Law Hum Behav (2009) 33:308–319

DOI 10.1007/s10979-008-9155-7

ORIGINAL ARTICLE

Predicting Short-Term Institutional Aggression in Forensic

Patients: A Multi-Trait Method for Understanding Subtypes
of Aggression
Michael J. Vitacco Æ Gregory J. Van Rybroek Æ
Jill E. Rogstad Æ Laura E. Yahr Æ James D. Tomony Æ
Emily Saewert

Published online: 29 October 2008

Ó American Psychology-Law Society/Division 41 of the American Psychological Association 2008

Abstract Accurately predicting inpatient aggression is an themselves from the same (Carmel & Hunter, 1993; Martin
important endeavor. The current study investigated inpa- & Daffern, 2006). Not surprisingly, inpatient aggression
tient aggression over a six-month time period in a sample extracts high emotional and financial costs from staff,
of 152 male forensic patients. We assessed constructs of patients, and institutions (Duffy, Doyle, & Ryan, 2004;
psychopathy, anger, and active symptoms of mental illness Hunter & Carmel, 1992). Clearly, a need exists for spe-
and tested their ability to predict reactive and instrumental cialized assessment protocols and the development of
aggression. Across all levels of analyses, anger and active empirically informed treatment plans aimed at reducing
symptoms of mental illness predicted reactive aggression. aggression and violence within forensic hospitals. Devel-
Traits of psychopathy, which demonstrated no relation- opment of these plans can only take place once there is a
ship to reactive aggression, were a robust predictor of more complete understanding of factors related to aggres-
instrumental aggression. This study (a) reestablishes psy- sion. To that end, there has been a recent upswing in the
chopathy as a clinically useful construct in predicting quantity and quality of research focused on the prediction
inpatient instrumental aggression, (b) provides some vali- of inpatient aggression (Grevatt, Thomas-Peter, & Hughes,
dation for the reactive/instrumental aggression paradigm in 2004; McDermott, Edens, Quanbeck, Busse, & Scott,
forensic inpatients, and (c) makes recommendations for 2008).
integrating risk assessment results into treatment Extant research highlights difficulties in making accu-
interventions. rate predictions of inpatient aggression. Two main issues
hinder accurate prediction. First, individuals who commit
Keywords Institutional aggression
Psychopathy
aggressive acts within an institution constitute a heteroge-
Prediction neous group, both in terms of symptoms and previous
aggression (Gadon, Johnstone, & Cooke, 2006; McDermott
et al., 2008). As such, the prototypical aggressive forensic
Mental health practitioners and direct care staff working in patient often cannot be easily identified. Second, scholars
inpatient settings face the daunting task of treating difficult creating ratings for aggression have failed to achieve uni-
and aggressive patients while at times having to defend formity in their operational definitions. As noted by Guy,
Edens, Anthony, and Douglas (2005), operational confu-
sion over what constitutes aggression may explain
M. J. Vitacco (&)
G. J. Van Rybroek
discrepant results across constructs and studies. Com-
J. D. Tomony
E. Saewert
pounding the issue is that many scholars in the field
Mendota Mental Health Institute, Madison, WI, USA
e-mail: [email protected] recognize a distinction between reactive/hostile and
instrumental aggression (see Fontaine, 2007), yet few
J. E. Rogstad researchers incorporate this distinction into their studies of
University of North Texas, Denton, TX, USA
inpatient aggression. As described by Cornell et al. (1996),
L. E. Yahr the two hallmarks of instrumental aggression are goal-
The University of Wisconsin-Madison, Madison, WI, USA directedness and planning, while the hallmarks of reactive

