied on a combination of measures to provide
coverage of these three constructs or used measures
that were extremely long. For example, Krueger et al.
[2007] recently developed a comprehensive set
of true/false items that covers the full spectrum of
externalizing problems. However, the length of
this instrument (415 items for the entire battery,
114 of which explicitly tap the constructs of physical
aggression, rule-breaking, and social aggression)
could create practical problems for many applica-
tions (i.e., longitudinal panel studies) where space
and subject time constraints are a limiting factor
(e.g., their true-false response option necessitates
more items to achieve adequate variance and
reliability). In addition, most existing measures do
not assess all three of these constructs. This includes
the 29-item Aggression Questionnaire [Buss and
Perry, 1992] which provides good coverage of
physical aggression and what they term ‘‘verbal’’
aggression but lacks coverage of rule-breaking, the
Achenbach Adult Self-Report [Achenbach and
Rescorla, 2003], which provides coverage of physical
aggression and rule-breaking but lacks a scale
explicitly tapping social aggression, or the Relational
Aggression questionnaire [Loudin et al., 2003], which
does not cover physical aggression or rule-breaking.
Finally, many instruments are proprietary (e.g., the
Achenbach, the Personality Assessment Inventory)
and thus researchers have to pay to use them. This
creates demands on limited research budgets and may
force researchers to limit sample sizes, a pragmatic
decision that undermines statistical power. In short,
there is a gap in the literature for researchers who
want a public domain and comprehensive yet
relatively short measure that has separate scales
for physical aggression, social aggression, and
rule-breaking.
These considerations motivated us to develop
the Subtypes of Antisocial Behavior Questionnaire
(STAB). Our goal was to obtain a measure of
around 30 items in length that would reliably
and validly assess each of the three major
subtypes of antisocial behavior. We also wanted a
measure that was suitable for research on college
students [especially given their high rates of
adolescent-limited antisocial behavior; Burt and
Mikolajewski, 2008; Moffitt, 1993] and community
samples, as well as ‘‘clinical’’ samples (i.e., indivi-
duals involved in the criminal justice system).
Finally, we planned to make the measure to be
freely available for other researchers in the hopes
that this would lead to a faster accumulation of
findings.
Overview of the Present Studies
We developed and evaluated the STAB across five
independent and diverse samples. The final measure
has 32 items. Study 1 details the development of the
STAB using a relatively large sample of college
students (N5400; 50% women). Once the scales
were constructed, we then examined how well the
STAB scales related to personality as assessed via
the Multidimensional Personality Questionnaire-
Brief Form [MPQ-BF; Patrick et al., 2002]. We
used Study 2 to further examine the psychometric
properties of the STAB in an independent sample of
college students (n5500; 50% women). Moreover,
we sought to establish convergent and discriminant
validity with relevant subscales in the Externalizing
Spectrum Model (ESM) [Krueger et al., 2007], and
to extend the personological findings from Study 1
to the Big Five domains. Study 3 was designed to
provide initial evidence of criterion-related validity
for the STAB by evaluating whether the scales can
successfully discriminate antisocial behaviors in
normative populations from those in criminally
active populations. We thus collected data on a
sample of 218 adjudicated adults (15% women)
currently under court supervision (i.e., probation or
parole) and in one of three treatment groups (i.e.,
substance use problems, domestic violence, or anger
management). Study 4 was designed to provide
additional evidence of criterion-related validity,
both via mean-level comparisons with normative
populations and via associations with the ESM. To
do so, we collected data on a second, independent
sample of adjudicated adults (n5155, 17% women)
currently under court supervision and in treatment.
Study 5 assessed a community sample of adults
(n5402, 60% women) to allow for mean-level
comparisons with the other sample types assessed
herein (i.e., adjudicated, college student).
378 Burt and Donnellan
Aggr. Behav.

STUDY 1: DEVELOPMENT OF THE 32-ITEM STAB
Sample
The sample consisted of 400 undergraduate
students (50% women; average age 519) enrolled
in psychology courses at a large public university
in the Midwest. They participated in exchange
for course credits or extra credit during Spring
Semester 2007. Data were collected over the Internet
using an anonymous web-based interface. The
ethnic breakdown was Caucasian (87%), African–
American (5%), Asian or Pacific Rim (5%),
Hispanic/Latino (1%), and other (2%) ethnicities.
Research protocol was approved by the Michigan
State University IRB. All participants provided
informed consent.
Item Content
Items were written by the authors, both of whom
study antisocial behavior. Existing instruments that
measure physical aggression, rule-breaking, and
social aggression were first consulted to form an
initial item pool. These included the following
measures: the Aggression Questionnaire [Buss and
Perry, 1992]; the Delinquent Behavior Index [Burt
and Donnellan, 2008; Farrington and West, 1971]; a
measure of Workplace Deviance [Bennett and
Robinson, 2000]; the Displaced Aggression Ques-
tionnaire [Denson et al., 2006]; a measure of peer
delinquency [the Friends questionnaire; Walden
et al., 2004]; a measure of antisocial behaviors
among college students modified from the National
Youth Survey [Elliott and Ageton, 1980; Paulhus
and Williams, 2002]; the Verbal Aggressiveness
Scale [Infante and Wigley, 1986]; the Richardson
Conflict Response Questionnaire [Richardson and
Green, 2003]; the Relational Aggression Question-
naire [Loudin et al., 2003]; and the Self-Report
Aggressive Driving Questionnaire [Hennessy and
Wiesenthal, 2001]. We also consulted diagnostic
criteria for Conduct Disorder and Antisocial Per-
sonality Disorder as listed in the DSM-IV-TR. Items
were selected for inclusion in the initial item pool
based on estimated severity of the behavior in
question, relevance to one of the three constructs,
and readability. We then developed additional items
to assess aspects of physical aggression, rule-break-
ing, and social aggression that were otherwise not
adequately assessed. A similar strategy was followed
by Denson et al. [2006] to create the Displaced
Aggression Questionnaire. We derived 120 items for
possible inclusion (i.e., 38 physical aggression items,
47 rule-breaking items, and 35 social aggression
items), all of which were administered to Study 1
participants.
We used the following instructions: ‘‘The follow-
ing items describe a number of different behaviors.
Please read each item and report how often you
have done this during the past year using the
following scale.’’ Items were administered using a
five-point scale (1 5‘‘never,’’ 2 5‘‘hardly ever,’’
35‘‘sometimes,’’ 4 5‘‘frequently,’’ and 5 5‘‘nearly
all the time’’). The items were written so that
the time frame could be changed (e.g., past year
to lifetime) and so that observers could use the
same core items to provide informant reports
of antisocial behavior. In this study, we made
use of the stem ‘‘in the past year’’ and relied on
self-reports.
Personality
Participants completed the 155-item MPQ-BF
[Patrick et al., 2002]. The MPQ-BF is composed of
ten primary scales that coalesce into three higher-
order factors: Positive Emotionality (PEM; the
dispositional tendency to experience positive affect/
emotions), Negative Emotionality (NEM; the dis-
positional tendency to experience negative affect/
emotions), and Constraint (CON; reverse-scored
impulsivity and behavioral restraint). Note that
NEM includes an Aggression scale as one of its
three subscales. Given this, although we focused our
primary attention on the three higher-order factors,
we also examined the MPQ-BF Aggression scale
separately.
STUDY 1 RESULTS
Initial Item Winnowing
We first conducted a series of principal axis
exploratory factor analyses (EFA) on the pool of
items that were developed for each expected dimen-
sion. We forced a single-factor solution for each
dimension and selected items that loaded highly on
that factor (i.e., 4.40). Following this procedure, 20
physical aggression items, 18 rule-breaking items,
and 21 social aggression items were selected for
further analyses. We submitted these 59 items to
another principal axis EFA where we forced a three-
factor solution using a promax (i.e., oblique)
rotation. We selected those items that loaded cleanly
on only one of the three factors (i.e., loadings
greater than .40 on one factor, and less than .30 on
the other factors). This process yielded 10 physical
379STAB Questionnaire
Aggr. Behav.

aggression items, 11 rule-breaking items, and 11
social aggression items.
1
EFA of 32-Item STAB
We subjected the final set of 32 items to an
additional EFA with normalized promax rotation.
We used FACTOR 7.0 [Lorenzo-Seva and Ferrando,
2006] and analyzed the polychoric correlation matrix
given that many of the items had asymmetric
distributions. Five eigenvalues were greater than
1.0 (12.549, 3.844, 2.263, 1.311, and 1.277); however,
a parallel analysis with 500 randomly generated
matrices suggested three dimensions (95th percentile
of random eigenvalues: 1.672, 1.576, and 1.510).
A parallel analysis is superior to the K1 rule (i.e.,
extract the number of factors based on eigenvalues
above 1.00) for deciding on the appropriate number
of factors to extract [e.g., Goldberg and Velicer,
2006; Russell, 2002]. This approach is based on the
consideration of the eigenvalues that emerge from
random datasets that have the same sample size and
items as the actual dataset. The underlying logic is
that the factor analyst should only extract the
number of factors that correspond to the number
of eigenvalues in the real dataset that are well above
the eigenvalues obtained from simulated data. See
Russell [2002, p 1633] for more detail about parallel
analyses.
Table I displays the pattern coefficients from the
three-factor solution (structure coefficients are
available upon request). The factors were moder-
ately correlated: r5.34 for the rule-breaking (I) and
social aggression factors (II); r5.49 for the rule-
breaking (I) and physical aggression factors (III);
r5.39 for the social aggression (II) and physical
aggression factors (III). Items generally showed
relatively high independent associations (i.e., Z.40)
with their respective factor and generally low cross-
loadings (i.e., r.30). However, a handful of the
physical aggression items with content involving
threatening, hitting, or fighting others tended to
cross-load on the rule-breaking factor (e.g., ‘‘threa-
tened others,’’ ‘‘felt better after hitting,’’ ‘‘Got into
fights more than the average person’’). In psycho-
logical terms, this may reflect the fact that acts of
physical aggression also tend to violate local rules
and norms. Moreover, selecting items that are
purely unidimensional would be a stringent require-
ment for our scale development efforts that is very
difficult to achieve in practice; indeed, prominent
factor analysts have noted that ‘‘there are few factor
univocal items, most items have secondary factor
loadings of substantial size’’ (p 230) [Goldberg and
Velicer, 2006]. Similar pattern loadings and factor
correlations were obtained when we analyzed the
correlation matrix rather than the polychoric
correlation matrix. Scales created from these items
had relatively high levels of internal consistency
as seen in Table I (all a’s Z.85). The average
inter-item rfor physical aggression items was .37,
the average inter-item rfor social aggression items
was .36, and the average inter-item rfor rule-
breaking items was .38.
Correlations with Personality Traits
Associations between the resulting STAB scales
and the MPQ-BF superfactors are reported in
Table II. Table II displays zero-order correlations
and partial correlations controlling for the overlap
between the STAB scales. The latter are important
given the oblique nature of our scales (i.e., these
partial correlations capture subtype-specific associa-
tions with personality). Importantly, results suggest
that physical aggression, social aggression, and
rule-breaking were linked to relatively distinct
personality configurations, even when controlling
for overlap with the other STAB scales. Physical
aggression was associated with high NEM and low
CON whereas social aggression was uniquely
associated with high NEM. Rule-breaking was
associated with all three personality dimensions.
The association with physical aggression may
partially reflect the fact that NEM includes an
Aggression subscale (which was correlated .58 with
the STAB physical aggression measure as compared
with .24 for social aggression and .42 for rule-
breaking). In any case, such findings hint at
potentially important distinctions between subtypes
of antisocial behavior in terms of their personality
correlates.
1
Of our ten physical aggression items, two appear to relate more
specifically to anger (i.e., ‘‘got angry quickly,’’ ‘‘had trouble
controlling temper’’). Importantly, however, several of the most
frequently used measures of physical aggression include items that
directly or indirectly tap anger. For example, the nine-item Physical
Aggression scale in the Buss and Perry [1992] Aggression Ques-
tionnaire includes items that at least indirectly tap anger (i.e., ‘‘I have
become so mad that I have broken things;’’ ‘‘There are people who
pushed me so far that we came to blows’’), as do at least two of six
items on the Physical Aggression scale in the Personality Assessment
Inventory (PAI). In short, items tapping anger are frequently
included in measures of physical aggression. Nonetheless, we re-ran
analyses linking our scales to criterion-related variables omitting
those items to ensure that our results were not dependent on these
items. Associations were essentially identical to those reported
herein. Such findings suggest that the presence of anger items does
not change the positioning of our physical aggression scale in the
broader nomological net.
380 Burt and Donnellan
Aggr. Behav.

