Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

AGGRESSIVE BEHAVIOR

Volume 36, pages 338–349 (2010)

Role of Executive Dysfunction in Predicting Frequency


and Severity of Violence
Megan Hancock, Jennifer L. Tapscott, and Peter N.S. Hoaken

The University of Western Ontario, London, Ontario, Canada

: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
The adverse consequences of violence on society are tremendous. The proportion of offenders incarcerated for violent offenses is
large, and the cost of keeping these offenders incarcerated is startling. Understanding and treating the causal underpinnings of
violent crime is of utmost importance for individuals and society as a whole. Several factors have been identified as potential
contributors to violent crime, including cognitive deficits in executive functioning [Hoaken et al., 2007]. To investigate this further,
77 offenders from Fenbrook Institution, a federal facility, were tested on a battery of executive functioning measures. Offenders
were found to have broad and pervasive dysfunction in their executive abilities. In addition, specific scores from the battery were
found using regression techniques, to predict the frequency and severity of past violent offending but not nonviolent offending. This
speaks of the possibility of a new type of correctional rehabilitation program, one that focuses on the rehabilitation of basic
executive functions. Aggr. Behav. 36:338–349, 2010. r 2010 Wiley-Liss, Inc.

: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :

Keywords: executive functioning; violence; Delis–Kaplan Executive Function System; Poisson regression; binary logistic regression

Violence has profound social, legal, and political [for a review, see Burt, 2009]. Clearly, the con-
costs. The victims of violent crime can lose valuable tributors to violence are myriad.
possessions, their health, ability to work, and A more recent approach to better understanding
sometimes their lives [World Health Organization, violent behavior has involved the assessment of
2002]. In the United States in 2008, 4,856,510 violent specific aspects of cognition. Rather than focusing
offenses were committed [Bureau of Justice Statistics, on environmental variables, some researchers have
2009]. Violence results in significant financial con- begun to examine individual cognitive abilities to see
sequences for society; so, researchers urgently need whether there is a pattern of functioning character-
to understand factors predisposing individuals to istic of those who manifest aggressive or violent
use violence, both in order to develop preventative tendencies. The rationale for evaluating the relation-
strategies for individuals who are at risk and also to ship between cognitive variables and violent beha-
develop rehabilitation programs for individuals who vior comes from research examining brain
are already violent. anomalies in aggressive and violent individuals
Historically, researchers have worked at identify- [Marsh and Martinovich, 2006]. For example,
ing psychosocial risk factors in childhood that are research has repeatedly shown that acquired brain
related to an increased likelihood of aggression and injury is much more prevalent in incarcerated
violence later in life. Several reviews indicate that the populations than in the general public. The
predictors of aggression in adolescence are diverse incidence of mild brain injury in the general public
and include the characteristics of adolescents’ family
(e.g. parental attitudes toward aggression), school Grant sponsor: Social Sciences and Humanities Research Council of
experience (e.g. strict and inflexible classroom rules), Canada; Grant number: 410-2008-2461.
peer groups (e.g. poor peer relations), and commu- Correspondence to: Megan Hancock, Department of Psychology,
nity [e.g. neighborhoods low in socioeconomic Westminster Hall, The University of Western Ontario, London,
status; Flannery et al., 2005; Kashani et al., 1999]. ON N6A 3K7, Canada. E-mail: mhancoc6@uwo.ca
Furthermore, research has demonstrated that genetic Received 18 December 2009; Accepted 4 June 2010
factors are a major risk factor for violent behavior Published online 30 June 2010 in Wiley Online Library (wileyonline
when compared with nonviolent antisocial behavior library.com). DOI: 10.1002/ab.20353

r 2010 Wiley-Liss, Inc.


Predicting Frequency and Severity of Violence 339

of North America varies from 0.1 to 0.6%, been implicated in both executive abilities and
depending on the data collection method [Cassidy violence. Given the profound effects that violent
et al., 2004]; conversely, much higher rates, ranging behavior has on society, it is of interest whether
from 29 to 100%, have been reported in various there is a relationship between executive functioning
violent forensic samples [Marsh and Martinovich, and violent behavior and, more specifically, what
2006; Martell, 1992; Slaughter et al., 2003]. This aspects of executive function predispose individuals
research suggests that diminished functional brain to these behaviors. That information could prove
capacity may be implicated in aggressive and violent very important to theory-based efforts at prediction
behavior. and treatment.
A specific area of the brain that has received a Years of research from a broad range of fields
great deal of attention in the forensic literature is the certainly suggests that impaired executive functioning
frontal cortex. Evidence implicating this region correlates with aggressive and violent behavior
comes from a series of studies finding structural [Broomhall, 2005; Hoaken et al., 2003; Marsh and
and functional abnormalities in the frontal lobes of Martinovich, 2006; Raaijmakers et al., 2008;
individuals who are aggressive and violent [for Séguin et al., 2004]. However, although a number
reviews, see Brower and Price, 2001; Bufkin and of studies have examined the relationship between
Luttrell, 2005]. Although aggression and violence executive functioning and criminality, it has not yet
cannot be attributed exclusively to brain dysfunc- been consistently demonstrated which particular
tion, it is important to integrate this information components of executive functioning are impaired
into an understanding of the variables that are in offenders. Furthermore, some researchers argue
related to an increased likelihood of aggressive and that executive deficits are characteristic of all
violent behavior. individuals who commit crime [Greenfield and
A number of researchers have suggested that the Valliant, 2007; Hoaken et al., 2007], whereas others
relationship between these brain anomalies and argue that these deficits are more characteristic
violent behavior may be mediated by an inability of those who commit violent crime specifically
to adaptively utilize what have been referred to as [Baker and Ireland, 2007; Barker et al., 2007;
‘‘executive functions’’ [Bufkin and Luttrell, 2005; Miura, 2009].
Marsh and Martinovich, 2006]. A large number of studies have examined the
Although executive functioning is frequently relationship between executive functioning and
discussed in the literature, an agreed-upon definition criminality, antisocial behavior, aggression, and
has not yet been established [Suchy, 2009]. One violence; yet, each of these studies used a slightly
point of agreement is that executive abilities allow different conceptual definition of executive functioning.
individuals to shift mindsets quickly in a constantly Moreover, a number of different measures were used
changing environment (i.e. adapt to novel and to assess executive abilities. However, not all these
diverse situations while simultaneously inhibiting measures are thought to assess the same abilities or
inappropriate behaviors). Executive abilities are share the same psychometric properties. To compli-
essential for success in school, in work, and in cate matters more, some researchers evaluate
everyday living as they mediate the ability to aggression and violence using self-report and others
organize thoughts in a goal-directed way. examine criminal histories or directly observe
Early executive functioning research was stimu- behavior. Some researchers only consider an offen-
lated by neuropsychological patients with frontal ders’ index offense, whereas others take into account
lobe damage. Although able to perform well on their complete criminal history. All being said, the
other cognitive tasks, including tests of intelligence overall theme within the literature is that executive
[Shallice and Burgess, 1991], these patients exhibited functioning, however loosely defined, is related to
difficulty controlling or regulating their behavior criminality. For example, in a meta-analysis of 39
and displayed deficits on tests of organization, studies, individuals who took part in antisocial
set-shifting, and goal-directed behavior. This early behaviors performed .62 standard deviations worse
research led neuropsychologists to suspect that the on measures of executive abilities than individuals
frontal lobes of the brain may be largely responsible who did not, despite controlling for age, sex,
for executive functioning. Although that view is now ethnicity, and intelligence [Morgan and Lilienfeld,
thought to be oversimplistic with a contemporary 2000].
understanding of some limited limbic involvement A recent development to this line of investigation
[Jurado and Rosselli, 2007; Stuss and Alexander, has been the publication of the Delis–Kaplan
2000], the point remains that the frontal lobes have Executive Function System [D-KEFS; Delis et al.,