123
Law Hum Behav (2009) 33:308–319 309

aggression include provocation and arousal to hostility. Taylor, DuQueno, & Novaco, 2004) and symptoms of
One hypothesis that warrants testing is determining if dif- mental illness (Doyle & Dolan, 2006; Nolan et al., 2003;
ferential relationships exist between predictor variables of Waldheter, Jones, Johnson, & Penn, 2005). McDermott
aggression and aggression subtypes. If differences do exist, et al. (2008) found anger predicted staff-directed aggres-
then differentiating between reactive and instrumental sion, while symptoms of mental illness predicted patient-
aggression likely will provide a framework to assist in the directed aggression. Results with mentally disordered
reconciliation of mixed findings. Moreover, there is solid offenders incarcerated on a correctional mental health unit
theoretical and empirical support for understanding also supported the idea that anger was prognostic of
aggression through the paradigm of instrumental and inpatient aggression (Wang & Diamond, 1999).
reactive subtypes (Fontaine, 2007; Meloy, 2006). The presence of positive symptoms of mental illness has
Failure to account for types of aggression may be a predicted inpatient aggression across studies; fortunately,
primary reason for the lack of consistent findings. A pri- these symptoms often respond to medication management
mary example of a construct yielding mixed predictive (Foley et al., 2005; Nolan et al., 2005). It is precisely the
validity for inpatient aggression is psychopathy. Psychop- malleability of dynamic factors that presents an opportu-
athy is defined by interpersonal arrogance, callousness, and nity to treat and minimize inpatient aggression. Another
poor behavioral controls and was initially thought to be important consideration is that base rates of inpatient vio-
underpinned by two separate, yet interrelated factors. lence and aggression are generally low (Rogers & Shuman,
Factor 1 measures superficiality, grandiosity, and lack of 2005), likely due to strict medication monitoring and the
empathy and remorse. Factor 2 is made up of traits of consistent presence of staff. Low base rate events can be
impulsive, irresponsible, and antisocial behavior (see Hare, more difficult to predict (Harrison, 2002; Martin & Terris,
1996, 2003; Rogers et al., 2000). In looking at predictive 1991; Wollert, 2006). Given limitations inherent in pre-
validity of psychopathy in inpatient settings, several earlier dicting violence, it is not surprising that mental health
studies found a fairly robust relationship between psycho- professionals were significantly more accurate in predict-
pathic traits and institutional aggression (Buffington- ing which patients would not be aggressive compared to
Vollum, Edens, Johnson, & Johnson, 2002; Hare & predicting which patients would aggress (Haim, Rabino-
McPherson, 1984; Hildebrand, de Ruiter, & Nijman, 2004; witz, Lereya, & Fenning, 2002). In sum, multiple factors
Hill, Rogers, & Bickford, 1996; Morrissey, Mooney, are responsible for how inpatient aggression is manifested
Hogue, Lindsay, & Taylor, 2007; Reiss, Grubin, & Meux, and clinicians must rely on both static and dynamic vari-
1999). A longitudinal study by Hill, Neumann, and Rogers ables if they hope to develop accurate prediction models.
(2004) found scores on the Psychopathy Checklist: Given the substantial costs associated with inpatient
Screening Version (PCL:SV, Hart, Cox, & Hare, 1995) aggression and inconsistent findings across studies, it is an
accounted for 31% of the variance associated with area ripe for further research. With proper evaluative
aggressive behavior in 149 male inpatients followed over a techniques, institutions can continue to devote resources
six-month period. aimed at identifying patients most at-risk for aggressive
Yet, other studies have questioned the predictive power behavior. Knowledge of empirically based predictors
of psychopathy for inpatient aggression. Guy and col- improves the likelihood that mental health practitioners and
leagues’ (2005) meta-analysis casted doubt on the ability of researchers will be able to accurately identify and treat
psychopathy to consistently predict inpatient aggression, as patients most at-risk for institutional aggression.
it yielded wide-ranging effect sizes across 38 independent
samples. In a more recent meta-analysis that included
multiple effect sizes evaluating the relationship between THE PRESENT STUDY
PCL instruments and institutional aggression, Leistico,
Salekin, DeCoster, and Rogers (2008) found moderate Research (Daffern et al., 2007; Quanbeck et al., 2007) has
weighted Cohen’s ds (.53, .41, and .51) for PCL-R total, found inpatient aggressive incidents can be identified on
Factor 1, and Factor 2 scores, respectively. the basis of their function (i.e., reactive versus instrumen-
Beyond a single static variable, the necessity of using tal). Capitalizing on the importance of measuring inpatient
dynamic variables has been discussed in several influential infractions and aggression, this study aims to improve on
articles (Daffern & Howells, 2007a; Douglas & Skeem, earlier methodologies by: (a) considering both static and
2005; Miller, 2006; Olver, Wong, Nicholaichuk, & Gor- dynamic factors in the short-term prediction of physical
don, 2007; Simourd, 2004). To that end, we identified two and verbal aggression, (b) measuring aggression through
dynamic variables with substantial research linking them to the reactive and instrumental paradigm, and (c) evaluating
inpatient aggression: anger (Daffern, Howells, & Ogloff, whether differential relationships exist between criterion
2007; Mills & Kroner, 2003; Novaco & Taylor, 2004; variables (i.e., psychopathy, anger, and mental illness) and