STUDY 2: PSYCHOMETRIC REPLICATION AND
CRITERION-RELATED VALIDITY OF THE STAB
Sample
This sample consisted of 500 undergraduate
students (50% women; average age 519) enrolled
in psychology courses at a large public university in
the Midwest. College students participated in
exchange for course credits or extra credit during
Fall Semester 2007. Data were collected over the
Internet using an anonymous web-based interface.
The ethnic breakdown was Caucasian (84%),
African–American (5%), Asian or Pacific Rim
(8%), Hispanic/Latino (1%), and other (2%)
TABLE I. Pattern Coefficients for 32-Item STAB (Study 1)
I II III
Physical aggression (a5.85)
Felt like hitting people .09 .13 .61
Got angry quickly .25 .29 .45
Hit back when hit by others .12 .23 .66
Threatened others .42 .08 .53
Had trouble controlling temper .02 .32 .54
Hit others when provoked .30 .13 .69
Got into fights more than the average person .49 .01 .37
Swore or yelled at others .04 .24 .42
Got into physical fights .55 .25 .48
Felt better after hitting .40 .02 .49
Social aggression (a5.86)
Blamed others .15 .51 .04
Tried to hurt someone’s feelings .11 .41 .27
Made fun of someone behind his/her back .44 .81 .48
Excluded someone from group activities when angry with him/her .06 .68 .00
Intentionally damaged someone’s reputation .38 .50 .05
Tried to turn others against someone when angry with him/her .16 .91 .17
Gave someone the silent treatment when angry with him/her .03 .51 .04
Called someone names behind his/her back .06 .62 .02
Revealed someone’s secrets when angry with him/her .41 .53 .17
Was rude towards others .12 .54 .04
Made negative comments about other’s appearance .05 .55 .07
Rule-breaking (a5.87)
Broke into a store, mall or warehouse .89 .00 .02
Broke the windows of an empty building .91 .02 .07
Shoplifted things .73 .14 .05
Littered public areas by smashing bottles, tipping trash cans, etc. .73 .05 .03
Stole a bicycle .88 .06 .09
Stole property from school or work .63 .16 .07
Left home for an extended period of time without telling family/friends .72 .09 .06
Sold drugs, including marijuana .83 .01 .06
Was suspended, expelled, or fired from school or work .77 .07 .10
Had trouble keeping a job .80 .05 .01
Failed to pay debts .56 .21 .04
TABLE II. Associations Between the STAB Scales and Personality Traits (Study 1)
Positive Emotionality Negative Emotionality Constraint
rPartial rr Partial rrPartial r
STAB Scale
Physical aggression .04 .05 .42

.26

.21

.12

Social aggression .05 .01 .35

.20

.06 .06
Rule-breaking .17

.16

.32

.11

.24

.17

Partial correlations control for other STAB scales. For example, the partial correlations for personality and Physical aggression control for Social
Aggression and Rule-breaking.

Po.05.
381STAB Questionnaire
Aggr. Behav.

ethnicities. Research protocol was approved by the
Michigan State University IRB. All participants
provided informed consent.
Measures
Antisocial behaviors. Participants completed
the 32-item STAB (all a4.77; see Table III), again
reporting on behaviors during the past year.
Participants also completed the 114 antisocial
behavior items from the ESM [Krueger et al.,
2007]. They specifically completed the following
ESM scales: Physical Aggression, Relational
Aggression, Theft, Rebelliousness, Destructive
Aggression, Honesty, and Fraud. The scale names
are generally quite reflective of item content with the
exception of Destructive Aggression, which includes
items that inquire about vandalism and the destruc-
tion of property as opposed to acts of personal
violence perpetrated on others (e.g., ‘‘I have
vandalized public property just for kicks’’). We
selected these particular scales both because their
content maps to the STAB scales and because this
measure was not consulted during the item genera-
tion phase of STAB. Accordingly, comparison with
these scales allows us to meaningfully assess both the
convergent and discriminant validity of the STAB.
We expected that the physical aggression and social
aggression scales would be strongly associated with
Physical Aggression and Relational Aggression,
respectively, and more weakly associated with Theft,
Rebelliousness, Destructive Aggression, Honesty,
and Fraud. We further expected rule-breaking to
be strongly associated with Theft, Rebelliousness,
Destructive Aggression, Honesty, and Fraud, and
more weakly associated with Physical Aggression
and Relational Aggression. Reliability information
for these scales is reported in Table V.
TABLE III. Pattern Coefficients for 32-Item STAB (Study 2)
I II III
Physical aggression (a5.84)
Felt like hitting people .18 .15 .69
Got angry quickly .36 .39 .48
Hit back when hit by others .03 .09 .68
Threatened others .11 .18 .55
Had trouble controlling temper .04 .26 .45
Hit others when provoked .13 .12 .74
Got into fights more than the average person .37 .00 .49
Swore or yelled at others .02 .31 .39
Got into physical fights .31 .14 .68
Felt better after hitting .18 .03 .65
Social aggression (a5.85)
Blamed others .07 .63 .01
Tried to hurt someone’s feelings .17 .53 .08
Made fun of someone behind his/her back .05 .58 .07
Excluded someone from group activities when angry with him/her .11 .74 .12
Intentionally damaged someone’s reputation .50 .44 .04
Tried to turn others against someone when angry with him/her .26 .71 .14
Gave someone the silent treatment when angry with him/her .04 .49 .04
Called someone names behind his/her back .01 .68 .03
Revealed someone’s secrets when angry with him/her .21 .56 .07
Was rude towards others .20 .62 .19
Made negative comments about other’s appearance .02 .58 .03
Rule-breaking (a5.78)
Broke into a store, mall or warehouse .99 .10 .06
Broke the windows of an empty building .82 .08 .07
Shoplifted things .77 .04 .02
Littered public areas by smashing bottles, tipping trash cans, etc. .76 .06 .05
Stole a bicycle .75 .01 .04
Stole property from school or work .67 .13 .01
Left home for an extended period of time without telling family/friends .54 .16 .01
Sold drugs, including marijuana .86 .10 .03
Was suspended, expelled, or fired from school or work .70 .08 .15
Had trouble keeping a job .67 .07 .04
Failed to pay debts .59 .09 .09
382 Burt and Donnellan
Aggr. Behav.

Personality. Participants completed the 50-
item International Personality Item Pool-Five
Factor Model [IPIP-FFM; Goldberg, 1999], a
measure of the Big Five factors of personality:
Extraversion, Agreeableness, Conscientiousness,
Neuroticism, and Intellect/Imagination (or Open-
ness). As compared with the MPQ-BF, the Big Five
represents the more dominant structural model of
personality structure in the literature, and it was
thus important to extend our prior personality
findings to the Big Five. That said, the two
personality models tap many of the same constructs:
Neuroticism is akin to NEM (especially the Stress
Reaction scale), Conscientiousness is akin to CON,
and Extraversion is akin to PEM. Agreeableness
includes aspects of low NEM (i.e., the Aggression
and Alienation scales) and high PEM (i.e., the Social
Closeness scale). Big Five reliability information is
reported in Table VI.
STUDY 2 RESULTS
EFA
We subjected the 32 STAB items to an EFA with a
normalized promax rotation based on the polycho-
ric correlation matrix using FACTOR 7.0 [Lorenzo-
Seva and Ferrando, 2006]. Six eigenvalues were
greater than 1.0 (11.914, 3.452, 2.100, 1.226, 1.179,
and 1.039); however, a parallel analysis with 500
random matrices once again suggested three dimen-
sions (95th percentile of random eigenvalues: 1.588,
1.506, and 1.445). Pattern loadings (see Table III)
were quite similar to those reported in Study 1. To
quantify this impression, we calculated Tucker
congruence coefficients for the pattern loadings for
the three factors; these were acceptably high
(coefficients 5.95, .96, and .94 for factors I, II,
and III, respectively). Moreover, many of the cross-
loadings observed in Study 1 did not clearly replicate
in this study. We thus concluded that there was
additional support for the three-factor structure of
the STAB.
Confirmatory Factor Analysis (CFA)
Although the EFA results were consistent across
Studies 1 and 2, a more rigorous evaluation of the
three factor model underlying the STAB would be
accomplished using CFA. However, the evaluation
of model fit within a CFA context is adversely
affected by unspecified but nonetheless minor cross-
loadings [Lee and Ashton, 2007]. Indeed, the two
previous EFA results indicated that most items had
minor relations with other factors and thus cross-
loadings would be present. At the same time, we also
suspected that many of the cross-loadings would be
sample specific given the subtle differences between
the EFA results in Studies 1 and 2. One solution to
the issue of fluctuating cross-loadings is to construct
item parcels for use in the CFA such that item
specific idiosyncrasies are potentially ‘‘washed out’’
in the process of aggregation. The advantages of
creating parcels of related items is that true
error variance and item-specific variance are
minimized (i.e., they account for less overall
variance in the newly created composite) whereas
true score variance is increased [see Little et al.,
2002, pp 155–158]. Put differently, combining
a number of relatively imperfect items results in
a composite measure that typically has more
desirable psychometric properties than any of
the individual single items. This is the same
rationale behind the practice of summing related
items into scales for use in typical data analytic
contexts.
We thus created three parcels of items for each of
our hypothesized latent variables. We first rank-
ordered the pattern coefficients from Study 1 for
each primary factor and then selected items
according to this ordering. For example, for
physical aggression, Parcel 1 contained the 1st best
coefficient (i.e., ‘‘Hit others when provoked’’), the
6th best coefficient (‘‘Felt better after hitting’’), and
the 7th best coefficient (‘‘Got into physical fights’’).
Parcel 2 contained the 2nd best coefficient (‘‘Hit
back when hit by others’’), the 5th best coefficient
(‘‘Threatened others’’), and 8th best coefficient
(‘‘Got angry quickly’’). Parcel 3 contained the
3rd best coefficient (‘‘Felt like hitting people’’), the
4th best coefficient (‘‘Had trouble controlling
temper’’), the 9th best coefficient (‘‘Swore or yelled
at others’’), and the 10th best coefficient (‘‘Got
into fights more than the average person’’).
A similar strategy was followed to create parcels
out of the social aggression and rule-breaking
items. A total of 498 participants had parcel level
data.
An inspection of descriptive statistics indicated
that the three rule-breaking parcels were consider-
ably kurtotic (range of values: 6.709–10.494) and
skewed (range of values: 2.431–2.934). As the
data departed from multivariate normality assump-
tions, we elected to use the Mplus program
(version 5.2) to conduct the CFA so we could
use maximum likelihood estimation with robust
standard errors (MLR). Factor variances were
fixed to 1.0 and all parcel loadings and factor
383STAB Questionnaire
Aggr. Behav.