Aggr. Behav.
340 Hancock et al.

2001], a valid and normed tool that measures METHOD


several aspects of executive function. Broomhall
Participants
[2005] was the first in the forensic field to make use
of this tool, assessing the capacity of violent The sample for this study consisted of 80 adult
offenders on three D-KEFS subtests. Broomhall male offenders from Fenbrook Institution, a
found that violent offenders were significantly medium security federal prison in Gravenhurst,
impaired on these tests, relative to the general Ontario. Participants were assessed on two inclusion
public, but noted that ‘‘the results did not provide criteria: proficiency in English and normal or
the depth of understanding required to inform corrected-to-normal vision. Only 3 of the 80
adequate prediction of future dangerousness, participants were excluded for failing to meet these
assessment, or treatment of individuals who criteria. Approximately 65% of the offenders
commit violent crime’’ (p 379). recruited had served a sentence for a violence-
This review has established that incarcerated related crime. The offenders ranged in age from
offenders are characterized by executive dysfunction. 19 to 57 (M 5 33.35, SD 5 9.19).
Although some evidence suggests that dysfunction is
more characteristic of individuals who are aggres-
Materials
sive, more research is needed. In addition, the
executive deficits that have been identified are rather Kaufman Brief Intelligence Test, Second Edition
vague, and it is not well understood how these (KBIT-2): Intelligence was measured using the
deficits translate into everyday behavior. The goals KBIT-2 which can be administered in approximately
of this study were: (A) to examine a sample of 20–30 min and provides Verbal and Nonverbal
offenders to see whether deficits in executive Scores, as well as a composite IQ score [Kaufman
functioning were identified, using a valid and and Kaufman, 2004]. The KBIT-2 was administered
normed measure (the D-KEFS), (B) to examine to ensure that any relationships that were detected
whether these deficits would be correlated with the between executive functioning and offending were
frequency of offending in general, with the fre- not solely a function of differences in intelligence.
quency of nonviolent offending or with the Delis– Kaplan Executive Function System (D-KEFS):
frequency of violent offending, and (C) to assess The D-KEFS is a battery of nine tests that
whether scores were able to predict crime in general, comprehensively assesses the key components of
nonviolent offending, violent offending, or serious executive functioning. In light of the available data
violent offending. Violence was assessed using a on the psychometric properties of the D-KEFS, this
continuous scale, which took into account the battery is considered among the most valid means
frequency and severity of violent acts. In compar- of assessing executive functioning [Baron, 2004;
ison with earlier research in which offenders are Homack et al., 2004]. Owing to time constraints,
often dichotomized as violent or nonviolent, in this only four of the subtests could be administered
study we examined and recorded each offender’s [i.e. the Tower Test, Verbal Fluency Test, Color-
entire adult criminal history, thereby avoiding Word Interference Test (CWIT), and Sorting Test;
categorization all together. Violence severity was see Shunk et al., 2006, for a detailed description of
also considered because, though an offender with a the subtests], but this decision was not thought to
murder charge and an individual with an assault affect the psychometric properties of the subtests,
charge are both ‘‘violent offenders,’’ these indivi- given that the D-KEFS subtests were designed to
duals may be quite dissimilar. ‘‘stand alone.’’
This study was primarily exploratory in nature; During the Tower Test, participants assembled a
therefore, no hypotheses were made about which number of towers by moving disks of different sizes
specific executive functioning scores would be across three vertical pegs. The Tower Test assesses
impaired or about which scores would predict executive abilities, including planning, rule learning,
offending. However, it was hypothesized that inhibition of impulsive and perseverative responding,
offenders would evidence a significant pattern of and the ability to establish and maintain the
executive dysfunction and that this dysfunction instructional set [Delis et al., 2001].
would be more characteristic of those who commit The Verbal Fluency Test consists of three condi-
violent offenses. Furthermore, it was hypothesized tions: Letter Fluency, Category Fluency, and
that an increased frequency and an increased Category Switching. This test measures participants’
severity of violence would each be related to greater ability to generate words fluently in an effortful,
executive dysfunction. phonemic format (Letter Fluency), from overlearned