123
310 Law Hum Behav (2009) 33:308–319

subtypes of aggression. In assessing criterion variables, we (2005), the BPRS is composed of 18 items that load on five
relied on structured interviews with forensic patients, col- scales: Affect (anxiety, guilt, depression, and somatic),
lateral information from multiple sources, official legal Positive Symptoms (thought content, conceptual disorga-
documentation from medical records, and ratings from nization, hallucinatory behavior, and grandiosity),
experienced mental health practitioners and research Negative Symptoms (blunted affect, emotional withdrawal,
assistants. This strategy afforded the best opportunity for and motor retardation), Resistance (hostility, uncoopera-
reliable and valid measurement of the criterion variables. tiveness, and suspiciousness), and Activation (excitement,
tension, and mannerisms-posturing). Forensic psycholo-
gists, all with training in scoring the BPRS and familiar
METHOD with the patient’s symptoms, completed BPRS ratings. To
ensure accurate measurement, every 10th BPRS was
Participants selected for evaluation of interrater reliability. A two-way
mixed model intraclass correlation coefficient for the total
Participants in this study included 152 male forensic score was .89, demonstrating consistent measurement
inpatients adjudicated Not Guilty by Reason of Mental across ratings for the duration of the study.
Disease or Defect (i.e., NGRI) who were involved in this The Ward Anger Rating Scale (WARS; Novaco, 1994)
research from June 2005 through August 2006. Patients was utilized because it is specifically designed as an anger
were found NGRI for a wide variety of crimes ranging assessment tool ‘‘for ease of recording in busy clinical
from simple assault to first degree intentional homicide. settings’’ (Novaco & Taylor, 2004, p. 45) and because
Commitment lengths ranged from one year to the possi- scoring relies on staff members’ direct observation of the
bility of life in a state mental health facility.1 The patients patient’s behavior. The WARS is a specialized measure
ranged in age from 18 to 70 (M = 38.30, SD = 12.22). designed for assessing anger and its correlates in inpatient
The majority of the participants were European American settings.
(n = 95, 62.6%); however, ethnic minorities were repre- The WARS consists of two parts. Part A is composed of
sented by African Americans (n = 40, 26.3%), Hispanic 18 items scored yes/no based for the presence or absence of
Americans (n = 10, 6.6%), and a combination of Native verbal (e.g., yelling) and physical (e.g., tantrum) correlates
Americans and biracial patients (n = 7, 4.6%). The of anger that occurred over the past week. Part B consists
majority of patients had a primary diagnosis of psychotic of seven items scored on a five-point Likert-type scale (not
disorder (65.0%); however, other disorders were present at all, very little, sometimes, fairly often, and very often)
and included major mood disorders (major depression and measuring additional correlates of anger (e.g., irritability).
bipolar disorders) as well as Axis II disorders. The majority The WARS has proven to be highly reliable and valid in
of the sample (70.0%) had more than one diagnosis. assessing anger in inpatient settings (Doyle & Dolan, 2006;
The state hospital where this research was conducted has Novaco & Taylor, 2004; Taylor et al., 2004). In the current
181 forensic beds divided over three security levels: min- study, ratings on Part A ranged from 0 to 8 (M = 1.61,
imum, medium, and maximum. To achieve a greater SD = 2.04); ratings on Part B ranged from 0 to 28
degree of score variance, patients were recruited and par- (M = 6.82, SD = 6.82); total WARS ratings ranged from 0
ticipated from all security levels. This research was to 36 (M = 8.43, SD = 7.70). Interrater reliability with a
approved by the Institutional Review Board for the state two-way mixed model intraclass correlation coefficient for
hospital prior to the onset of data collection. the total score was .74, demonstrating adequate measure-
ment across ratings.
Criterion Measures The Psychopathy Checklist: Screening Version
(PCL:SV; Hart, Cox, & Hare, 1996) is a 12-item rater-
The Brief Psychiatric Rating Scale (BPRS; Overall & based instrument designed to measure psychopathic traits
Gorham, 1962) was used to measure mental health symp- that consists of four factors: Interpersonal, Affective,
toms. The BPRS is widely applied in mental health Lifestyle, and Antisocial tendencies (for information on the
research and inpatient settings to monitor the presence and PCL:SV four-factor model see Jackson, Neumann, & Vit-
subsequent change in symptoms following the introduction acco, 2007; Vitacco, Neumann, & Jackson, 2005). The
of psychotropic medications. As elucidated by Shafer PCL:SV has been previously used with robust results in
studies of inpatient violence (Hill et al., 2004) and is
1
In the state the study occurred in, individuals found not guilty by strongly correlated (r = .80) with the Psychopathy
reason of insanity (NGRI) have the legal right to petition for
Checklist-Revised (Hare, 2003), often used as the criterion
conditional release every 6 months. As such, even a patient who
could remain in the hospital for the rest of his life could be measure in studies of psychopathy. Given the PCL:SV has
conditionally released 6 months after being found NGRI. been successfully used in previous studies of inpatient