covariances were freely estimated. The fit for the
initially specified model was more or less acceptable
(w
2
575.102, df 524; CFI 5.973; TLI 5.960;
RMSEA 5.065). This judgment was based on the
common rule of thumb that adequate models should
have CFI and TLI values of around .95 or higher as
well as RMSEA values less than .08 [see Brown,
2006, p 87].
Nonetheless, we examined modification indices to
identify sources of misfit. We then evaluated
whether the inclusion of any additional parameters
would change the substantive interpretations we
drew from the model. We focused only on modifica-
tions associated with parcel cross-loadings (i.e., we
ignored correlated parcel-specific residuals). The
largest value was associated with a secondary
loading from the first physical aggression parcel to
the rule-breaking latent factor. Accordingly, we
specified this loading and the fit generally improved
(w
2
560.825, df 523; CFI 5.980; TLI 5.969;
RMSEA 5.057). Relevant results are reported in
Table IV. As seen there, the unanticipated pattern
loading was not substantial (.196). Correlations
between latent factors were as follows (note that
because latent factors are attenuated for measure-
ment error, these correlations are slightly higher
than those among the observed STAB scales):
physical aggression and social aggression: r5.61;
physical aggression and rule-breaking: r5.44; and
social aggression and rule-breaking: r5.42. When
viewed in conjunction with the EFA results from
Study 2, these results suggest that a three-factor
structure for STAB was more or less reasonable and
replicable.
Associations with Krueger et al. [2007] ESM
Scales
The associations between the STAB scales and the
Krueger et al. [2007] ESM scales are presented in
Table V. As seen there, the STAB physical aggres-
sion scale was strongly correlated with the ESM
Physical Aggression scale (r5.67). This strong
association across physical aggression scales per-
sisted even when controlling for the other two STAB
scales (partial r5.59) and, moreover, was signifi-
cantly larger than the partial correlations between
STAB physical aggression and the other ESM scales
(Dw
2
Z58.5 on 1 df, all P’so.0001; remaining partial
r’sr.19). Furthermore, the unique association of
ESM Physical Aggression with STAB physical
aggression was significantly larger (Dw
2
Z29.8 on 1
df, both P’so.0001) than its corresponding associa-
tions with STAB social aggression and rule-breaking
(partial r’s 5.15 and .32, respectively). Similarly,
the STAB social aggression scale was strongly
associated with the Relational Aggression scale
(r5.64; partial r5.50). Associations between the
STAB social aggression scale and all other ESM
scales were significantly smaller (remaining partial
r’sr.26; Dw
2
Z19.9 on 1 df, all P’so.000). More-
over, the unique association of ESM Relational
Aggression with STAB social aggression was sig-
nificantly larger than its corresponding partial
correlations with STAB physical aggression and
rule-breaking (partial r’s 5.19 and .15, respectively;
Dw
2
Z31.6 on 1 df, both P’so.001). Such results
provide clear evidence of convergent and discrimi-
nant validity for the STAB physical aggression and
social aggression scales.
As expected, the STAB rule-breaking scale was
strongly correlated with several measures in the
ESM, namely Theft, Fraud, Rebelliousness, and
Destructive Aggression (r’s between .52 and .64).
This overlap persisted when variance shared with
the other STAB scales were controlled (i.e., the
partial correlations were also strong, ranging from
.46 to .59). Moreover, these unique associations of
ESM Theft, Fraud, Rebelliousness, and Destructive
Aggression with STAB rule-breaking were signifi-
cantly larger than their corresponding partial
correlations with STAB physical and social
aggression (Dw
2
Z13.3 on 1 df, all P’so.001). By
contrast, the unique associations of the STAB
rule-breaking scale with ESM Physical Aggression
and Relational Aggression scales (i.e., partial
r’s 5.32 and .15, respectively) were significantly
smaller than its partial correlations with ESM
Theft, Fraud, Rebelliousness, and Destructive
Aggression (Dw
2
Z6.8 on 1 df, all P’so.01). Indeed,
the only ESM scale that did not perform as
expected was Honesty, which was moderately
associated with all three STAB scales. Nevertheless,
the totality of the evidence leads us to conclude that
the STAB rule-breaking scale also has convergent
and discriminant validity. In sum, these results
provide good evidence of criterion-related validity
for all three STAB scales.
Associations with the Big Five
Associations between the STAB scales and the Big
Five scales are reported in Table VI. Results again
suggest that the three STAB scales were uniquely
associated with relatively distinct personality attri-
butes (i.e., personality results are more distinct when
considering partial correlations as opposed to the
zero-order correlations). As in study 1, both social
384 Burt and Donnellan
Aggr. Behav.

and physical aggression were associated with high
Neuroticism, whereas physical aggression was also
associated with low Agreeableness. Rule-breaking
was associated with Agreeableness, Conscientious-
ness, and to a lesser extent, Neuroticism. Such
findings further highlight the presence of relatively
distinct personality correlates of the three subtypes
of antisocial behavior.
TABLE IV. CFA Results for the STAB Using Item Parcels (Studies 2, 3, 4, and 5)
Study Parcels
Physical
aggression
Social
aggression Rule-breaking R
2
Study 2
(college sample,
n5500)
Physical aggression
Parcel 1 .604 .196 .508
Parcel 2 .839 .704
Parcel 3 .863 .745
Social aggression
Parcel 1 .838 .702
Parcel 2 .792 .628
Parcel 3 .887 .787
Rule-breaking
Parcel 1 .782 .611
Parcel 2 .739 .545
Parcel 3 .789 .622
Study 3
(adjudicated
sample, n5218)
Physical aggression
Parcel 1 .631 .281 .694
Parcel 2 .884 .781
Parcel 3 .913 .833
Social aggression
Parcel 1 .798 .636
Parcel 2 .784 .615
Parcel 3 .835 .697
Rule-breaking
Parcel 1 .811 .657
Parcel 2 .464 .215
Parcel 3 .734 .539
Study 4
(adjudicated
sample, n5155)
Physical aggression
Parcel 1 .632 .294 .736
Parcel 2 .853 .728
Parcel 3 .910 .829
Social aggression
Parcel 1 .819 .670
Parcel 2 .715 .512
Parcel 3 .898 .806
Rule-breaking
Parcel 1 .735 .540
Parcel 2 .655 .430
Parcel 3 .845 .714
Study 5
(community adult
sample, n5398)
Physical aggression
Parcel 1 .577 .211 .500
Parcel 2 .883 .780
Past Year report Parcel 3 .911 .831
Social aggression
Parcel 1 .819 .671
Parcel 2 .825 .681
Parcel 3 .891 .794
Rule-breaking
Parcel 1 .802 .644
Parcel 2 .504 .254
Parcel 3 .827 .683
Completely Standardized Pattern Coefficients Reported.
385STAB Questionnaire
Aggr. Behav.

STUDY 3: INITIAL EVIDENCE OF CRITERION-
RELATED VALIDITY FOR THE STAB IN A
CLINICAL SAMPLE
The primary objectives of Study 3 were to confirm
the factor analytic structure of the STAB in a
criminally active sample and to establish links
between the STAB scales and reports of criminal
convictions. We also wanted to compare mean levels
of the STAB scales across different types of
offenders and compare average levels of the STAB
scales in a criminally active/clinical sample to those
in college and community samples (the latter of
which is presented in Study 5).
Sample
Participants were drawn from an outpatient
psychiatric treatment facility for adjudicated adults
(n5218; 15% women). Clients at these facilities
were either on parole or probation and were
mandated to attend group therapy by the court
system. Treatment groups were centered on one of
three themes: substance abuse (n582), domestic
violence (n5117), and anger management (n519).
The average participant had been convicted of
just over two crimes (with a range of 1–22
convictions) ranging from petty theft to sexual
offenses. Of these, approximately 40% were violent
or physically aggressive offenses (e.g., mugging,
assault, sexually based offenses). The average
age was 30 years old (SD 510 years; range 518–65
years). The ethnic breakdown was 59% Caucasian,
21% African–American, 13% Hispanic/Latino,
and 7% other. Participants received a small
financial incentive ($15). Research protocol
was approved by the Michigan State University
IRB. All participants provided informed
consent.
TABLE V. Associations Between the STAB Scales and the Externalizing Spectrum Model (ESM) Scales (Study 2)
STAB Scales
Physical aggression (a5.84) Social aggression (a5.85) Rule-breaking (a5.78)
rPartial rrPartial rrPartial r
ESM Scales (a)
Physical aggression (.77) .67

.59

.30

.15

.48

.32

Relational aggression (.85) .49

.19

.64

.50

.37

.15

Theft (.75) .28

.02 .23

.01 .64

.59

Fraud (.71) .41

.10

.45

.26

.57

.46

Rebelliousness (.81) .29

.09

.20

.02 .52

.46

Destructive aggression (.84) .32

.11

.22

.04 .59

.53

Honesty (.83) .30

.07 .37

.23

.34

.21

Partial correlations control for other STAB scales. For example, the partial correlation between ESM Physical Aggression and STAB Physical
Aggression controls for STAB Social Aggression and Rule-breaking.

Po.05.
TABLE VI. Associations Between the STAB Scales and the Big Five (Study 2)
STAB Scales
Physical aggression Social aggression Rule-breaking
rPartial rrPartial rrPartial r
Big Five Dimension (a)
Extraversion (.89) .00 .01 .03 .03 .00 .00
Agreeableness (.82) .33

.20

.22

.02 .32

.20

Conscientiousness (.82) .17

.02 .20

.09 .28

.21

Neuroticism (.88) .38

.24

.41

.28

.11

.11

Intellect/Imagination (.80) .01 .01 .01 .00 .06 .06
Partial correlations control for other STAB scales. For example, the partial correlation between Agreeableness and Physical Aggression are
calculated controlling for Social Aggression and Rule-breaking.

Po.05.
386 Burt and Donnellan
Aggr. Behav.