Aggr. Behav.
Predicting Frequency and Severity of Violence 341

concepts (Category Fluency), and while simulta- TABLE I. List of Abbreviations From the D-KEFS
neously shifting between overlearned concepts
Abbreviation Variable name
[Category Switching; Delis et al., 2001].
The CWIT is similar to the classic Stroop Test, but CWITI-CN Color-Word Interference Test—Inhibition Condition
it consists of four conditions. The first two condi- minus Color Naming. Scaled Score
tions provide a baseline measure of the two basic CWITICE Color-Word Interference Test—Inhibition Condition
Corrected Errors. Cumulative Percentile Rank
skills that are required to complete the higher-level CWITIUE Color-Word Interference Test—Inhibition Condition
tasks: naming of color patches and reading of color Uncorrected Errors. Cumulative Percentile Rank
words. The third condition assesses inhibition, as CWITI/S-CN Color-Word Interference Test—Inhibition/Switching
participants must inhibit reading the words in order Condition minus Color Naming. Scaled Score
to name the dissonant ink colors in which those CWITI/S-R Color-Word Interference Test—Inhibition/Switching
Condition minus Reading. Scaled Score
words are printed [Delis et al., 2001]. The fourth CWITI/SCE Color-Word Interference Test—Inhibition/Switching
condition measures both inhibition and cognitive Condition Corrected Errors. Cumulative Percentile
flexibility [Delis et al., 2001]. Rank
Finally, the Sorting Test consists of two condi- CWITI/SUE Color-Word Interference Test—Inhibition/Switching
tions—a Free Sorting condition and Sort Recogni- Condition Uncorrected Errors. Cumulative
Percentile Rank
tion condition. The Sorting Test assesses a number STCCS Sorting Test—Confirmed Correct Sorts. Scaled Score
of important component processes of executive STFSDS Sorting Test—Free Sorting Description Score. Scaled
functioning, including (a) initiation, (b) concept Score
formation skills, (c) modality-specific problem- STSRDS Sorting Test—Sort Recognition Description Score.
solving skills (verbal vs. perceptual), (d) the ability Scaled Score
STPR Sorting Test—Perceptual Rules. Scaled Score
to explain the sorting concepts abstractly, (e) the STVR Sorting Test—Verbal Rules. Scaled Score
ability to transfer sorting concepts into action, STOAD Sorting Test—Overly Abstract Descriptions.
(f) the ability to inhibit previous sorting responses Cumulative Percentile Rank
in favor of flexibility of behavior, and (g) the ability STPSA Sorting Test—Percent Sorting Accuracy. Scaled Score
to inhibit previous description responses in favor of TTMAR Tower Test—Move Accuracy Ratio. Scaled Score
TTMFMT Tower Test—Mean First Move Time. Scaled Score
flexibility of thinking [Delis et al., 2001]. TTTAS Tower Test—Total Achievement Score. Scaled Score
The D-KEFS was normed on a sample of more VFLF-TC Verbal Fluency—Letter Fluency—Total Correct.
than 2,000 individuals. Consequently, most of Scaled Score
the D-KEFS raw scores were transformed into VFCF-TC Verbal Fluency—Category Fluency—Total Correct.
age-adjusted standard scores, ranging from 1 to 20 Scaled Score
VFCS-TC Verbal Fluency—Category Switching—Total Correct.
(M 5 10, SD 5 3). However, because some raw Scaled Score
scores were restricted in range in the normative VFCS-SA Verbal Fluency—Category Switching—Switching
sample, these raw scores were instead transformed Accuracy. Scaled Score
into age-adjusted cumulative percentile ranks.
Unless otherwise noted, a higher score reflects
counterbalanced, but complete counterbalancing
superior performance, and those scores that were
was not possible owing to the number of tests
converted to cumulative percentile ranks will be
administered.
identified where appropriate. See Table I for a list of
As prescribed by the Correctional Service of
abbreviations for the D-KEFS scores.
Canada, offenders were not eligible to receive
compensation for participation. The overall time
Procedure to complete the experiment for the offenders was
approximately 2 hr. To prevent experimenter bias,
Offenders were initially recruited during an
no questions regarding current or past criminal
information session held within the prison, but
activities were asked during the testing session.
additional offenders were recruited when they
approached the researchers and expressed interest
in participating. The researchers made it explicit that
File Review
participation was voluntary and that the results
would have no bearing on any subsequent correc- Only after an offender had completed the test
tional decisions. After obtaining informed consent, battery was his file reviewed. Two of the authors
the researchers explained that no feedback would (M.H. and J.L.T.) collaboratively reviewed offen-
be provided. The order of administration of the four ders’ complete adult criminal history through the
D-KEFS subtests and the KBIT-2 was partially Offender Management System, the computerized

Aggr. Behav.
342 Hancock et al.

case file management system that is used by the of the offenders in the sample had committed more
Correctional Service of Canada and the National than one severe violent offense. Therefore, this
Parole Board. The first 20 files were recoded to dependent variable was dichotomous. Consequently,
ensure that there had been no coder drift. binary logistic analyses were conducted to examine
The term violence has various proposed defini- predictors of severe violent offending.
tions, but this study defined violence as behavior An explanation of how regression coefficients
involving an intentional act of physical aggression were interpreted needs to be provided before
against another individual that is likely to cause presenting the results. Briefly, because Poisson
physical injury [Meloy, 2006]. All violent offenses regression is ‘‘linear in the logarithm’’ [Coxe et al.,
were coded according to Cornell’s [1996] guidelines. 2009], when all other variables are held constant, a
That is, severity of violence was coded along a 1-unit increase in a predictor results in an increase of
6-point scale that had the following anchors: 1 (assault the natural logarithm of the predicted count that is
without injury), 2 (minor injury; e.g. bruises, minor equal to the value of the unstandardized regression
medical treatment, attempted rape), 3 (serious injury coefficient (b). This interpretation is straightfor-
requiring substantial hospital treatment; e.g. broken ward, but has the disadvantage of interpreting the
limb, rape, gunshot wound), 4 (severe injury resulting change in the unit of a transformation of the
in lasting impairment or life-threatening injury), 5 outcome (i.e. the natural logarithm of the predicted
(homicide), and finally 6 (extreme homicide; e.g. count). As such, for the Poisson models, the
multiple killings, mutilation). It should be noted that exponentiated unstandardized regression coefficients
a slight modification to Cornell’s coding scheme was (i.e. eb) are presented in the tables, which can
made such that, in accordance with our definition of be interpreted as incidence rate ratios. That is, for a
violence, threats of violence were coded as nonviolent 1-unit change in the predictor, the predicted rate is
offenses. Dependent variables were then generated multiplied by eb. Similarly, for the binary logistic
from these ratings. On the basis of the file review, the model, exponentiating the regression coefficients
frequency of nonviolent, violent, severe violent, and produces an odds ratio. Odds ratios are interpreted
total offending was determined. Severe violent as the change in the likelihood of an offender having
offenses included the subset of the violent offenses committed a severe violent offense given a 1-unit
that produced serious, lasting physical harm (i.e. change in the predictor.
severity of 4, 5, or 6). Poisson regression assumes that the period of risk,
also known as exposure, is the same for all
observations. This assumption was violated in this
RESULTS study, considering that offenders were of different
ages and, consequently, had been at risk of
Analytic Rationale
committing adult offenses for different periods of
This study investigated frequency of offending time. To control for nonuniform exposure times, the
during adulthood. Therefore, the dependent variables natural logorithm of years at risk of committing an
of interest are count variables. Count data present a adult offense was entered as a covariate in each of
challenge to researchers in the correctional and the regression analyses with its parameter fixed at
forensic fields; however, these challenges can be 1.00. As a result of the inclusion of this variable, the
managed with statistical techniques designed specifi- outcome variable predicted in each analysis was the
cally for this type of data [for reviews, see Hutchinson rate of log-offenses per unit of exposure (i.e. per year
and Holtman, 2005; Walters, 2007]. The benchmark of adult life) instead of simply the frequency of
model for fitting count data is the Poisson distribu- offenses.2
tion and the standard regression model for analyzing
count data is the Poisson regression.1
2
Another factor that needs to be considered when deciding whether
Although the initial intent was to use Poisson or not to use Poisson-class regression is whether the number of zero
counts exceeds what the Poisson model can handle [Walters, 2007].
regression to predict frequency of nonviolent offending, Zero-inflated models for Poisson regressions allow for an excessive
violent offending, and severe violent offending, none number of observed zeroes [Ridout et al., 1998]. A zero-inflated
Poisson regression was conducted to test whether it provided a better
1
Psychology data are often overdispersed and, therefore, better fit than the standard model. The Vuong test compares the zero-
analyzed by a negative binomial model. An estimate of the dispersion inflated model to the standard model to determine which model is a
parameter can be calculated through Stata which indicates whether a better fit. The z-values were not significant when the outcome
Poisson or negative binomial model is more appropriate. This predicted was frequency of violent offending meaning that the zero-
statistic indicated that a Poisson model was most appropriate for the inflated Poisson regression is not a better fit than the standard
present data. Poisson regression.