123
Law Hum Behav (2009) 33:308–319 311

aggression and its strong correlation to the PCL-R, we aggression, (b) goal-directed act, (c) perception that the act
decided to use the PCL:SV to measure the construct of was unprovoked, and (d) presence of intense anger during
psychopathy in the present study. the aggression. Scoring on the first two are indicative of
Ratings for the current study were based on both instrumental aggression, while the latter are indicative of
intensive clinical interviewing and file review of both their reactive aggression. This scoring system further divided
current inpatient stay and community history. Raters for acts into reactive offenders (i.e., aggression within the
psychopathy were students trained in structured inter- institution was entirely reactive) and mixed/instrumental
viewing and supervised by Ph.D.-level clinicians trained in (aggression was instrumental as well as reactive). In total,
psychopathy assessment. In addition, all raters underwent 33 individuals were classified as engaging in only reactive
training at the hospital to discuss scoring issues that often aggressive and 11 were classified as engaging in at least
arise with mentally disordered offenders. Scores on the one act of instrumental aggression.
PCL:SV higher than 18 denote a categorical classification
of psychopathy (Hart et al., 1996); however, recent taxo- Procedure
metric evidence (Walters et al., 2007) indicates PCL:SV
scores are most appropriately considered as dimensional. Patients throughout the institution and residing on all
Scores on the PCL:SV ranged from 4 to 22 (M = 11.07, security units were eligible to participate in the research
SD = 4.05). Every 10th PCL:SV (i.e., 15 total) was scored study. Participation required a willingness to be inter-
by a second rater. A two-way mixed model intraclass viewed by a research assistant completing the PCL:SV; the
correlation coefficient for the total PCL:SV score was .88, other primary instruments did not require patient partici-
showing consistent measurement during the study. pation. Standard institutional security protocol was
followed, which dictated that interviewers could be
Outcome Data observed but not heard by staff. Any patient who wished to
participate in the study signed an informed consent form.
Outcome data on aggression were collected by trained Consent documented that participation was voluntary, they
research assistants using the Overt Aggression Scale (OAS; could desist at any time, staff would have no access to the
Yudofsky, Silver, Jackson, Endicott, & William, 1986). psychopathy ratings obtained during the clinical inter-
The OAS has four categories that can be scored: (a) verbal views, and participation in the study would have no
aggression, (b) physical aggression against objects, (c) influence on privileges or security level at the hospital. As
physical aggression against self, and (d) physical aggres- a prospective longitudinal study, follow-up data were not
sion against others. For the purposes of this study, only collected until completion of all criterion measures.
verbal aggression and physical aggression against others
were used for outcome analyses. Statistical Analyses
Data collection was done via extensive file review for a
six-month period after completion of the assessment bat- This study was designed to evaluate factors related to the
tery. Hospital policy dictates that all aggressive behaviors prospective prediction of inpatient aggression in a sample
and threats be documented in the medical record. Previous of forensic patients committed under NGRI. In addition,
research using this methodology has resulted in highly the study aimed to evaluate the hypothesis that subtypes of
reliable and valid results (see Hill et al., 2004). Results for aggression will be differentially related to our variables of
aggression were coded two ways: First, frequency of verbal interest (i.e., psychopathy, anger, and symptoms of mental
aggression and physical aggression against others were illness). First, we evaluated our hypotheses by testing
tallied. Based on those tallies, each patient was assigned group differences on the predictor measures based on
two separate ratings (i.e., yes or no) based on whether there physical aggression (coded as 0 or 1) as well as verbal
were any incidents of physical or verbal aggression. To aggression (coded as 0 or 1). We ran an ANOVA with
avoid criterion contamination, file information used to Bonferroni corrections to evaluate the differences on
score the PCL:SV was collected within a week of the onset measures between groups with no physical aggression and
of the six-month follow-up period. There was no overlap those who engaged in reactive or instrumental aggression.
between the collection of the criterion variables and out- To supplement our group differences, we used receiver
come data. operating characteristics (ROC) curve analyses with area
Each incident of physical aggression toward others was under curves (AUCs) as the corresponding statistic. In this
scored by two raters for instrumental and reactive aggres- study, AUCs measured the accuracy of a measure based on
sion on a modified scoring system based on the work of its ability to separate those engaging in institutional
Cornell et al. (1996). For this study, four separate criteria aggression from those not engaging in institutional
were considered: (a) planning or preparation before the aggression. Values from .90 to 1 represent excellent

123
312 Law Hum Behav (2009) 33:308–319

Table 1 Correlations among criterion variables

WARS A WARS B BPRS Aff BPRS Pos BPRS Neg BPRS Res BPRS Act PCL Int PCL Affect PCL Lif PCL Ant

WARS A (.72) .73** .003 .30** .03 .42** .39** .02 .07 .12 .22**
WARS B (.87) .05 .27** .02 .45** .38** .07 .23** .12 .31**
BPRS Aff (.71) .16 .26** .25** .41** -.26** -.20** -.08 -.09
BPRS Pos (.83) .16 .59** .45** .09 -.08 .08 .07
BRPS Neg (.76) .22** .16 -.34** -.13 -.10 -.07
BPRS Res (.75) .60** -.25** .006 .04 .18
BPRS Act (.68) -.08 -.10 -.11 .06
PCL Int (.80) .45** .07 .08
PCL Affect (.77) .15 .23*
PCL Lif (.63) .49**
PCL Ant (.68)
Notes. WARS A = WARS Part A, WARS B = WARS Part B. BPRS Scales: Aff = Affect, Pos = Positive Symptoms, Neg = Negative
Symptoms, Res = Resistance, and Acc = Activation. PCL:SV scales: Int = Interpersonal, Affect = Affective, Lif = Lifestyle, and
Ant = Antisocial tendencies
* p \ .05, ** p \ .01

prediction while those from .50 to .60 represent chance variables via Pearson correlations. Table 1 demonstrates
selection (Pepe, 2003).2 An advantage of ROC/AUC the inter-relationships between BPRS scales and both
analyses is their strength in dealing with low base rate scales of the WARS. BPRS Resistance symptoms were
events, which makes them particularly well-suited for highly related to both parts of the WARS (r = .42 and .47,
predicting institutional aggression. respectively). The other notable feature is the relative
independence of the personality-based facets of the
PCL:SV; the Affective facet demonstrated a significant
RESULTS relationship with Part B of the WARS (i.e., r = .23), but no
other criterion.
The base rate of physical aggression in the current study
was relatively low with 29.0% (n = 44) of the total sample
engaging some form of physical aggression during the Group Differences for Aggression Subtypes
study period. Of those 44, 75.0% (n = 33) engaged in
reactive aggression only, while 25.0% (n = 11) engaged in Computing a power analysis for patients who aggressed
at least one incident of instrumental aggression. Overall, (n = 44) with individuals who did not aggress (n = 108)
these individuals were responsible for 139 documented indicated appropriate power (.92) to detect meaningful dif-
instances of physical aggression. There were no differences ferences. The overall ANOVA model based on group
in the mean number of aggressive incidents between those differences for aggression was significant, F = 6.49,
who committed only reactive aggression (M = 3.03, p \ .001. As illustrated in Table 2, results indicated that
SD = 2.79) versus those who committed at least one act of patients who aggressed had significantly higher scores on
instrumental aggression (M = 3.27, SD = 4.00, t = .22, both parts of the WARS (p \ .001), with large effect sizes.
p = .73). In comparing individuals who did not aggress Likewise, active symptoms of mental health measured by
with those that did aggress, both age F(1, 150) = .23, BPRS Affect symptoms (p = 0.01), BPRS Positive symp-
p = .23 and ethnicity v2 = .77, p = .86 were unrelated to toms (p = 0.01), BPRS Resistance symptoms (p \ .001)
the commission of institutional aggression. Verbal and BPRS Activation symptoms (p \ .001) were associated
aggression was a more common occurrence within the with inpatient aggression with moderate to large effect sizes.
institution, with 53.0% (n = 81) of the patients engaging in Regarding psychopathy, consistent with some earlier
verbal aggression. research (Guy et al., 2005; McDermott et al., 2008) only the
Prior to testing the predictive power of our criterion Antisocial tendencies factor of the PCL:SV reached signif-
measures, we evaluated the relationships among the icance (p = .009; Cohen’s d = 0.47), suggesting that the
majority of psychopathic traits are not associated with
2 overall physical aggression in an inpatient setting.
Interpretation of AUC curves are as follows: .90–1 excellent; .80–
.90 good; .70–.80 fair, .60–.70 poor; and .60–.50 are considered fail Table 3 lists the results for verbal aggression. Notably, a
and indicate chance performance. similar pattern emerged for individuals engaging in verbal