Measures
Participants completed the 32-item STAB (inter-
nal consistency reliabilities are again good, all
a4.70; as presented in Table VII). As with the
two prior studies, they again reported on behaviors
only during the last year. We also collected
information on the lifetime number of criminal
convictions (those offenses that were formally
prosecuted), and the number of these convictions
that were violent in nature (i.e., involved physically
hurting another person), via two single items.
Though these are admittedly crude measures of the
prevalence of criminal offenses, they nevertheless
allowed us to evaluate associations between the
STAB and a measure of criminal activity.
We also administered a reading screen to ensure
that participants had the requisite reading ability. In
particular, we made use of the Test of Word
Reading Efficiency [TOWRE; Torgesen et al.,
1999], a brief measure that captures sight word
reading ability. TOWRE data was missing for only
one participant. The average reading grade equiva-
lent was 9.80, with 12% of the sample falling at or
below a 5th grade reading level. Analyses were
conducted first on the full adjudicated sample, and
then repeated excluding those with reading difficul-
ties. Similar results emerged in either case, and thus
results are reported on the former (the latter are
available upon request).
STUDY 3 RESULTS
CFA
We used the same item parcels for the CFA as
were used in Study 2. Mplus 5.2 was again used for
analyses with the MLR estimator. We also specified
the secondary cross-loading identified in Study 2
(involving the first aggression parcel and the rule-
breaking factor). Results are reported in Table IV.
The fit for the initially specified model was more or
less acceptable (w
2
555.778, df 523; CFI 5.962;
TLI 5.941; RMSEA 5.081). The largest modifica-
tion index value suggested that a cross-loading from
the physical aggression factor to the last social
aggression parcel would improve fit. Although
overall fit improved (w
2
543.049, df 522;
CFI 5.976; TLI 5.961; RMSEA 5.066), the added
cross-loading was actually negative (.41) and it
caused the primary loading for this parcel to exceed
1.0 (1.19). Although a completely standardized
loading above 1.0 is possible [see Brown, 2006, p
149], we elected to retain our original model in light
of the possibility of over-fitting. Correlations be-
tween latent factors were as follows (as before,
because latent factors are attenuated for measure-
ment error, these correlations are slightly higher
than those among the observed STAB scales):
physical aggression and social aggression: r5.67;
physical aggression and rule-breaking: r5.61; and
social aggression and rule-breaking: r5.37.
Associations Between the Past-Year STAB
Scales and Lifetime Criminal Convictions
The past-year physical aggression and rule-break-
ing scales were correlated with self-reported lifetime
number of convictions (both r’s 5.24, P’so.05)
whereas social aggression was not (r5.13,
P5.057). The pattern of results using partial
correlations was similar (physical aggression: partial
r5.14, Po.05; rule-breaking: partial r5.14,
Po.05; social aggression: partial r5.03,
P5.634). In terms of violent crimes, all three scales
had significant zero-order associations (r5.31, .27,
TABLE VII. Comparison of Groups Within Adjudicated Sample (Study 3)
Substance abuse
(n582)
Domestic violence
(n5117)
Anger management
(n519)
M SD M SD M SD
Criminal History
No. of convicted crimes (F(2, 215) 52.6, P5.08) 2.01 1.23 2.44 1.84 1.74 1.79
No. of convicted violent crimes (F(2, 215) 559.7, Po.01) .28 .59 1.27 .64 1.32 .95
% violent crimes (F(2, 215) 5124.4, Po.01) .12 .25 .70 .31 .88 .21
STAB Scales, past year
Physical aggression (a5.91; F(2, 215) 54.5, P5.01) 20.21 6.84 22.80 7.57 24.68 7.69
Social aggression (a5.83; F(2, 215) 53.6, P5.03) 20.78 4.61 22.73 5.63 22.79 5.11
Rule-breaking (a5.71; F(2, 215) 54.0, P5.02) 14.06 4.03 15.70 4.06 15.26 3.00
Mean levels of the STAB scales generally differed across the treatment groups, as indicated by one-way ANOVAs (F-test results are presented on
the left side of the table).
387STAB Questionnaire
Aggr. Behav.

and .16, for physical aggression, rule-breaking,
and social aggression, respectively); however, only
physical aggression and rule-breaking had statisti-
cally significant partial correlations with number of
violent convictions (partial r5.20 and .14, respec-
tively). Variance unique to the social aggression
scale was not significantly associated with self-
reported violent crime (partial r5.04, P5.543).
Collectively, these results suggest that at least two of
the three STAB scales are associated with criminal
convictions, and that relations are primarily evident
for those scales actively assessing actual criminal
acts (viz. physical aggression and rule-breaking).
Comparison of Treatment Groups
We next compared treatment groups (see
Table VII) on the STAB scales, number of convicted
crimes, number of violent convicted crimes, and the
proportion of convicted crimes that were violent in
nature (i.e., no. of violent convictions/no. of total
convictions). Members of the substance abuse
treatment groups reported far fewer violent criminal
convictions than those in the other groups, as
evaluated via independent samples t-tests (Cohen’s
deffect size comparing the substance group to a
combined domestic violence and anger management
group was 1.56, Po.01) and somewhat fewer overall
criminal convictions, though not significantly so
(Cohen’s d50.22, ns). It thus appears that the
substance abuse treatment group contained indivi-
duals engaging in relatively low levels of violent
antisocial behavior. Importantly, this pattern per-
sisted to the STAB scales (Cohen’s dcomparing the
substance group to a combined domestic violence
and anger management group ranged from 0.38 to
0.40, all P’so.01; again evaluated via independent
samples t-tests), suggesting that the STAB is also
able to detect differences in antisocial activity even
within high-risk clinical samples.
STUDY 4: ADDITIONAL EVIDENCE OF THE
CRITERION-RELATED VALIDITY FOR THE STAB
IN A CLINICAL SAMPLE
The objectives of Study 4 were to confirm links
between the STAB scales and reports of criminal
convictions in an independent adjudicated sample,
as well as to constructively replicate treatment group
comparisons and associations with a related mea-
sure (i.e., the ESM) in a criminally active sample.
We also sought to offer additional confirmation of
the STAB factor analytic structure. Finally, we
compared average levels of the general STAB scales
in this sample to those in a community sample
(results are presented as part of Study 5).
Sample
Participants were drawn from the outpatient
psychiatric treatment facility described in Study 3.
Only participants who did not participate in Study 3
were eligible for participation in this study (n5155;
17% women). Treatment groups were centered on
one of three themes: substance problems/abuse
(n558), domestic violence (n582), and economic
crimes (n515; e.g., fraud, check forgery, etc.). The
average participant had been convicted of 2.73
crimes (with a maximum of 15 convictions) ranging
from petty theft to sexual offenses. The average age
was 33 years old (SD 511 years; range 518–65
years). The ethnic breakdown was 55% Caucasian,
22% African–American, 10% Hispanic/Latino, and
13% other. Participants received a small financial
incentive ($15) for their time. Research protocol was
approved by the Michigan State University IRB. All
participants provided informed consent.
Measures
Antisocial behaviors. Participants completed
the 32-item STAB (all aZ.80; as presented in
Table VIII). However, unlike in prior studies, they
reported on behaviors in general (i.e., without
regard to a specific time period), an assessment
strategy that we expected to capture typical or trait-
like levels of antisocial behavior. We also collected
information on the lifetime number of criminal
convictions and the number of these convictions that
were violent in nature via the same two items
administered in Study 3. Participants completed 114
antisocial behavior items from the ESM [Krueger
et al., 2007; ESM scales: Physical Aggression,
Relational Aggression, Theft, Rebelliousness,
Destructive Aggression, Honesty, and Fraud), as
examined in Study 2. Reliability information for
these scales is reported in Table VIII.
Reading ability. We again administered the
TOWRE reading screen [Torgesen et al., 1999] to
ensure that participants had the requisite reading
ability. The average reading grade equivalent was
9.32, with 17% of the sample falling at or below
a 5th grade reading level. As before, analyses
were conducted first on the full sample, and then
repeated excluding those with reading difficulties.
Similar results emerged in either case, and thus
results are reported on the former (the latter are
available upon request).
388 Burt and Donnellan
Aggr. Behav.

STUDY 4 RESULTS
CFA
We used the same item parcels and CFA analyses
as used in Studies 2 and 3. We fit the model
identified in Study 2 and the overall fit was
acceptable (w
2
531.072, df 523; CFI 5.988;
TLI 5.981; RMSEA 5.048). Table IV displays the
standardized parameters from this model. Correla-
tions between latent factors were as follows (as a
reminder, latent factors are attenuated for measure-
ment error, and thus these correlations are slightly
higher than those among the observed STAB scales):
physical aggression and social aggression: r5.74;
physical aggression and rule-breaking: r5.67; and
social aggression and rule-breaking: r5.695. When
combined with the factor analytic results from
Studies 1, 2, and 3, these results support the
hypothesized three-factor structure for the STAB.
Associations Between the STAB Scales and
Lifetime Criminal Convictions
The physical aggression and rule-breaking scales
were moderately correlated with the overall number
of lifetime convictions (r’s 5.39 and .50, respec-
tively, P’so.05), as was social aggression (r5.35,
Po.05). In terms of violent crimes, all three scales
again had significant zero-order associations
(r5.42, .35, and .31, for physical aggression, rule-
breaking, and social aggression, respectively). Both
sets of correlations appear stronger than those
reported in Study 3, which may reflect the fact that
the STAB scales in this study were assessed using
something akin to a lifetime or trait-like assessment
(rather than just the past year, as in Study 3).
Moreover, the pattern of partial correlations in these
data was far more specific. Only rule-breaking
remained significantly associated with the overall
number of crimes once we controlled for overlap
among the STAB measures (physical aggression:
partial r5.08, P5.35; rule-breaking: partial
r5.34, Po.05; social aggression: partial r5.04,
P5.62). By contrast, only physical aggression
remained significantly associated with number of
violent crimes (physical aggression: partial r5.24,
Po.05; rule-breaking: partial r5.10, P5.22; social
aggression: partial r5.03, P5.76). As with Study 3,
these results collectively suggest that only two of the
three STAB scales are associated with criminal
convictions, and that these relations are specific
to those scales assessing actual criminal acts (viz.
physical aggression and rule-breaking). The current
results extend conclusions from Study 3, however, as
the general STAB physical aggression and rule-
breaking scales evidenced unique associations with
violent crime and overall crime, respectively. Such
findings further augment the validity of those
particular scales.
We then compared our treatment groups (via
independent samples t-tests) on the STAB scales, the
overall number of convicted crimes, number of
violent convicted crimes, and the proportion of
convicted crimes that were violent in nature.
Because of the small sample size, the economic
crimes group was omitted from these treatment
group comparisons (but was included in other
analyses with these data). Members of the substance
abuse treatment groups again reported far fewer
violent criminal convictions than those in the
domestic violence group (mean (SD) 50.13 (0.39)
and 1.12 (0.84), respectively; Cohen’s d51.51,
Po.0001) and fewer criminal convictions overall
(mean (SD) 52.13 (1.98) and 3.31 (3.25), respectively;
TABLE VIII. Associations Between the STAB Scales and the Externalizing Spectrum Model (ESM) Scales (Study 4)
STAB Scales
Physical aggression (a5.89) Social aggression (a5.84) Rule-breaking (a5.80)
rPartial rrPartial rrPartial r
ESM Scales (a)
Physical aggression (.86) .80

.60

.58

.00 .65

.31

Relational aggression (.70) .50

.06 .62

.39

.50

.18

Theft (.85) .43

.00 .39

.03 .70

.60

Fraud (.81) .35

.19

.50

.29

.60

.48

Rebelliousness (.87) .43

.04 .50

.22

.62

.44

Destructive aggression (.91) .52

.10 .48

.08 .69

.52

Honesty (.81) .30

.16 .41

.18

.57

.46

Partial correlations control for other STAB scales. For example, the partial correlation between ESM Physical Aggression and STAB Physical
Aggression controls for STAB Social Aggression and Rule-breaking.