Aggr. Behav.
Predicting Frequency and Severity of Violence 343

Preliminary Analyses [FreqVO], severe violent offending [SevVO],


frequency of nonviolent offending [FreqNVO], and
One offender was missing data for one subtest
the total number of offenses [TotalOFF]) was
owing to his colorblindness. In order to retain as
computed (see Table III). This was a crude estima-
much meaningful data as possible, group means were
tion because the assumptions underlying correlation
utilized to fill the missing data points. See Table II for
analyses are similar to those underlying linear regres-
the mean and standard deviations of scores from the
sion and the present data violated those assumptions.
DKEFS and from the outcome variables. It should be
As such, it was concluded that an inclusive approach
noted that although 10% of the sample of offenders in
for entry into the regression model would be appro-
this study had committed an act of homicide, more
priate to ensure that all possible relationships between
than one-quarter of the incarcerated Canadian federal
the predictors and outcome variables were fully
population has a history of homicide [Motiuk and
explored. Therefore, predictor variables that were
Vuong, 2005]. As such, the results of this study may
significantly related to the outcome variables with a
fail to generalize to offenders who have engaged in
set at .10 were included in the regression equations. See
more severe forms of violence than was committed by
Table III for the correlations between each of the D-
the current sample.
KEFS scores and the covariates to be included in the
In order to determine which variables needed to be
regression analyses (i.e. age and the KBIT composite
entered into the regression analyses, the correlation
score).
between each of the predictor variables and the four
outcome variables (i.e. frequency of violent offending
Regression Analyses

TABLE II. Means, Ranges, and Standard Deviations of Poisson regression analyses were performed to
D-KEFS Scores (Tower Test, Verbal Fluency, Color-Word determine the relationship between D-KEFS scores
Interference, and Sorting Test), and Outcome Variables identified in the preliminary analyses and frequency
(Frequency of Violent Offending [FreqVO], Severe Violent of violent offending. Table IV shows the likelihood
Offending [SevVO], Frequency of Nonviolent Offending
[FreqNVO], and Frequency of Total Offending [TotalOff])
ratio chi-square (LR w2), z values, incident rate ratios,
and significance level for each predictor variable.
Variable Mean (range) SD Owing to the exploratory nature of the research, each
score was included in a separate regression analysis
TTTAS 10.10 (3–15) 2.14
TTMFMT 10.87 (4–14) 2.17 rather than including all scores in a single model. This
TTMAR 8.83 (1–14) 2.57 decision allowed us to identify all variables that were
VFLF-TC 9.17 (2–17) 3.43 individually related to violent offending rather than
VFCF-TC 10.38 (2–19) 3.20 identifying the variables that had unique effects. In
VFCS-TC 9.12 (2–15) 3.07
these analyses, age, intelligence, and frequency of
VFCS-SA 8.91 (1–15) 3.14
CWITI-CN 9.51 (1–19) 3.26 nonviolent offending served as covariates and time at
CWITcICE 38.96 (1–100) 41.96 risk was entered as an offset variable. Three of the
CWITcIUE 61.06 (1–100) 42.07 predictors were statistically significant in the predic-
CWITI/S-CN 9.27 (3–18) 3.11 tion of rate of violent offending. The exponentiation
CWITI/S-R 9.17 (1–15) 3.12
of the regression coefficient for STPR, e .11 5 .90, was
CWITcI/SCE 37.77 (1–100) 40.42
CWITcI/SUE 51.44 (1–100) 38.69 the predicted multiplicative effect of a 1-unit change
STCCS 7.79 (1–14) 3.01 in STPR on the number of violent offenses committed in
STFSDS 7.29 (1–13) 2.86 1 year. In other words, an offender with a STPR score of
STSRDS 7.10 (1–16) 3.09 7 was expected to have a rate of violent offending that
STPSA 10.32 (2–13) 2.57
was 0.90 times the rate of violent offending of an
STcOAD 53.40 (1–100) 42.20
STPR 6.97 (1–17) 3.61 offender with a scaled score of 6 (i.e. he was likely to
STVR 7.05 (1–14) 2.96 commit violent offenses less frequently). The expected
FreqVO 1.29 (0–7) 1.38 rate change in violent offending for a 1-unit change in
FreqNVO 10.35 (0–86) 12.05 CWITI/SCE was 1.01 and TTMFTM was .92.
SevVO 1.00 (0–1) 0.32
Binary logistic regression analyses were used to
TotalOff 11.64 (1–87) 12.27
predict whether an offender had committed a severe
N 5 77. See Table I for the list of abbreviations.
c
violent offense from scores from the D-KEFS.
indicates a cumulative percentile rank, which is out of 100. All other Table V shows the Wald w2, z-values, odds ratios,
D-KEFS scores are scaled scores, which have a mean of 10, a
standard deviation of 3, and a maximum of 20. The outcome and significance level for each predictor variable. As
variables are counts of the offense frequency. an example of how to interpret the coefficients the

Aggr. Behav.
344 Hancock et al.

TABLE III. Correlations Among the D-KEFS Scores (Tower Test, Verbal Fluency, Color-Word Interference and Sorting Test),
Age, KBIT Composite Score, and the Outcome Variables (Frequency of Violent Offending [FreqVO], Severe Violent Offending
[SevVO], Frequency of Nonviolent Offending [FreqNVO], and Frequency of Total Offending [TotalOff])

Variable Age KBIT composite FreqVO SevVO FreqNVO TotalOff

1. TTTAS .04 .20 .21 .17 .08 .14


2. TTMFMT .07 .15 .23 .17 .06 .10
3. TTMAR .01 .00 .14 .21 .07 .08
4. VFLF-TC .04 .38 .15 .02 .07 .09
5. VFCF-TC .02 .32 .06 .07 .01 .01
6. VFCS-TC .01 .29 .16 .21 .03 .05
7. VFCS-SA .00 .37 .05 .17 .08 .09
8. CWITI-CN .04 .09 .11 .03 .00 .06
9. CWITICE .30 .16 .10 .02 .15 .16
10.CWITIUE .02 .32 .03 .09 .06 .06
11. CWITI/S-CN .06 .12 .04 .28 .05 .05
12. CWITI/S-R .03 .18 .02 .15 .12 .12
13. CWITI/SCE .18 .14 .26 .08 .03 .00
14. CWITI/SUE .05 .34 .05 .18 .11 .10
15. STCCS .08 .51 .10 .06 .02 .01
16. STFSDS .13 .56 .13 .07 .05 .04
17. STSRDS .15 .67 .15 .13 .05 .03
18. STPSA .01 .04 .12 .30 .05 .04
19. STOAD .23 .29 .02 .32 .05 .06
20. STPR .14 .64 .21 .10 .02 .00
21. STVR .05 .49 .03 .10 .03 .03

N 5 77. See Table I for the list of abbreviations.