123
Law Hum Behav (2009) 33:308–319 313

Table 2 Group differences on

Criterion No physical Physical aggression F p Cohen’s
criterion measures for total
(n = 108) (n = 44) d-scores
physical aggression
M SD M SD

WARS A 1.08 1.48 2.91 2.59 29.95 \.001 0.98

WARS B 4.98 4.48 11.34 7.15 43.68 \.001 1.18
BPRS Affect 6.77 2.92 8.20 3.85 6.25 .01 0.44
BPRS Positive 9.69 4.66 12.02 5.90 6.71 .01 0.46
BPRS Negative 5.89 3.51 6.20 3.70 0.25 .62 0.09
BPRS Resistance 6.03 3.06 9.05 4.45 23.00 \ .001 0.86
BPRS Activation 4.71 1.89 7.32 3.44 35.62 \ .001 1.07
PCL Interpersonal 1.43 1.57 1.48 1.75 0.03 .86 0.03
PCL Affective 3.09 1.82 2.95 1.55 0.19 .66 -0.08
PCL Lifestyle 3.09 1.43 3.22 1.24 0.30 .59 0.09
PCL Antisocial 3.26 1.41 3.93 1.45 6.99 .009 0.47
PCL Total 10.86 4.02 11.59 4.11 1.02 .32 0.18

Table 3 Group differences on

Criterion No verbal (n = 71) Verbal aggression (n = 81) F p Cohen’s d-scores
criterion measures for verbal
aggression M SD M SD

WARS A 0.77 1.23 2.35 2.31 26.30 \.001 0.84

WARS B 3.87 3.92 9.41 3.49 39.08 \.001 1.50
BPRS Affect 7.08 3.60 7.27 2.96 0.12 .72 0.06
BPRS Positive 9.31 4.00 11.28 5.83 5.75 .02 0.39
BPRS Negative 6.34 3.72 5.67 3.40 1.35 .25 -0.19
BPRS Resistance 5.58 2.95 8.06 4.03 18.38 \.001 0.70
BPRS Activation 4.56 1.71 6.26 3.15 16.37 \.001 0.66
PCL Interpersonal 1.37 1.56 1.50 1.67 .67 .60 0.08
PCL Affective 2.93 1.76 3.16 1.74 .42 .42 0.13
PCL Lifestyle 2.79 1.48 3.43 1.20 8.71 .004 0.48
PCL Antisocial 3.03 1.22 3.83 1.54 12.34 .001 0.57
PCL Total 10.10 3.68 11.93 4.18 8.07 .005 0.44

aggression (n = 81) compared to those who did not engage level revealed a distinct pattern of results, strongly sug-
in verbal aggression (n = 71). BPRS Affect symptoms did gesting criterion variables were differentially associated
not differentiate groups on verbal aggression (p = .72); with aggression subtypes. Table 4 presents selected dif-
however, the Lifestyle facet (p = .004) from the PCL:SV ferences that emerged when subtyping physical aggression.
did differentiate verbal aggressors from patients who did Notably, the personality features of psychopathy, namely
not verbally aggress. Notably, all other differences between Interpersonal (p \ .001) and Affective (p = .02) features,
groups for physical aggression remained for verbal were significantly higher in patients who engaged in
aggression. Effect sizes were moderate for verbal aggres- instrumental aggression, demonstrating large effect sizes
sion, except for Part B of the WARS (Cohen’s d = 1.50). (Cohen’s ds = 2.36 and 1.41, respectively). The Antisocial
To conduct a finer-grained analysis to further explore tendencies facet was lower in the no-aggression group
group differences in inpatient aggression, we divided the compared to the instrumental aggressors (p = .02, Cohen’s
sample into three groups: instrumental aggressors d = 0.79).
(n = 11), reactive aggressors (n = 33), and those who did In considering active symptoms of mental illness, the
not commit any physical aggression (n = 108). However, BPRS Affect and Positive symptoms were higher in the
the following results should be considered preliminary as reactive aggression group compared to the no physical
power for this analysis was low (.39). Again, the overall aggression group, but did not differentiate based on sub-
ANOVA model revealed group differences, F = 5.27, types of aggression. The Resistance and Activation scales
p \ .001; Bonferroni corrections at the .025 significance of the BPRS successfully differentiated groups on reactive