Po.05.
389STAB Questionnaire
Aggr. Behav.

Cohen’s d50.44, Po.05). It thus appears that the
substance abuse treatment group again contained
individuals engaging in relatively low levels of
antisocial behavior. Importantly, this pattern per-
sisted to the STAB physical aggression (mean
(SD) 520.24 (6.10) and 24.02 (6.41) for substance
abuse and domestic violence groups, respectively;
Cohen’s d50.61, Po.01) and rule-breaking scales
(means (SD) 515.05 (4.91) and 17.75 (5.05) for
substance abuse and domestic violence groups,
respectively; Cohen’s d50.53, Po.01), but not to
the social aggression scale (mean (SD) 522.54 (5.13)
and 23.56 (5.62) for substance abuse and domestic
violence groups, respectively; Cohen’s d50.18, ns).
Such results offer additional evidence that the
physical aggression and rule-breaking scales on the
STAB are able to detect incremental differences in
actual antisocial activities even within criminally
active samples.
Associations with Krueger et al. [2007] ESM
Scales
The associations between the STAB scales and the
Krueger et al. [2007] ESM scales are presented in
Table VIII. As in study 2, the STAB physical
aggression scale was strongly correlated with the
ESM Physical Aggression scale (r5.80; partial
r5.60), an association that was significantly larger
than the partial correlations between STAB physical
aggression and the other ESM scales (Dw
2
Z21.5 on
1 df, all P’so.0001; remaining partial r’sr.16).
Moreover, the unique association of ESM Physical
Aggression with STAB physical aggression was
significantly larger than its corresponding associa-
tions with STAB social aggression and rule-breaking
(partial rs5.00 and .31, respectively; Dw
2
Z10.6 on 1
df, both P’sr.001). The STAB social aggression
scale was again strongly associated with the Rela-
tional Aggression scale (r5.62; partial r5.39).
Although a trend was observed, this association
was no longer significantly larger than associations
with all other ESM scales (remaining partial
r’sr.29). However, the unique association of ESM
Relational Aggression with STAB social aggression
was significantly larger than its corresponding
associations with STAB physical aggression and
rule-breaking (partial r’s 5.06 and .18, respectively;
Dw
2
Z4.01 on 1 df, both P’so.05).
Finally, as before, the STAB rule-breaking scale
was strongly and largely uniquely associated with
Theft, Fraud, Rebelliousness, and Destructive
Aggression (r’s between .60 and .70; partial r’s
between .44 and .60). This association was larger
than the partial correlation with ESM Relational
Aggression (Dw
2
Z6.40 on 1 df, P5.01; partial
r5.18), but was no longer significantly larger than
associations with ESM Physical Aggression (partial
r5.31). However, the unique associations of ESM
Theft, Fraud, Rebelliousness, and Destructive
Aggression with STAB rule-breaking were signifi-
cantly larger than their corresponding partial
correlations with STAB physical and social aggres-
sion (Dw
2
Z4.70 on 1 df, all P’sr.05). In short, the
only substantive difference between these results and
the corresponding analyses in Study 2 was the
association with Honesty. In the current sample,
Honesty performed as initially anticipated, eviden-
cing strong and largely unique associations with
rule-breaking (partial r5.46; this association is
significantly larger than the corresponding associa-
tions with STAB physical and social aggression,
Dw
2
Z7.56 on 1 df, both Pr.01). Such results
therefore provide important additional evidence of
convergent and discriminant validity for all three of
the STAB scales, and moreover, suggest that
such relations are not specific to college student
populations.
STUDY 5: COMPARISONS OF ADJUDICATED,
COLLEGE, AND COMMUNITY SAMPLES
The core objective of Study 5 was to offer further
evidence of the validity for the STAB via compar-
isons of mean levels of STAB physical aggression,
rule-breaking, and social aggression across the three
sample types (adjudicated, college, and community
adults). To achieve this goal, we collected additional
STAB data from a community sample of adults
using an Internet-based market research company.
These community data were then compared to our
adjudicated samples (i.e., samples 3 and 4) and
college students’ samples (i.e., samples 1 and 2). We
also examined the impact of gender and age on
STAB scale scores across the three sample types.
Finally, we sought to confirm the STAB factor
analytic structure in a sample of community adults.
Sample
The community sample consisted of 402 adults
(60% women) recruited through Zoomerang, a
market research company that maintains a database
of 2.5 million Internet users. Individuals opt-in to
take up to four surveys a month in exchange for
points that are redeemable for gifts, such as movies,
music, gift cards, and other merchandise. Data were
collected over the Internet. Four participants were
390 Burt and Donnellan
Aggr. Behav.

excluded from analysis because they did not indicate
gender so the sample size used for analyses was 398.
Participants’ average age was 43 years old (SD 516
years; range 518–66 or more years). The ethnic
breakdown was 70% Caucasian, 5% African–
American, 5% Hispanic/Latino, 4% Asian, and
14% other (2% of the sample did not provide this
information). Research protocol was approved by
the Michigan State University IRB. All participants
provided informed consent.
Measures
Antisocial behaviors. Participants completed
the 32-item STAB two times, reporting on their
behaviors over the last year and in general (all
aZ.80). The two administrations were separated by
a 40-item questionnaire unrelated to the goals of this
study. The order of administration was counter-
balanced across the sample (i.e., 50% reported on
the past year first followed by their general report,
while the other 50% completed the STAB adminis-
trations in the reverse-order). As compared with
men who completed the STAB questionnaires in
order 1 (i.e., past year, general), men who completed
the STAB in order 2 (i.e., general, past year)
reported equivalent levels of all past year antisocial
behaviors, as well as general rule-breaking. The two
administration groups did differ, however, on
general physical aggression and social aggression,
such that those completing the general assessment
first endorsed more general physical aggression and
social aggression than those completing the past
year assessment first (Cohen’s d50.56 and 0.34,
respectively, both P’so.05). Among women, differ-
ences emerged only for past-year rule-breaking
(which was slightly lower in those completing the
general STAB first; Cohen’s d50.36, Po.05) and
general physical aggression (which was again slightly
higher in those completing the lifetime STAB first;
Cohen’s d50.31, Po.05). The only consistent
difference across the STAB orderings was thus
observed for the general physical aggression scale.
Importantly, however, physical aggression scores
were significantly higher for general reports as
compared with past year reports regardless of
administration order (both P’so.001), a pattern
that persisted to the social aggression and rule-
breaking scales as well (all P’so.001). In short,
although mean levels of general physical aggression
(but not past year physical aggression) varied by
order of STAB administration, all participants
reported higher levels of general physical aggression,
social aggression, and rule-breaking than past year
physical aggression, social aggression, and rule-
breaking, respectively (as would be expected).
STUDY 5 RESULTS
CFA
We used the same item parcels (for past year
items) and CFA analyses as used in Studies 2, 3,
and 4. We fit the model identified in Study 2 and the
overall fit was acceptable (w
2
562.528, df 523;
CFI 5.970; TLI 5.954; RMSEA 5.066). Table IV
displays the standardized parameters from this
model. Correlations between latent factors were as
follows (as a reminder, latent factors are attenuated
for measurement error, and thus these correlations
are slightly higher than those among the observed
STAB scales): physical aggression and social aggres-
sion: r5.79; physical aggression and rule-breaking:
r5.50; and social aggression and rule-breaking:
r5.60. When combined with the factor analytic
results from Studies 1, 2, 3, and 4, these results offer
strong support for the hypothesized three-factor
structure of the STAB.
Demographic Correlates of the STAB Scale
Scores
We first examined the impact of age and gender on
the STAB scale scores, combining all five samples
(i.e., those from studies 1–5) for analysis. We
expected men to report higher levels of physical
aggression and rule-breaking than women, whereas
we expected that women would score higher than
men on social aggression. We also expected
antisocial behaviors to generally decrease with
age, consistent with prior research [Hirschi and
Gottfredson, 1983; Moffitt, 1993]. Gender (men 51,
women 52) was negatively associated with physical
aggression and rule-breaking, such that men
reported higher levels of these particular behaviors
than did women (point-biserial correlations ranged
from .11 to .25, all P’so.01). Such findings are
clearly consistent with our expectations and with
prior research. By contrast, there was no association
observed between gender and social aggression
(r5.03), results that were not consistent with our
hypotheses. That said, other studies have similarly
found that social aggression does not vary across
gender [Forrest et al., 2005]. Age was associated
with all three scales (both past year and general) in
the expected direction, with correlations ranging
from .11 to .36 (all P’so.01). In short, all three
forms of antisocial behavior were more pronounced
391STAB Questionnaire
Aggr. Behav.