Po.05; Po.10.

exponentiation of the regression coefficient for occurrences correctly predicted, was 11% and the
CWITI/S-CN, e .51 5 0.60, was the predicted multi- specificity, or percentage of nonoccurrences cor-
plicative effect of a 1-unit change in CWITI/S-CN on rectly predicted, was 100%.
the likelihood of having committed a severe violent Because sensitivity was low, a binary logistic
offense (vs. not committing a severe violent offense). regression was run containing all the significant
In other words, four scores from the subtests of the predictors identified by individual regressions.
D-KEFS were each related to whether an offender Table VI displays the z-values, odds ratios, and
had committed a severe violent offense or not, over significance level for each predictor variable within
and above age, intelligence, and frequency of this regression. None of the scores were able to
nonviolent offending.3 predict whether an offender had committed a severe
Although it is important to interpret the odds violent offense over and above the ability of the
ratios, it was equally important to examine how other scores, age, intelligence, and frequency of
accurate the regression models were at correctly nonviolent offending. However, more importantly,
identifying offenders, as those who had committed it was found that the regression model’s accuracy of
severe violent offenses and those who had not. It correctly classifying offenders improved when all the
was found that for each individual binary logistic predictors were input simultaneously. Using this
regression 90% offenders were correctly classified. model, sensitivity improved to 88.89% and specifi-
One of the 9 severe violent offenders was correctly city remained high at 100%. These predicted values
classified and all 68 offenders who had not are significantly higher than those calculated by
committed a severe violent offense were correctly chance alone (w2 5 58.84, Po.01).
classified. Therefore, the sensitivity, or percentage of

3
We acknowledge that we have made no mention of power or effect DISCUSSION
size throughout the discussion of the regression analyses; however,
this was done intentionally. The regression analyses used in this Central Findings
study are not completely developed in terms of the traditional types
of testing that are usually available. As there is no established way of Offenders scored below the mean of the normative
determining effect size or power in these types of regression analyses, sample on many, but not all, of the scores from the
we chose not to make mention of these concepts. D-KEFS tests, indicating a relative weakness in

Aggr. Behav.
Predicting Frequency and Severity of Violence 345

TABLE IV. Poisson Regression Models for the Prediction TABLE V. Binary Logistic Regression Models for the
of Frequency of Violent Offending From Individual D-KEFS Prediction of Severity of Violent Offending From Individual
Scores Covarying Age, Intelligence, and the Frequency of D-KEFS Scores Covarying Age, Intelligence, and the
Nonviolent Offending (FreqNVO) Frequency of Nonviolent Offending (FreqNVO)

Model IRR (95% CI) z LR w2(4) Model OR (95% CI) z Wald w2(4)

Predicted outcome: rate FreqVO Predicted outcome: rate SevVO 8.74


CWITcI/SCE 1.01 (1.00–1.02) 2.05 11.30 CWITI/S-CN 0.60 (0.38–0.93) 2.27
Age 0.97 (0.95–.99) 2.34 Age 1.00 (0.91–1.10) 0.09
IQ 0.99 (0.98–1.01) 1.00 IQ 1.02 (0.95–1.10) 0.63
FreqNVO 1.00 (0.98–1.02) 0.18 FreqNVO 0.40 (0.19–0.86) 2.35
Predicted outcome: rate FreqVO Predicted outcome: rate SevVO 9.14
STPR 0.90 (0.83–0.97) 2.82 15.66 STPAS 0.60 (0.39–0.95) 2.18
Age 0.97 (0.95–1.00) 1.79 Age 0.97 (0.88–1.06) 0.66
IQ 1.02 (0.99–1.04) 1.35 IQ 0.98 (0.91–1.05) 0.59
FreqNVO 1.00 (0.98–1.01) 0.54 FreqNVO 0.49 (0.28–0.86) 2.48
Predicted outcome: rate FreqVO Predicted outcome: rate SevVO 11.71
TTMFMT 0.92 (0.85–1.00) 2.04 11.19 STcOAD 1.03 (1.01–1.06) 2.38
Age 0.97 (0.95–0.99) 2.23 Age 0.95 (0.84–1.06) 0.37
IQ 1.00 (0.98–1.01) 0.30 IQ 0.97 (0.92–1.06) 0.32
FreqNVO 1.00 (0.98–1.01) 0.56 FreqNVO 0.56 (0.36–0.88) 2.51
Predicted outcome: rate FreqVO Predicted outcome: rate SevVO 9.08
TTTAS 0.92 (0.83–1.01) 1.79 10.40 TTMAR 1.41 (0.92–2.16) 1.58
Age 0.98 (0.95–1.00) 2.44 Age 0.98 (0.90–1.07) 0.53
IQ 1.00 (0.98–1.01) 0.28 IQ 0.99 (0.93–1.06) 0.33
FreqNVO 0.99 (0.97–1.01) 0.75 FreqNVO 0.54 (0.34–0.86) 2.58
Predicted outcome: rate SevVO 10.93
The four covariates listed under the predicted outcome were included VFCS-TC 0.68 (0.47–0.98) 2.03
in the same model (i.e. only four models were estimated) N 5 77. Age 1.01 (0.83–0.99) 0.21
Age, intelligence, and frequency of nonviolent offending served as
IQ 1.03 (0.96–1.11) 0.88
covariates. Time at risk was entered as an offset variable.
c FreqNVO 0.56 (0.36–0.88) 2.52
indicates a cumulative percentile rank, which is out of 100. All other
scores are scaled scores, which have a mean of 10, a standard The four covariates listed under the predicted outcome were included
deviation of 3, and a maximum of 20. IRR, incidence rate ratio in the same model (i.e. only five models were estimated) N 5 77. Age,
(i.e. the exponentiated unstandardized regression coefficient, eb) intelligence, and frequency of nonviolent offending served as
LR, likelihood ratio; see Table I for the list of abbreviations. covariates. Time at risk was entered as an offset variable.
Po.05; Po.01.
c
indicates a cumulative percentile rank, which is out of 100. All other
scores are scaled scores which have a mean of 10, a standard
these areas compared with the general population. deviation of 3, and a maximum of 20. OR, odds ratio (i.e. the
exponentiated unstandardized regression coefficient, eb); see Table I
When examining the relationship between measures for the list of abbreviations.
of executive functioning and rates of offending, Po.05; Po.01.