123
314 Law Hum Behav (2009) 33:308–319

Table 4 Group differences on

Criterion measures No physical Reactive (2) Instrumental Group differencesa Cohen’s d-scoresb
selected criterion measures for
(1) (n = 108) (n = 33) (3) (n = 11)
no aggression, reactive
aggression, and instrumental M SD M SD M SD
aggression
WARS A 1.08 1.48 3.12 2.75 2.27 2.00 1 \ 2*** 1.10
1 \ 3* 0.71
BPRS Resistance 6.03 3.06 8.73 4.44 10.00 4.56 1 \ 2*** 0.79
1 \ 3*** 1.23
BPRS Activation 4.71 1.89 7.42 3.46 7.00 3.52 1 \ 2** 1.15
a 1 \ 3*** 1.10
Group differences were based
on analyses with Bonferonni PCL Interpersonal 1.43 1.57 0.76 1.15 3.76 1.43 1 \ 3*** 1.56
correction 2 \ 3*** 2.36
b
Only d-scores for significant PCL Affective 3.09 1.82 2.48 1.33 4.36 1.36 2 \ 3** 1.41
findings are listed PCL Lifestyle 3.09 1.43 3.24 1.39 3.18 0.60 1=2=3 N/A
* p \ .05, ** p \ .01, PCL Antisocial 3.26 1.41 3.79 1.54 4.36 1.12 1 \ 3* 0.79
*** p \ .001

Table 5 Area under curve

Criterion measures Reactive Instrumental Any aggression
results for subtypes of
aggression AUC Error Sig AUC Error Sig AUC Error Sig

WARS A .69 .06 \.001 .63 .08 .14 .70 .05 \.001
WARS B .74 .05 \.001 .71 .08 .02 .77 .05 \.001
BPRS Affect .67 .05 .005 .45 .10 .58 .62 .04 .02
BPRS Positive .61 .06 .05 .61 .09 .24 .63 .05 .01
BPRS Negative .56 .06 .31 .39 .09 .24 .51 .05 .80
BPRS Resistance .65 .06 .008 .72 .08 .02 .70 .05 \.001
BPRS Activation .71 .06 \.001 .66 .09 .07 .73 .05 \.001
PCL Interpersonal .35 .05 .01 .87 .04 \.001 .50 .05 .99
PCL Affective .38 .05 .04 .73 .06 .01 .48 .05 .66
PCL Lifestyle .55 .06 .40 .52 .06 .84 .55 .05 .38
PCL Antisocial .59 .06 .12 .70 .07 .03 .64 .05 .007
PCL Total .43 .05 .20 .85 .04 \.001 .54 .05 .73

and instrumental aggression compared to patients who both subscales of the WARS and four scales of the BPRS
engaged in the no physical aggression group. Part A of the predicted aggression.3
WARS acted in much the same manner, where it was lower The results for instrumental aggression revealed a dis-
in the no physical aggression group; however, no differ- tinctly different pattern. First, three facets of the PCL:SV
ence in the WARS scales were found when comparing (i.e., Affective, Interpersonal, and Antisocial) evidenced
reactive from instrumental aggressors. significant AUCs (range from .87 to .70) in the expected
direction. Both the WARS Part B (AUC = .71) and BPRS
AUCs Predicting Aggression Subtypes Resistance scale (AUC = .72) demonstrated fair predictive
ability. In sum, Affective and Interpersonal features of
Table 5 summarizes the ROC/AUC analyses for this psychopathy, which were unrelated to reactive aggression,
patient sample based on type of aggression committed. predicted instrumental aggression in this sample of forensic
Review of the resulting AUCs demonstrated a high con- inpatients.
sistency with the tests of group differences. Findings
indicated significant results for Part A of the WARS 3
For the sake of comparison we predicted verbal aggression. The
(AUC = .69), Part B of the WARS (.74), and four scales of criterion variables that were significant predictors of verbal aggres-
the BPRS (AUCs ranged from .61 to .71). The results from sion included WARS Part A (AUC = .71) and Part B (AUC = .77),
BPRS Resistance (AUC = .68), BPRS Activation (AUC = .66), and
the PCL:SV were contrary to the expected relationship PCL:SV Lifestyle (AUC = .65) and Antisocial tendencies
(i.e., low scores predicted reactive aggression). Likewise, (AUC = .67).