in younger participants. Such findings further
buttress the validity of the STAB physical aggres-
sion, rule-breaking, and social aggression scales.
Mean Comparisons Across Sample Types
The primary objective in Study 5 was to examine
mean differences on the STAB scales across our
three sample types (i.e., college students, community
adults, and adjudicated adults), analyses that would
provide a final and important indication of the
validity of the STAB scales. The specific types of
samples analyzed here were thought to be particu-
larly advantageous in regards to this objective.
Adults in the community were expected to report
the lowest levels of antisocial behavior. By contrast,
antisocial behavior in college students should be
relatively common, reflecting the transient, adoles-
cent-limited antisocial behavior characteristic of late
adolescence/emerging adulthood [Moffitt, 1993].
Antisocial behavior in an adjudicated sample was
expected to be the most common, and should
represent clinically significant and longer term
antisocial behavior. Given this, we expected that
the adjudicated sample would report higher mean
scores on the STAB physical aggression and rule-
breaking scales than both the college students and
the community adults. We also expected that college
students would report higher levels of STAB
physical aggression and rule-breaking than would
the community adults. We did not have strong prior
expectations regarding social aggression, as social
aggression is not generally prosecuted as a criminal
act (and thus the adjudicated sample need not have
higher scores on social aggression).
We conducted analyses separately by sex because
there were proportionally more females in the
college and community adult samples than in the
adjudicated samples and mean levels of physical
aggression and rule-breaking vary across sex.
Analyses were also conducted separately for past
year STAB administrations and general STAB
administrations. The former set of analyses com-
pared samples 1 and 2 (combined), 3, and 5, whereas
the latter set of analyses compared samples 4 and 5.
Results are presented in Tables IX–XI. Whenever
possible (i.e., Tables IX and XI), we report the
means actually observed in the data, with the goal of
facilitating discussion of the magnitude of sample
differences (via standardized Cohen’s deffect sizes).
There is now increasing recognition that significance
testing per se is not particularly informative [Cohen,
1994; Kline, 2004], yielding information only on the
presence and direction of mean differences. Effect
sizes, by contrast, provide a direct assessment of the
magnitude of a given difference, thereby offering
important additional information regarding the
validity of the current instrument.
As expected, adults from the community had
significantly lower scores on all three STAB scales
when compared with the college students and the
adjudicated adults (see Table IX).
2
Specifically, as
evaluated via independent samples t-tests, adults on
parole or probation reported substantially higher
levels of physical aggression, rule-breaking, and
social aggression during the last year than did adults
in the community, with an average Cohen’s deffect
size of 0.58 (considered moderate-to-large in magni-
tude). The differences observed for physical aggres-
sion and rule-breaking appeared slightly stronger in
men than in women (Cohen’s d’s ranged from 0.59
to 0.72 for men and 0.30 to 0.57 for women),
whereas differences on social aggression appeared
stronger in women than in men (Cohen’s deffect
sizes of 0.72 vs. 0.59, respectively), although all
effects were moderate-to-large in magnitude. That
said, the very small number of women in the
adjudicated sample prohibits any firm conclusions.
Next, college students reported somewhat lower
levels of physical aggression and rule-breaking than
did the adjudicated adults, although these effect
sizes were only small-to-medium in magnitude
(ranging from .15 to .44). The two samples did not
differ in their social aggression scores, however (and,
in any case, the trend was for higher social
aggression in the college students than in the
adjudicated adults). Finally, we compared the
community adult and college student samples.
College students reported significantly more physi-
cal aggression, social aggression, and rule-breaking
than did adults in the community, with an average
Cohen’s deffect size of 0.45 (with the exception of
2
Importantly, these main effects of sample persisted even when
statistically controlling for the effects of sex. We conducted a 2 (sex)
3 (sample) ANOVA for each STAB scale: physical aggression
(main effect of sample F(2,1507) 535.0, Po.001; main effect of sex
F(1,1507) 514.2, Po.001; sample-type by sex interaction
F(2,1507) 53.7, Po.05), rule-breaking (main effect of sample
F(2,1507) 58.9, Po.001; main effect of sex F(1,1507) 525.5,
Po.001; sample-type by sex interaction F(2,1507) 59.7, Po.001),
and social aggression (main effect of sample F(2,1507) 583.0,
Po.001; main effect of sex F(1,1507) 53.0, P5.08; sample-type by
sex interaction F(2,1507) 50.3, P5.75). This main effect of sample
persisted even when additionally modeling age as a covariate
(physical aggression: F(2,1500) 512.0, Po.001; rule-breaking:
F(2,1500) 57.4, P5.001; social aggression: F(2,1500) 55.4,
Po.01). Such findings suggest that the main effects of sample
observed in Tables IX and X are robust to simultaneous considera-
tions of sex.
392 Burt and Donnellan
Aggr. Behav.

rule-breaking in women). As before, the differences
observed for physical aggression and rule-breaking
appeared stronger in men (average Cohen’s deffect
size was 0.49 for men and 0.08 for women), whereas
differences on social aggression across samples
appeared larger in women (Cohen’s deffect sizes
of 0.69 vs. 0.84, respectively).
The above results thus suggest that adults in the
community (and particularly men) demonstrate
significantly less physical aggression and rule-break-
ing than either the adjudicated adults or the college
students, the latter of whom also evidence less
physical aggression and rule-breaking than the
adjudicated adults. Such findings are fully consistent
with our expectations. That said, however, the
samples differed on key demographic variables that
were also associated with the STAB scales. Specifi-
cally, although we accounted for the different
proportion of women across samples, we have not
accounted for age effects (mean ages were 19 years
for college students, 30 years for adjudicated adults,
and 43 years for community adults). We thus
conducted mean comparisons using a general linear
model analysis with age as a covariate. We expected
that adjudicated adults, as the only high-risk clinical
sample, would continue to evidence higher levels of
physical aggression and rule-breaking than either
the college students or the community adults.
However, because adolescent-limited antisocial be-
havior is expected to be largely transient [Moffitt,
TABLE X. Comparison of Past Year STAB Across Samples, With Age as a Covariate
Estimated marginal means
Adults College Adjudicated
mean (SE) mean (SE) mean (SE) Significant differences, Po.05
Males n5158 n5440 n5185
AGG (F(2,777) 513.6, Po.01) 19.79 (0.69) 19.94 (0.36) 22.66 (0.47) Adult, collegeoadjudicated
RB (F(2,777) 56.1, Po.01) 14.12 (0.48) 14.18 (0.25) 15.46 (0.33) Adult, collegeoadjudicated
SA (F(2,777) 54.3, P5.01) 20.03 (0.63) 21.91 (0.33) 22.13 (0.42) Adultocollege, adjudicated
Females n5240 n5457 n533
AGG (F(2,722) 58.3, Po.01) 19.15 (0.45) 17.26 (0.29) 20.29 (0.88) Collegeoadult, adjudicated
RB (F(2,722) 512.5, Po.01) 13.40 (0.22) 12.09 (0.14) 3.65 (0.44) Collegeoadult, adjudicated
SA (F(2,722) 52.2, P5.12) 21.11 (0.46) 22.34 (0.30) 22.63 (0.92) None
AGG, RB, and SA represent the past year STAB scales of physical aggression, rule-breaking, and social aggression, respectively. These analyses
compare samples 1 and 2 (i.e., college), 3 (i.e., adjudicated), and 5 (i.e., adults). Mean levels of the STAB scales generally differed across the
sample types, as indicated by ANOVAs (F-test results are presented on the left-hand side of the table). Estimated marginal means are presented
with age as a covariate. Least Significant Difference (LSD) pairwise comparisons were used to statistically compare these estimated marginal
means across samples.
TABLE IX. Comparison of Past Year STAB Across Samples
Cohen’s d
Adults College Adjudicated Adjudicated Adjudicated College
mean (SD) mean (SD) mean (SD) vs. adults vs. college vs. adults
Males n5158 n5440 n5185
AGG (F(2,780) 527.3, Po.01) 17.39 (5.92) 20.96 (5.91) 22.28 (7.62) .72

.19

.60

RB (F(2,780) 511.8, Po.01) 13.05 (3.42) 14.63 (4.78) 15.29 (4.12) .59

.15.38

SA (F(2,780) 530.4, Po.01) 18.44 (6.18) 22.58 (5.73) 21.88 (5.38) .59

.13 .69

Females n5240 n5457 n533
AGG (F(2,727) 511.0, Po.01) 16.84 (6.16) 18.47 (4.64) 20.36 (6.09) .57

.35.30

RB (F(2,727) 54.8, Po.01) 12.78 (2.59) 12.42 (2.44) 13.67 (3.24) .30 .44

.14
SA (F(2,727) 558.1, Po.01) 18.87 (5.78) 23.52 (5.28) 22.70 (4.75) .72

.16 .84

AGG, RB, and SA represent the past year STAB scales of physical aggression, rule-breaking, and social aggression, respectively. STAB scales
were summed across items. These analyses compare samples 1 and 2 (i.e., college), 3 (i.e., adjudicated), and 5 (i.e., adults). Mean levels of the
STAB scales differed significantly across the sample types, as indicated by ANOVAs (F-test results are presented on the left-hand side of the
table). A positive Cohen’s deffect size (right-hand side of the table) indicates that mean levels are higher in the adjudicated sample (columns 1 and
2 on the right side of the table) or in the college sample (column 3 on the right side of the table).

and

indicate that sample means were
significantly different in an independent-samples t-test at Po.01 and Po.05, two-tailed, respectively. indicates that sample means were
marginally different at Po.10, two-tailed.
393STAB Questionnaire
Aggr. Behav.

1993], and because both college students and adults
in the community constitute nonclinical samples
(albeit in different developmental stages), we
expected that college students and adults would
not evidence different levels of physical aggression
and rule-breaking once we controlled for age.
Results of these analyses are reported in Table X
(note that Cohen’s dcannot be computed for
estimated marginal means, and thus our discussion
of these results centers entirely on the statistical
significance of the differences). Consistent with our
predictions, we again observed a main effect of
sample for the physical aggression and rule-breaking
scales (although social aggression differed across
sample type only in men). Simple pairwise compar-
isons of the estimated marginal means (via Least
Significant Difference) specifically revealed that
adjudicated men continued to report higher levels
of physical aggression and rule-breaking as com-
pared with men in the community and male college
students. Perhaps more importantly, however, male
college students and community adults no longer
differed on either scale. In short, the differences we
had observed on physical aggression and rule-
breaking between college males and men in the
community were largely a function of the age
differences between the two samples. By contrast,
differences with the adjudicated adults appear to be
more robust (or at least, are not a function of age or
ethnicity; latter analyses not shown). These results
did not extend to women, however. Indeed, adult
women in the community reported higher levels of
both physical aggression and rule-breaking than the
college students once age was taken into account.
However, community adults did not differ from the
adjudicated adults (though again, given the very
small sample of adjudicated adults, interpretative
caution is warranted).
Comparisons of the general STAB scales across
the adjudicated and community adult samples (see
Table XI) were also encouraging (as a reminder,
general STAB reports were not collected on college
students). As the particularly small number of
adjudicated females in sample 4 (n527) precluded
meaningful comparisons across the two samples,
these analyses were restricted to men only. Adult
men in the community reported less aggressive,
rule-breaking, and socially aggressive behaviors
than did adult men on parole or probation, as
evaluated via independent samples t-tests (average
Cohen’s deffect size was 0.36, which is considered
small-to-moderate in magnitude). When combined
with comparisons using past year STAB data, such
results are collectively thought to provide important
support for the validity of the STAB scales.
OVERALL DISCUSSION
There is growing recognition that meaningful and
substantively important behavioral distinctions exist
within the broader category of antisocial behavior.
In particular, there is emerging interest in the
distinctions among physically aggressive behaviors,
rule-breaking behaviors, and socially aggressive
behaviors. One potential limitation to accumulating
additional scientific insights into the correlates and
origins of these three varieties of antisocial behavior
is the lack of an efficient self-report assessment in
the public domain. We developed the STAB as a
relatively short but comprehensive assessment of
these three types of antisocial behavior to fill this
gap. The final STAB questionnaire is presented in
Appendix.
The present series of studies provides initial
evidence of the factorial validity, internal consis-
tency, and criterion-related validity of the STAB
scales. The factor structure of the STAB was initially
established a sample of college students, and was
then confirmed in a second sample of college
students, a sample of community adults, and two
samples of adjudicated adults. Internal consistency
reliabilities were also found to be quite good across
all five samples. as ranged from .84 to .91 for
physical aggression, from .83 to .90 for social
aggression, and from .71 to .87 for rule-breaking.
In short, there is consistent evidence supporting the
three factor structure of the STAB as well as
evidence for the internal consistency reliability of
the STAB scales.
TABLE XI. Comparison of the General STAB Scales Across
Samples
Cohen’s d
Adults College Adjudicated Adults vs.
mean (SD) mean (SD) mean (SD) adjudicated
Males n5158 – n5128
AGG 20.77 (6.54) 23.06 (6.60) .35

RB 14.99 (4.72) 17.10 (5.17) .43

SA 21.26 (6.29) 23.09 (5.50) .31

AGG, RB, and SA represent the general STAB scales of physical
aggression, rule-breaking, and social aggression, respectively. STAB
scales were summed across items. These analyses compare samples 4
(i.e., adjudicated) and 5 (i.e., adults) using an independent samples
t-test. A positive Cohen’s deffect size indicates that mean levels are
higher in the adjudicated sample.

indicates that sample means were
significantly different at Po.01.
394 Burt and Donnellan
Aggr. Behav.