it was found that, as hypothesized, scores from the


D-KEFS were related to the frequency and severity performed better on these measures. Similarly,
of violent offending. However, none of the scores offenders who performed poorly on a measure of
were related to either the frequency of nonviolent concept formation and cognitive flexibility (STPR)
offending or the frequency of total offending. These were more likely to commit an increasingly large
findings are consistent with research suggesting that number of violent offenses. Finally, the fewer
executive dysfunction is characteristic of offenders corrected errors an offender made on the switching
who commit violent crimes [Baker and Ireland, condition of the CWIT (CWITI/SCE), the more likely
2007; Barker et al., 2007; Miura, 2009], although they were to have an increased rate of violent
inconsistent with other research that has suggested offending. This latter finding was surprising because
that offenders who commit violent offenses and the offenders, as a group, committed a large number
those who commit nonviolent offenses do not differ of corrected errors. The current finding is idiosyn-
in their executive abilities [Greenfield and Valliant, cratic, and not easily explained by anything cur-
2007; Hoaken et al., 2007]. rently in the literature. This finding may be the result
Predictors of frequency of violence: Results of a Type I error; alternatively, it may indicate that
suggested that offenders who performed poorly on offenders who commit high rates of violent offenses
measures sensitive to impulsivity (TTMFMT) were become less likely to correct their errors on the
more likely to have committed an increasingly large CWIT because they lack the ability to monitor their
number of violent offenses than were offenders who behavior as they perform tasks.

Aggr. Behav.
346 Hancock et al.

TABLE VI. Binary Logistic Regression Model for the committed a severe violent offense were correctly
Prediction of Severity of Violent Offending From D-KEFS classified and that 100% of the offenders who had
Scores Covarying Age, Intelligence, and the Frequency not committed a severe offense were correctly
of Nonviolent Offending (FreqNVO)
classified. These results indicate that executive
Model OR (95% CI) z Wald w2(7) functioning may well be a construct that warrants
consideration for inclusion in future research on risk
Predicted outcome: rate SevVO 8.42
CWITI/S-CN 0.58 (0.30–1.14) 1.57
assessment, a possibility that is discussed in more
STPAS 0.66 (0.32–1.34) 1.16 detail below.
STcOAD 1.03 (0.99–1.06) 1.63 As mentioned earlier, research investigating
VFCS-TC 0.65 (0.30–1.40) 1.10 executive abilities in offenders tends to dichotomize
Age 0.99 (0.85–1.16) 0.09 offenders as violent or nonviolent. A major problem
FreqNVO 0.49 (0.20–1.18) 1.58
with studying offenders in this manner has been
The seven covariates listed under the predicted outcome were confusion over what criteria should be used to
included in the same model (i.e. only one model was estimated) identify violent offenders. Serin and Preston [2001]
N 5 77. Age, intelligence, and frequency of nonviolent offending
served as covariates. Time at risk was entered as an offset variable. reviewed a variety of the criteria used to classify
c
indicates a cumulative percentile rank, which is out of 100. All other offenders as violent or nonviolent, including index
scores are scaled scores which have a mean of 10, a standard offense, attitudes (e.g. hostility), emotions, (e.g.
deviation of 3, and a maximum of 20. OR, odds ratio (i.e. the
exponentiated unstandardized regression coefficient, eb); see Table I
anger), and victim selection (e.g. spousal assault).
for the list of abbreviations. Inconsistencies in the use of these criteria contribute
Po.05; Po.01.
to the ambiguity of findings. Our approach to
dealing with offense history was to avoid classifica-
Predictors of severity of violence: Results from the tion all together and to examine the frequency of
second set of regression analyses suggested that the violent offenses, whereas at the same time also
more impaired an offender was at inhibiting considering severity. This approach, we believe,
prepotent verbal responses (CWITI/S-CN) and in constitutes the strength of this study and should
concept formation (STPAS), the more likely they result in more generalizable conclusions.
were to have committed a severe violent offense. In summary, if an individual lacks intact executive
Similarly, offenders who scored lower on measures abilities, criminal activity—or more specifically,
of cognitive flexibility (VFCS-TC) were more likely to violent behavior—may be the default response for
have committed a severe violent offense. Unexpect- navigating through difficult situations (e.g. solving
edly, the more intact offenders’ expressive language financial strain, coping with a provocative inter-
abilities (as evidenced by a low number of abstract personal encounter). For example, an individual
sorting descriptions; STOAD), the more likely they who is not able to accurately examine the outcome
were to have committed a severe violent offense. of earlier decisions (owing to difficulties in altering
This latter finding was surprising because, as a their behavior in response to environmental
group, the offenders were impaired in expressive changes, and/or owing to difficulties with inhibition
language abilities. However, it seems that offenders and problem solving) will likely continue to make
who are less impaired are more likely to have poor decisions in the future.
committed a severe violent offense. Although
unexpected, an explanation for this finding may
Theoretical Considerations
involve the dynamic of escalation in violent
encounters. Many incidents of conflictual violence Despite the range of executive functioning deficits
are preceded by a verbal exchange (i.e. insults and associated with violence in this study, a few
provocations). Offenders with more intact expres- theoretical considerations need to be discussed given
sive language abilities may well incite the escalation the lack of agreement in the literature regarding the
of the conflict to the point of serious violence. measurement of executive functioning, the relation-
Also of interest was whether these significant ship between cognitive and social variables, and the
predictors were able to correctly classify offenders as conceptualization of violence as a unitary construct.
having committed a severe violent offense or as not First, executive dysfunction in violent offenders
having committed a severe violent offense. The seems to be clinically significant; however, the
individual ability of each predictor to correctly ecological validity of experimental tasks and
classify offenders was weak. When the significant traditional neuropsychological tests of executive
predictors were run in a regression together, it was functioning has been criticized [Chan et al., 2008].
found that 88.89% of the offenders who had That is, although conventional experimental tasks