123
Law Hum Behav (2009) 33:308–319 315

Polytomous Logistic Regression Predicting difficult, yet crucial, ventures. As Monahan (1981) stated,
Aggression Subtypes ‘‘knowledge of the appropriate base rate is the most
important single piece of information necessary to make
The final evaluative approach utilized a polytomous an accurate prediction’’ (p. 60). In this statement, Mona-
logistic regression. Moreover, given the significant corre- han was directly referring to the prediction of aggression
lations among our criterion variables, a refined test of and violent behavior. He recognized the importance of
significance was warranted to evaluate which criteria integrating base rates with available clinical information
account for the greatest amount of variance. Polytomous to accurately predict aggression.
logistic regression is appropriate for dealing with multiple In moving beyond base rates to greater predictive
categorical dependent variables with continuous indepen- accuracy, it is necessary to identify, and then utilize
dent variables (Unrau & Coleman, 1998). In this case, the appropriate predictors variables. Recently, there has been
three categories are: (1) no aggression, (2) reactive literature suggesting that the construct of psychopathy,
aggression, and (3) instrumental aggression. The final once thought to be highly predictive of inpatient aggres-
model was significant v2(20) = 89.07, p \ .001, with an sion, is weakly associated with aggression in psychiatric
overall correct classification rate of 81.6%. The McFadden inpatients (Guy et al., 2005; McDermott et al., 2008). The
pseudo R2 was .38. In other words, the criterion variables current findings paint a different picture, instead suggesting
accounted for approximately 38% of the variance associ- that the minimization of psychopathy for predicting inpa-
ated with aggression classification. The results from the tient aggression has been oversimplified. In this study,
polytomous regression prediction model confirmed our psychopathy was not predictive of reactive aggression, but
earlier findings in that the Interpersonal facet from the traits of psychopathy were strongly predictive of instru-
PCL:SV was the strongest predictor of instrumental mental aggression; both Interpersonal and Affective facets
aggression (Wald = 7.69, p = .006) while Part B of the had good to fair AUCs, .87 and .73, respectively (see also
WARS was the strongest predictor of reactive aggression Vitacco, Neumann, Caldwell, Leistico, & Van Rybroek,
(Wald = 10.02, p = .002). 2006; Woodworth & Porter, 2002). The PCL:SV total score
was also a significant predictor of instrumental aggression.
Moreover, in the regression analyses, the Interpersonal
DISCUSSION facet was the strongest predictor of instrumental aggres-
sion. Although preliminary, this study extends the
Identifying individuals most at-risk for inpatient aggression relationship between psychopathy and instrumental
is a critical endeavor that is increasingly becoming the aggression to a secure forensic hospital setting. The present
purview of mental health practitioners. If the institution’s finding that psychopathy helped predict instrumental but
environment is not reasonably safe for staff and other not reactive aggression might help to understand incon-
patients, there is little chance for therapeutic endeavors to sistencies in findings that did not distinguish reactive from
take place. The more that inpatient aggression can be instrumental aggression or between facets of psychopathy.
predicted and minimized, a better chance exists for the Strong support exists for differentiating reactive from
creation of an effective treatment environment. In mak- instrumental aggression in forensic patients (Kockler,
ing such predictions of future inpatient aggression there are Stanford, Nelson, Meloy, & Sanford, 2006; Meloy, 2006),
several considerations that stem from our results that and the current study provides an additional layer of sup-
should improve clinical assessments of inpatient port for this distinction. As noted by Meloy (2000),
aggression. characteristics and behaviors associated with instrumental
An initial suggestion involves generating greater aggression are more difficult to predict. Thus, this study
knowledge of base rates of inpatient aggression. As noted provides some guidance as to identifying which individuals
by Meehl and Rosen (1955) base rates provide funda- are most likely to engage in instrumental aggression.
mental information to clinicians and assists in risk The current results suggest that failure to consider sub-
formulations. In this study, the base rate for physical types of aggression obscures potentially important
aggression was 29.0% and the rate for instrumental relationships between psychopathy and aggression. Essen-
aggression was much lower at 7.0%. The overall level of tially, failing to assess psychopathy may misidentify
aggressive incidents was consistent with recent research patients most at-risk for planned, goal-directed aggression.
evaluating inpatient aggression in forensic mental health Although individuals engaging in instrumental aggression
patients (Haggard-Grann, 2007; McDermott et al., 2008; represented only a small subgroup (n = 11) of the total
Quinsey, Jones, Book, & Barr, 2006). Thus, consideration sample, understanding and preventing aggression is salient
of base rates in predicting inpatient aggression will benefit to staff and patients working in mental health institutions
clinicians by providing a useful starting point to such (Daffern & Howells, 2007b). Unfortunately, the

123
316 Law Hum Behav (2009) 33:308–319

importance of differentiating instrumental from reactive LIMITATIONS AND FUTURE DIRECTIONS