We also found consistent support for the criterion-
related validity of the STAB. We first examined
associations between the STAB and a related
measure, the ESM, in a college student sample,
and then sought to extend these findings to a high-
risk clinical sample. Results revealed that, as
compared with other STAB scales, the STAB
physical aggression scale evidenced an especially
strong association with the ESM Physical Aggres-
sion scale and significantly smaller associations with
all other ESM scales. Similarly, the STAB social
aggression scale demonstrated a particularly strong
association with the ESM Relational Aggression
scale and less substantial associations with all other
ESM scales (although the latter were only signifi-
cantly smaller in sample 2). Finally, the STAB rule-
breaking scale was strongly and largely uniquely
associated with ESM Theft, Fraud, Rebelliousness,
Destructive Aggression, and to a lesser extent,
Honesty.
We also evaluated associations with two fre-
quently used conceptualizations of personality (the
Big Three and the Big Five) across two samples.
STAB social aggression was uniquely associated
with NEM/Neuroticism in both samples. STAB
physical aggression, by contrast, was positively
associated with NEM/Neuroticism, and was nega-
tively associated with CON and Agreeableness.
Finally, STAB rule-breaking was associated with
NEM/Neuroticism, Agreeableness, PEM, and Con-
scientiousness/CON. Thus, the various STAB scales
appear to be associated with unique configurations
of personality traits. Moreover, consistent with prior
research [Burt and Donnellan, 2008], only rule-
breaking was consistently associated with impulsiv-
ity, whereas physical aggression appears to have a
stronger link with affective dysregulation. When
combined with the ESM findings detailed above, this
pattern of results collectively provides compelling
evidence of convergent and discriminant validity for
the STAB.
Next, the STAB scales demonstrated the expected
associations with participant demographics. Sex was
negatively associated with physical aggression and
rule-breaking, such that men reported higher levels
of these particular behaviors than did women. By
contrast, there was no evidence for sex differences
for social aggression. The latter finding buttresses
prior reports suggesting that the prevalence of social
aggression does not vary across sex [Forrest et al.,
2005], but is not consistent with other research
indicating that social aggression is more common in
girls as compared to boys [Crick and Zahn-Waxler,
2003]. Future research should further explore the
role of sex in social aggression. Age was associated
with all three scales in the expected direction, such
that all three forms of antisocial behavior were more
common in younger participants than older partici-
pants.
Finally, we evaluated whether the STAB demon-
strated expected mean differences across the various
sample types and clinical treatment groups. The
adjudicated men reported higher levels of physical
aggression and rule-breaking on the STAB as
compared with college students (Cohen’s
d50.15–0.19) and community adults (Cohen’s
d50.59–0.72 past year, 0.35–0.43 in general),
differences that persisted even when controlling for
age effects. Similarly, male college students reported
more STAB physical aggression and rule-breaking
that the community adults (Cohen’s d50.38–0.60).
Importantly, however, the increased prevalence in
college students as opposed to community adults
appears to be primarily a function of normative
differences with age, as controlling for age elimi-
nated (or in the case of women, reversed) these
differences between groups. Social aggression, by
contrast, was more common in adjudicated adults
and college students than in community adults,
across both men and women (Cohen’s
d50.59–0.84), but was equivalent in college and
adjudicated samples. In men, this difference per-
sisted even when controlling for age. Our confidence
in the validity of the STAB was further bolstered by
the fact that the STAB scales detected differences in
criminal convictions across clinical treatment
groups. Specifically, members of the substance abuse
treatment groups reported far fewer violent criminal
convictions than those in the other groups (Cohen’s
d51.56 and 1.51) and somewhat fewer overall
criminal convictions (Cohen’s d50.22 and 0.44),
differences that were detected by the STAB scales.
In sum, these data suggest that the STAB is able
to meaningfully detect differences in antisocial
activity both within and across high-risk and
normative samples. All in all, we believe that there
is a good deal of initial support for the validity of the
STAB.
Furthering the Construct Validity of Subtypes
of Antisocial Behavior
We believe that the STAB will be a useful research
instrument when exploring the construct validity of
different varieties of antisocial behavior. As noted
in the Introduction, there is converging evidence
that physical aggression, rule-breaking, and social
aggression constitute meaningfully distinct, albeit
395STAB Questionnaire
Aggr. Behav.

overlapping, components of the broader construct of
antisocial behavior. These subtypes demonstrate
different developmental trajectories, different
demographic correlates, and personological under-
pinnings, and evidence important etiological distinc-
tions. However, there is much work to be done to
more firmly validate these constructs and ground
them in the literature as a whole. For example, recent
evidence indicates that associations with particular
candidate genes vary across antisocial behavior
subtypes [Burt and Mikolajewski, 2008], and that
the magnitude of genetic and environmental influ-
ences varies across physical aggression and rule-
breaking (i.e., genetic influences are more important
for physical aggression whereas shared environmental
influences are more important for rule-breaking).
Given the latter findings, it is entirely possible
that gene–environment interplay also varies across
the subtypes. Indeed, because gene–environment
interactions typically load on the genetic proportion
of variance in standard twin modeling, the finding
of higher genetic influences on physical aggression
than on rule-breaking is circumstantially consistent
with this possibility. Moreover, we know of no
study exploring etiological distinctions across all
three facets of antisocial behavior. It thus remains
unclear how social aggression fits into this emerging
literature.
Although these gaps in the literature are surprising
given the possible implications of such differences,
we suspect it is driven at least in part by the need to
collect multiple or very long measures to assess
physical aggression, rule-breaking, and social
aggression (particularly since twin studies are known
for the use of brief measures). The STAB would
remedy this situation, thereby allowing future twin
projects [e.g., Klump and Burt, 2006] to do this sort
of work. Similarly, longitudinal work that examines
differential outcomes for physical aggression, rule-
breaking, and social aggression (e.g., conventional
adult life versus prison) would augment the
aforementioned work on etiological distinctions,
serving to firmly cement these three subtypes of
antisocial behavior within the literature. In short,
the STAB may serve as an important tool for
longitudinal and genetically informed studies
(among others), making them easier and cheaper
to conduct since the measure is relatively short and
in the public domain.
Limitations and Conclusions
There are several limitations of the current set of
studies. The first is that we relied in part on
convenience samples of college students, as well as
relatively small samples of adults on parole or
probation and adults in the community. We thus do
not have a good basis for developing norms that
could be defended as representative of a targeted
population. That said, rather than attempting to
estimate mean levels in given populations, our
express goal was to compare average levels
across these sample types so as to provide evidence
that the STAB could meaningful distinguish these
groups. Second, we used late-adolescent and adult
samples. Future work is needed to validate the
STAB with younger participants. Next, additional
work is needed to examine the convergence of
the STAB scales across self- and informant
reports of antisocial behavior. As it stands, we
focused on mono-method strategies (i.e., self-reports
with self-reports) to validate the STAB scales.
Finally, there is a significant body of literature
highlighting subtypes within physical aggression
that are not assessed within the STAB:
proactive (i.e., premeditated, instrumental physical
aggression) and reactive (i.e., impulsive, affective
physical aggression) [Barratt et al., 1997; Davidson
et al., 2000]. Excellent measures of these two
forms of physical aggression already exist [e.g., the
Reactive-Proactive Aggression questionnaire;
Raine et al., 2006], and thus we saw little need to
include them here. Moreover, our goal was to
develop a measure of dimensions within the
broader construct of antisocial behavior (rather
than making finer although no less important
distinctions within the subdomain of physical
aggression per se).
In conclusion, the present report suggests that
the STAB is a promising self-report measure of
physically aggressive, rule-breaking, and socially
aggressive antisocial behaviors. It appears to have
a stable factor structure, reliable scales, and
convergent validity with other longer self-report
measures of antisocial behavior. It also appears
to be suitable for use with community, college,
and adjudicated samples. Furthermore, the STAB
differentiates various groups of offenders (i.e.,
violent vs. substance using), correlates with
self-reports of criminal activity within an adjudi-
cated sample, and demonstrates expected mean
differences across normative and criminally
active samples. All in all, these findings suggest
that the STAB is a useful tool for researchers
who are interested in studying the origins and
correlates of these different forms of antisocial
behavior with a quick and efficient assessment
tool.
396 Burt and Donnellan
Aggr. Behav.