Aggr. Behav.
Predicting Frequency and Severity of Violence 347

demand relatively simple responses to single events, adulthood. Given these features, executive function-
more complex multistep tasks in daily life may ing is a good candidate for the evaluation of
require a more complicated series of responses. potential inclusion in risk assessment.
Future research should attempt to include more Potential for rehabilitation: Beyond simply asses-
ecologically valid measures. sing individuals’ risk for future violence, researchers
Second, the heterogeneity of criminal behavior need to develop interventions to manage or reduce
makes it unlikely that all forms of crime share the this risk. Given the findings of this study, executive
same etiology. This study examined neuropsycholo- functioning is a construct that warrants further
gical dysfunction in isolation, without considering exploration as a potential target for improvement
the psychosocial predictors of crime. Scarpa and during the rehabilitation of violent offenders.
Raine [2007] reviewed what little is known on the Executive dysfunction is not specific to incarcerated
interaction between neurocognitive deficits and offenders, but rather characteristic of a number of
social variables in relation to antisocial and violent clinical populations, such as individuals with schizo-
behavior. These authors concluded that the combi- phrenia [Velligan and Bow-Thomas, 1999], atten-
nation of neurocognitive deficits and early adverse tion-deficit disorders [Pliszka, 2007], Tourette’s
life events significantly increased risk for committing disorder [Bornstein, 1990], and some traumatic brain
violent offenses over possessing only one set of risk injury [Krpan et al., 2007]. Research in cognitive
factors. Joint assessment of both psychosocial and rehabilitation suggests that with the right techniques,
cognitive contributors to violence is an approach executive dysfunction can be improved significantly
that warrants further exploration. in these populations [Rath et al., 2003; Worthington,
Third, owing to the exploratory nature of the 2005]. Given these findings, perhaps similar strategies
research, violence was conceptualized as homoge- could be applied to developing rehabilitation pro-
nous. However, other researchers have distinguished grams for offenders identified as having deficits in
between instrumental violence, used to obtain some executive functioning, and such treatment may teach
identifiable goal, such as a monetary gain, and offenders an alternative to violence.
reactive violence, used as a defense after a perceived
threat [Bushman and Anderson, 2001]. In fact,
Limitations
Broomhall [2005] found preliminary evidence to
suggest that instrumental and reactive aggressors As with all research, the findings of this study
may be characterized by different executive deficits. must be interpreted within the context of their
Instrumental and reactive violence should be differ- limitations. The sample for this study was composed
entiated in future research in this area. of offenders who had been convicted of violent
offenses of a relatively low severity. The infrequency
with which serious violence had been committed
Implications of the Current Research
may be a consequence of sampling from a medium
Violence risk assessment: The value of being able security forensic facility. Although the total number
to estimate the likelihood that someone may be of cases (N 5 77) was not restrictively small for
violent in the future cannot be overstated. Currently, conducting binary logistic regression (the model
executive functions (or any cognitive abilities for used to evaluate severe violent offending), the
that matter) are not considered in the most widely number of cases in the severe violence cell was small
used and validated violence risk assessment tools (N 5 9), and small groups reduce the efficiency of
[e.g. Level of Service Inventory—Revised: Andrews the estimates [Stata Data Analysis Examples, 2008].
and Bonta, 2003; Violence Risk Appraisal Guide: Research conducted in forensic settings is asso-
Quinsey et al., 2006; Historical, Clinical, Risk ciated with additional limitations that threaten both
Management-20: Webster et al., 1997]. Given that the internal and external validity of its findings.
violence risk assessment is far from perfect, and Psychological testing is considered suspicious within
given that this study identified specific scores from the prison culture. It is possible that offenders with
the D-KEFS that were able to predict frequency and the most violent histories were the most suspicious,
severity of past violent offending, executive func- as they had been incarcerated the longest and were,
tioning may be a construct that merits further therefore, less likely to participate. As such, the
exploration in the field of violence risk assessment. offender population who volunteered to participate
Measures of executive functioning are easy to in this study may not be truly representative of
administer, they are objective, and executive func- all incarcerated offenders. An additional problem
tions are thought to be relatively stable across with testing incarcerated offenders is the issue of

Aggr. Behav.
348 Hancock et al.

motivation; although encouraged to try their best criminology and criminal justice. Trauma Violence Abuse
and despite perceived effort, the offenders may not 6:176–191.
have been completely motivated to complete the Bureau of Justice Statistics. 2009. The National Crime Victimi-
zation Survey: Criminal Victimization 2008. Washington, DC:
battery to the best of their ability, which threatens U.S. Department of Justice, Bureau of Justice Statistics, September
the internal validity of the research. 2009.
Burt AS. 2009. Are there meaningful etiological differences within
antisocial behavior? Results from a meta-analyses. Clin Psychol
CONCLUSION Rev 29:163–178.
Bushman BJ, Anderson CA. 2001. Is it time to pull the plug on the
This study was the first of its kind. Offenders were hostile versus instrumental aggression dichotomy? Psychol Rev
found to be characterized by patterns of broad and 108:273–279.
Cassidy JD, Carroll LJ, Pelso PM, Borg J, Holst H, Holm L, Kraus J,
pervasive executive dysfunction. More importantly, Coronada VG. (2004). Incidence, risk factors and prevention of
executive dysfunction was found to be related mild traumatic brain injury: Results of the WHO collaborating
specifically to violence and to the frequency and center task force on mild traumatic brain injury. J Rehabil Med
severity of violent offending. Despite the aforemen- 43:28–60.
tioned limitations, the results from this study are Chan RC, Shum D, Toulopoulou T, Chen EY. 2008. Assessment of
executive functions: Review of instruments and identification of
exciting and, if replicated, have significant implica-
critical issues. Arch Clin Neuropsychol 2:201–216.
tions for society. An understanding of the complex Cornell DG. 1996. Coding Guide for Violent Incidents: Instrumental
and interconnected variables underlying violence Versus Hostile/Reactive Aggression. Virginia: University of Virginia.
and criminality is a prerequisite for the early Coxe S, West SG, Aiken LS. 2009. The analysis of count data:
identification of at-risk youth and for the develop- A gentle introduction to Poisson regression and its alternatives.
ment of interventions for rehabilitating offenders. J Pers Assess 9:121–136.
Delis DC, Kaplan E, Kramer JH. 2001. Delis-Kaplan Executive
Identification of distinct predictors of frequent and Function System: Examiner’s Manual. San Antonio, TX: The
severe violent offending calls into question a ‘‘one Psychological Corporation.
size fits all’’ approach to remediation. Flannery DJ, Hussey D, Jefferis E. 2005. Adolescent delinquents and
violent behavior. In: Gullotta TP, Adams GR (eds). Handbook
ACKNOWLEDGMENTS of Adolescent Behavioral Problems: Evidence-Based Approaches
to Prevention and Treatment. New York: Springer Science1
This research was funded by an operating grant Business Media, pp 415–438.
(awarded to P.N.S.H.) from the Social Sciences and Greenfield R, Valliant PM. 2007. Moral reasoning, executive
function, and personality in violent and nonviolent adult
Humanities Research Council of Canada
offenders. Psychol Rep 101:323–333.
(Ref ]410-2008-2461). The authors acknowledge Hoaken PNS, Shaughnessy V, Pihl RO. 2003. Executive cognitive
the invaluable help of Marta Shaw, Joel Ginsburg, functioning and aggression: Is it an issue of impulsivity. Aggr
Chris Rowntree, and Bradley Corbett. Behav 29:15–30.
Hoaken PNS, Allaby DB, Earle J. 2007. Executive cognitive
functioning and the recognition of facial expressions of emotion
in incarcerated violent offenders, non-violent offenders, and
REFERENCES controls. Aggress Behav 33:412–421.
Andrews DA, Bonta J. 2003. Level of Service Inventory–Revised Homack S, Lee D, Riccio CA. 2004. Test review: Delis-Kaplan
(LSI-R). Toronto, ON: Multi-Health Systems. Executive Function System. J Clin Exp Neuropsychol 25:
Baker S, Ireland JL. 2007. The link between dyslexic traits, executive 599–609.
functioning, impulsivity, and social self-esteem among an Hutchinson MK, Holtman MC. 2005. Analysis of count data using
offender and non-offender sample. Law Psychiatry 30:492–503. Poisson regression. Res Nurs Health 28:408–418.
Barker ED, Séguin JR, White HR, Bates ME, Lacourse É, Jurado MB, Rosselli M. 2007. The elusive nature of executive
Carbonneau R, Tremblay RE. 2007. Developmental trajectories functions: A review of our current understanding. Neuropsychol
of male physical violence and theft. Arch Gen Psychiatry Rev 17:213–233.
64:592–599. Kashani JH, Jones MR, Bumby KM, Thomas LA. 1999. Youth
Baron IS. 2004. Delis-Kaplan executive function system. Child violence: Psychosocial risk factors, treatment, prevention, and
Neuropsychol 10:147–152. recommendations. J Emot Behav Disord 7:200–210.
Bornstein RA. 1990. Neuropsychological performance in children Kaufman AS, Kaufman NL. 2004. Kaufman Brief Intelligence
with Tourette’s syndrome. Psychiatry Res 33:73–81. Test-2. Circle Pines, MN: American Guidance Service.
Broomhall L. 2005. Acquired sociopathy: A neuropsychological Krpan K, Levine B, Stuss D, Dawson D. 2007. Executive function
study of executive dysfunction in violent offenders. Psychiatry and coping at one-year post traumatic brain injury. J Clin Exp
Psychol Law 12:367–387. Neuropsychol 29:36–46.
Brower MC, Price BH. 2001. Neuropsychiatry of frontal lobe Marsh NV, Martinovich WM. 2006. Executive dysfunction and
dysfunction in violent and criminal behavior: A critical review. domestic violence. Brain Injury 20:61–66.
J Neurol Neurosurg Psychiatry 71:720–726. Martell DA. 1992. Estimating the prevalence of organic brain
Bufkin JL, Luttrell VR. 2005. Neuroimaging studies of aggressive dysfunction in maximum-security forensic psychiatric patients.
and violent behavior: Current findings and implications of J Forensic Sci 37:878–893.