APPENDIX: THE STAB
The following items describe a number of different
behaviors. Please read each item and report how
often you have done this using the following scale.
12 3 4 5
never hardly ever sometimes frequently nearly all the time
1.______ Felt like hitting people
2.______ Broke into a store, mall, or warehouse
3.______ Blamed others
4.______ Hit back when hit by others
5.______ Broke the windows of an empty building
6.______ Tried to hurt someone’s feelings
7.______ Got angry quickly
8.______ Shoplifted things
9.______ Made fun of someone behind their back
10.______ Threatened others
11.______ Littered public areas by smashing bottles, tipping trash
cans, etc.
12.______ Excluded someone from group activities when angry
with him/her
13.______ Had trouble controlling temper
14.______ Stole a bicycle
15.______ Gave someone the silent treatment when angry with
him/her
16.______ Hit others when provoked
17.______ Stole property from school or work
18.______ Revealed someone’s secrets when angry with him/her
19.______ Got into fights more than the average person
20.______ Left home for an extended period of time without telling
family/friends
21.______ Intentionally damaged someone’s reputation
22.______ Swore or yelled at others
23.______ Sold drugs, including marijuana
24.______ Tried to turn others against someone when angry with
him/her
25.______ Got into physical fights
26.______ Was suspended, expelled, or fired from school or work
27.______ Called someone names behind his/her back
28.______ Felt better after hitting
29.______ Failed to pay debts
30.______ Was rude towards others
31.______ Had trouble keeping a job
32.______ Made negative comments about other’s appearance
REFERENCES
Achenbach TM, Rescorla LA. 2003. Manual for ASEBA Adult
Forms Profiles. Burlington, VT: University of Vermont, Research
Center for Children, Youth, & Families.
Barratt ES, Stanford MS, Kent TA, Felthous A. 1997. Neuropsy-
chological and cognitive psychophysiological substrates of
impulsive aggression. Biol Psychiatry 41:1045–1061.
Bennett RJ, Robinson SL. 2000. Development of a measure of
workplace deviance. J Appl Psychol 85:349–360.
Bjorkqvist K, Lagerspetz MJ, Kaukiainen A. 1992. Do girls
manipulate and boys fight? Developmental trends in regard to
direct and indirect aggression. Aggr Behav 18:117–127.
Brown TA. 2006. Confirmatory Factor Analysis for Applied
Research. New York: Guilford.
Burt SA. 2009. Are there meaningful etiological differences within
antisocial behavior? Results of a meta-analysis. Clin Psychol Rev
29:163–178.
Burt SA, Donnellan MB. 2008. Personality correlates of aggressive
and non-aggressive antisocial behavior. Pers Individ Dif 44:
53–63.
Burt SA, Larson CL. 2007. Differential affective responses in those
with aggressive versus non-aggressive antisocial behaviors. Pers
Individ Dif 43:1481–1492.
Burt SA, Mikolajewski AJ. 2008. Preliminary evidence that specific
candidate genes are associated with adolescent antisocial
behavior. Aggr Behav 34:257–265.
Buss AH, Perry M. 1992. The Aggression Questionnaire. J Pers Soc
Psychol 63:452–459.
Cairns RB, Cairns BD, Neckerman HJ, Ferguson LL, Gariepy J-L.
1989. Growth and aggression: 1 childhood to early adolescence.
Dev Psychol 25:320–330.
Cohen D, Strayer J. 1996. Empathy in conduct-disordered and
comparison youth. Dev Psychol 32:988–998.
Cohen J. 1994. The Earth is round (Po.05). Am Psychol
49:997–1003.
Crick NR. 1997. Engagement in gender normative versus nonnor-
mative forms of aggression: links to social-psychological adjust-
ment. Dev Psychol 33:610–617.
Crick NR, Bigbee MA. 1998. Relational and overt forms of peer
victimization: A multi-informant approach. J Counsel Clin
Psychol 66:337–347.
Crick NR, Grotpeter JK. 1995. Relational aggression, gender, and
social-psychological adjustment. Child Dev 66:710–722.
Crick NR, Zahn-Waxler C. 2003. The development of psychopathol-
ogy in males and females: Current progress and future challenges.
Dev Psychopathol 15:719–742.
Crick NR, Casas JF, Mosher M. 1997. Relational and overt
aggression in preschool. Dev Psychol 33:579–588.
Crick NR, Casas JF, Nelson DA. 2002. Toward a more compre-
hensive understanding of peer maltreatment: Studies of relational
victimization. Curr Dir Psychol Sci 11:98–101.
Davidson RJ, Putnam KM, Larson CL. 2000. Dysfunction in the
neural circuitry of emotion regulation—A possible prelude to
violence. Science 289:591–594.
DeMarte JA. 2008. The Heterogeneity of Antisocial Behavior:
Evidence for Distinct Dimensions of Physical Aggression,
Rule-breaking, and Social Aggression. Doctoral Dissertation,
Michigan State University, East Lansing, MI.
Denson TF, Pedersen WC, Miller N. 2006. The Displaced
Aggression Questionnaire. J Pers Soc Psychol 90:1032–1051.
Elliott DS, Ageton SS. 1980. Reconciling race and class differences in
self-reported and official estimates of delinquency. Am Sociol
Rev 45:95–110.
Farrington DP, West DJ. 1971. A comparison between early
delinquents and young aggressives. Br J Criminol 11:341–358.
Forrest S, Eatough V, Shevlin M. 2005. Measuring adult indirect
aggression: The development and psychometric assessment of the
Indirect Aggression Scales. Aggr Behav 31:84–97.
Frick PJ, Lahey BB, Loeber R, Tannenbaum L, Van Horn Y,
Christ MAG, Hart EA, Hanson K. 1993. Oppositional defiant
disorder and conduct disorder: A meta-analytic review of factor
analyses and cross-validation in a clinic sample. Clin Psychol Rev
13:319–340.
Goldberg LR. 1999. A broad-bandwidth, public-domain, personality
inventory measuring the lower-level facets of several five-factor
models. In: Mervielde I, Deary I, De Fruyt F, Ostendorf F, (ed).
Personality Psychology in Europe, Vol. 7. Tilburg, The Nether-
lands: Tilburg University Press, pp 7–28.
397STAB Questionnaire
Aggr. Behav.

Goldberg LR, Velicer WF. 2006. Principles of exploratory factor
analysis. In: Strack S, (ed). Differentiating Normal and Abnormal
Personality, 2nd edition. New York, NY: Springer, pp 209–237.
Hennessy DA, Wiesenthal DL. 2001. Further validation of the
driving vengeance questionnaire. Violence Vict 16:565–573.
Hirschi T, Gottfredson M. 1983. Age and the explanation of crime.
Am J Sociol 89:552–584.
Infante DA, Wigley III CJ. 1986. Verbal aggressiveness: An
interpersonal model and measure. Commun Monogr 53:63–69.
Kline RB. 2004. Beyond Significance Testing: Reforming Data
Analysis Methods in Behavioral Research. Washington, DC:
American Psychological Association.
Klump KL, Burt SA. 2006. The Michigan State University Twin
Registry (MSUTR): Genetic, environmental, and neurobiological
influences on behavior across development. Twin Res Hum
Genet 9:971–977.
Krueger RF, Markon KE, Patrick CJ, Benning SD, Kramer MD.
2007. Linking antisocial behavior, substance use, and personality:
An integrative quantitative model of the adult externalizing
spectrum. J Abnorm Psychol 116:645–666.
Lahey BB, Waldman ED. 2003. A developmental propensity model
of the origins of conduct problems during childhood and
adolescence. In: Lahey BB, Moffitt TE, Caspi A, (eds). Causes
of Conduct Disorder and Juvenile Delinquency. New York:
Guilford Press, pp 76–117.
Lee K, Ashton MC. 2007. Factor analysis in personality research. In:
Robins RW, Fraley RC, Krueger RF, (eds). Handbook of
Research Methods in Personality Psychology. New York, NY:
Guilford, pp 424–443.
Little TD, Cunningham WA, Shahar G, Widaman KF. 2002. To
parcel or not to parcel: Exploring the question, weighing the
merits. Struct Equation Model 9:151–173.
Loeber R, Schmaling KB. 1985. Empirical evidence for overt and
covert patterns of antisocial conduct problems: A meta-analysis.
J Abnorm Child Psychol 13:337–352.
Loeber R, Stouthamer-Loeber M. 1998. Development of juvenile
aggression and violence. Am Psychol 53:242–259.
Lorenzo-Seva U, Ferrando PJ. 2006. FACTOR: A computer
program to fit the exploratory factor analysis model. Behav
Res Methods Instrum Comput 38:88–91.
Loudin JL, Loukas A, Robinson S. 2003. Relational aggression in
college students: Examining the roles of social anxiety and
empathy. Aggr Behav 29:430–439.
McNeilly-Choque M, Hart CH, Robinson CC, Nelson LJ, Olsen SF.
1996. Overt and relational aggression on the playground:
Correspondence among different informants. J Res Child Educ
11:47–67.
Moffitt TE. 1993. Adolescence-limited and life-course-persistent
antisocial behavior: A developmental taxonomy. Psychol Rev
100:674–701.
Moffitt TE. 2003. Life-course persistent and adolescence-limited
antisocial behavior: A research review and a research agenda. In:
Lahey B, Moffitt TE, Caspi A, (eds). The Causes of Conduct
Disorder and Serious Juvenile Delinquency. New York: Guilford.
Murray-Close D, Han G, Cicchetti D, Crick NR, Rogosch FA. 2008.
Neuroendocrine regulation and physical and relational aggres-
sion: The moderating roles of child maltreatment and gender.
Dev Psychol 44:1160–1176.
Offord DR, Bennett KJ. 1994. Conduct disorder: Long-term
outcomes and intervention effectiveness. J Am Acad Child
Adolesc Psychiatry 33:1069–1078.
Osterman K, Bjorkqvist K, Lagerspetz K, Kaukiainen A,
Landau SF, Fraczek A, et al. 1998. Cross-cultural evidence of
female indirect aggression. Aggr Behav 24:1–8.
Pardini DA, Lochman JE, Frick PJ. 2003. Callous/unemotional
traits and social cognitive processes in adjudicated youth. J Am
Acad Child Adolesc Psychiatry 42:364–371.
Patrick CJ, Curtin JJ, Tellegen A. 2002. Development and validation
of a brief form of the Multidimensional Personality Question-
naire. Psychol Assess 14:150–163.
Paulhus DL, Williams KM. 2002. The dark triad of personality:
Narcissism, Machiavellianism, and psychopathy. J Res Pers
36:556–563.
Raine A, Dodge K, Loeber R, Gatzke-Kopp L, Lynam D,
Reynolds C. 2006. The Reactive-Proactive Aggression (RPQ)
questionnaire: Differential correlates of reactive and proactive
aggression in adolescent boys. Aggr Behav 32:159–171.
Richardson DS, Green LR. 2003. Defining direct and indirect
aggression: The Richardson Conflict Response questionnaire.
Revue Internationale De Psychologie Sociale 16:11–30.
Russell DW. 2002. In search of underlying dimensions: The use (and
abuse) of factor analysis in Personality and Social Psychology
Bulletin. Pers Soc Psychol Bull 28:1629–1646.
Stranger C, Achenbach TA, Verhulst FC. 1997. Accelerated long-
itudinal comparisons of aggressive versus delinquent syndromes.
Dev Psychopathol 9:43–58.
Tackett JL, Krueger RF, Sawyer MG, Graetz BW. 2003. Subfactors
of DSM-IV conduct disorder: Evidence and connections with
syndromes from the Child Behavior Checklist. J Abnorm Child
Psychol 31:647–654.
Tackett JL, Krueger RF, Iacono WG, McGue M. 2005. Symptom-
based subfactors of DSM-defined conduct disorder: Evidence for
etiologic distinctions. J Abnorm Psychol 114:483–487.
Torgesen JK, Wagner RK, Rashotte CA. 1999. Test of Word
Reading Efficiency. Austin: PRO-ED.
Tremblay RE. 2003. Why socialization fails: The case of chronic
physical aggression. In: Lahey B, Moffitt TE, Caspi A, (eds). The
Causes of Conduct Disorder and Serious Juvenile Delinquency.
New York: Guilford.
Vaillancourt T, Brendgen M, Boivin M, Tremblay R. 2003.
A longitudinal confirmatory factor analysis of indirect and
physical aggression: Evidence of two factors over time? Child
Dev 74:1628–1638.
Walden SB, McGue M, Iacono WG, Burt SA, Elkins I. 2004.
Identifying shared environmental contributions to early sub-
stance use: The importance of peers and parents. J Abnorm
Psychol 113:440–450.
White HR, Bates ME, Buyske S. 2001. Adolescence-limited versus
persistent delinqeuncy: Extending Moffitt’s hypothesis into
adulthood. J Abnorm Psychol 110:600–609.
Xie H, Cairns BD, Cairns RD. 2005. The development of aggressive
behaviors among girls: Measurement issues, social functions, and
differential trajectories. In: Pepler DJ, Madsen KC, Webster C,
Levence KS, (eds). The Development and Treatment of Girlhood
Aggression. Mahwah: Lawrence Erlbaum Associates Publishers,
pp 105–136.
398 Burt and Donnellan
Aggr. Behav.