Aggr. Behav.
Predicting Frequency and Severity of Violence 349

Meloy JR. 2006. Empirical basis and forensic application of affective Serin RC, Preston DL. 2001. Managing and treating violent
and predatory violence. Aust N Z J Psychiatry 40:539–547. offenders. In: Ashford JLB, Sales BD, Reid W (eds). Treating
Miura H. 2009. Differences in frontal lobe function between violent Adult and Juvenile Offenders with Special Needs. Washington,
and nonviolent conduct disorder in male adolescents. Psychiatry DC: American Psychological Association, pp 249–271.
Clin Neurosci 63:161–166. Shallice T, Burgess PW. 1991. Deficits in strategy application
Morgan AB, Lilienfeld SO. 2000. A meta-analytic review of following frontal lobe damage in man. Brain 114:727–741.
the relation between antisocial behavior and neuropsycho- Shunk AW, Davis AS, Dean RS. 2006. Test review: Delis Kaplan
logical measures of executive function. Clin Psychol Rev executive function system (D-KEFS). Appl Neuropsychol 13:275–279.
20:113–136. Slaughter B, Fann JR, Ehde D. 2003. Traumatic brain injury in a
Motiuk LL, Vuong B. 2005. Homicide, Sex, Robbery, and Drug county jail population: Prevalence, neuropsychological function-
Offenders in Federal Corrections: An End-of-2004 Review (Research ing, and psychiatric disorders. Brain Injury 17:731–741.
Brief No. B-37). Ottawa, Ontario: Correctional Service Canada. Stata Data Analysis Examples: Logit Regression. 2008. UCLA:
Pliszka S. 2007. Practice parameter for the assessment and treatment Academic Technology Services, Statistical Consulting Group.
of children and adolescents with attention-deficit/hyperactivity Retrieved May 20, 2008 from http://www.ats.ucla.edu/stat/stata/
disorder. J Am Acad Child Adolesc Psychiatry 46:894–921. dae/logit.htm
Quinsey VL, Harris GT, Rice ME, Cormier CA. 2006. Violent Stuss DT, Alexander M. 2000. Executive functions and the frontal
Offenders: Appraising and Managing Risk. Washington, DC: lobes: A conceptual view. Psychol Res 63:289–298.
American Psychological Association. Suchy Y. 2009. Executive functioning: Overview, assessment, and
Raaijmakers MAJ, Smidts DP, Sergeant JA, Maassen GH, research issues for non-neuropsychologists. Ann Behav Med
Posthumus JA, van Engeland H, Matthys W. (2008). Executive 37:106–116.
functions in preschool children with aggressive behavior: Impair- Velligan D, Bow-Thomas C. 1999. Executive function in schizo-
ments in inhibitory control. J Abnorm Child Psychol 36:1097–1107. phrenia. Semin Clin Neuropsychiatry 4:24–33.
Rath JF, Simon D, Langenbahn DM, Sherr RL, Diller L. 2003. The Walters GD. 2007. Using Poisson class regression to analyze count
construct of problem solving in higher level neuropsycho- data in correctional and forensic psychology. Crim Justice Behav
logical assessment and rehabilitation. Arch Clin Neuropsychol 35:1659–1674.
19:613–635. Webster CD, Douglas KS, Eaves D, Hart SD. 1997. HCR-20:
Ridout M, Demétrio CGB, Hinde J. 1998. Models for count data Assessing Risk for Violence, Version 2. Burnaby, BC: Mental
with many zeros. Proc Int Biometr Conf, Cape Town 98:179–192. Health, Law, and Policy Institute, Simon Fraser University.
Scarpa A, Raine A. 2007. The biosocial basis of violence. In: World Health Organization. 2002. World report on violence and
Flannery DJ, Vazsonyi AT, Waldman ID (eds). The Cambridge health. Retrieved May 20, 2009 from http://www.who.int/
Handbook of Violent Behavior and Aggression. New York: violence
Cambridge University Press, pp 151–169. Worthington A. 2005. Rehabilitation of executive deficits: Effective
Séguin JR, Nagin D, Assaad J, Tremblay RE. 2004. Cognitive- treatment of related disabilities. In: Halligan P, Wade D (eds).
neuropsychological function in chronic physical aggression and The Effectiveness of Rehabilitation for Cognitive Deficits.
hyperactivity. J Abnorm Psychol 113:603–613. London: Oxford University Press, pp 257–270.

Aggr. Behav.

You might also like