aggression is frequently overlooked in inpatient settings.
All too often in post-event debriefings, direct-care staff The results of this study must be considered in light of
report that aggression ‘came out of the blue.’ For instru- several limitations. The primary limitation is the small
mental aggression, that may indeed be the case. Better number of patients (n = 11) who engaged in instrumental
understanding of idiographic factors related to aggression aggression. This was affirmed by the low power when this
should assist in the proactive development of treatment sample was evaluated. This small sample limits the overall
plans and safety measures designed to ameliorate generalizability of the study and subsequent findings.
aggression. Although we were encouraged by the consistency of results
The pattern of results from this study reaffirms the across all levels of analyses pertaining to instrumental
importance of using dynamic variables in assessing risk. aggression, the small number of patients only allows us to
Reactive aggression, the most common or ‘‘basic’’ type of arrive at preliminary conclusions. We are continuing to
aggression, was predicted by anger and mental health collect data on instrumental aggression and encourage
symptoms. Verbal aggression was predicted by the same. other researchers studying inpatient violence to differenti-
Quite possibly, the relationship between psychotic symp- ate between instrumental and reactive aggression.
toms and aggression stems from the lack of behavioral The results of the current study are promising with
control for individuals with severe psychotic symptoms respect to understanding factors related to inpatient
(Doyle & Dolan, 2006; Nolan et al., 2003). Likewise, aggression. However, more research is required to more
anger, as measured by the two scales of WARS, predicted fully understand the multifaceted phenomenon of inpatient
reactive aggression (AUCs = .69 and .74, respectively). aggression. For example, the criterion variables used in the
Part B of the WARS also predicted instrumental aggression polytomous logistic regression explained a significant
(AUC = .71). Anger is a predictor of inpatient aggression amount of variance associated with aggression (i.e., 38%);
(Daffern et al., 2007; Mills & Kroner, 2003) and was a however, that still leaves a great deal (i.e., 62%) of unac-
verified predictor in the current sample. A noteworthy counted variance. Clearly, there is a great deal left to
finding of the current study is the further identification of understand about the nature of inpatient aggression and
the WARS as a measure for assessing anger and predicting how to best employ assessments and interventions designed
inpatient aggression (Novaco & Taylor, 2004). A primary to predict and prevent violent behavior.
advantage of the WARS is that it relies on observable A recent development in predicting inpatient aggression
characteristics of anger where staff ratings of patients can is the creation of specialized measures. Instruments such as
occur on a daily basis, if desired. With careful observation, the Violence Risk Scale (VRS, Wong & Gordon, 2006) or
staff can use the WARS to identify which patients are in the Inventory of Offender Risks, Needs, and Strengths
need of immediate intervention. This practice, commonly (IORNS, Miller, 2006) are specifically designed to assess
referred to as risk-responsivity (Ward & Eccleston, 2004; risk in inpatient settings and for offenders on the cusp of
Wong & Hare, 2004), is an effective strategy for dealing release. Initial results for these instruments have been
with emerging problems in institutions (Logan & Gaes, promising; however, further study is needed to validate and
1993). strengthen initial findings. We must determine how spe-
Conceptually, the current results fit well with existing cialized instruments relate to subtypes of aggression and if
research, but also suggest new ideas for implementing they can be integrated within risk management plans. As of
intervention and treatment programs. First, the current this writing, there is no standard protocol for predicting
paper underscores the need for employing both static and inpatient aggression. It must yet be determined how for-
dynamic factors in predicting inpatient aggression. malized, structured instruments can be incorporated with
Dynamic factors can be reliably measured and effectively other measures, such as the WARS and PCL:SV, to aid in
treated (see Douglas & Skeem, 2005). Second, it reestab- the comprehensive assessment of potential for inpatient
lishes psychopathy as a salient predictor of aggression in an aggression. The clinical field would advance with refine-
inpatient setting. Research (Guy et al., 2005; McDermott ment of preventative safety and treatment intervention
et al., 2008) failing to distinguish between reactive and measures utilized with forensic patients, who, empirically,
instrumental aggression discounted the importance of are believed to be at risk for hospital aggression.
psychopathy for predicting aggression within institutional
settings. Quite possibly, psychopathy may mediate the
relationship between anger, symptoms of mental illness, SUMMARY
and institutional aggression. In other words, the presence of
psychopathic traits may lead to a basic propensity to use The current study sheds new light on methodological issues
aggression instrumentally. concerning the prediction of inpatient aggression. By

123
Law Hum Behav (2009) 33:308–319 317

assessing subtypes of aggression, the current data reestab- Fontaine, R. G. (2007). Disentangling the psychology and law of
lish psychopathy as a construct useful in understanding and instrumental and reactive subtypes of aggression. Psychology,
Public Policy, and Law, 13, 143–165. doi:10.1037/1076-
predicting inpatient aggression. However, other constructs 8971.13.2.143.
such as anger and active symptoms of mental illness were Gadon, L., Johnstone, L., & Cooke, D. (2006). Situational variables
valuable predictors of reactive aggression. The presence of and institutional violence: A systematic review of the literature.
differential relationships between predictors and instru- Clinical Psychology Review, 26, 515–534. doi:10.1016/j.cpr.
2006.02.002.
mental and reactive aggression underscores the necessity of Grevatt, M., Thomas-Peter, B., & Hughes, G. (2004). Violence,
assessing clinical constructs over multiple domains. mental disorder, and risk assessment: Can structured clinical
Assessing multiple domains empirically linked to aggres- assessments predict the short-term risk of inpatient violence?
sion offers clinicians the best opportunity for: (a) Journal of Forensic Psychiatry & Psychology, 15, 278–292. doi:
10.1080/1478994032000199095.
appropriately identifying patients most at-risk for subtypes Guy, L. S., Edens, J. F., Anthony, C., & Douglas, K. S. (2005). Does
of aggression, and (b) providing insight into which treat- psychopathy predict institutional misconduct among adults? A
ment can reduce which type of aggression. Practitioners meta-analytic investigation. Journal of Consulting and Clinical
can use information gleaned from risk assessments to Psychology, 73, 1056–1064. doi:10.1037/0022-006X.73.6.1056.
Haggard-Grann, U. (2007). Assessing violence risk: A review and
create idiographic interventions consistent with the goal of clinical recommendations. Journal of Counseling and Develop-
creating a safer environment for staff and patients alike. ment, 85, 294–301.
Haim, R., Rabinowitz, J., Lereya, J., & Fenning, A. (2002).
Predictions made by psychiatrists and psychiatric nurses of
violence by patients. Psychiatric Services (Washington, D.C.),
53, 622–624. doi:10.1176/appi.ps.53.5.622.
REFERENCES Hare, R. D. (1996). Psychopathy: A clinical construct whose time has
come. Criminal Justice and Behavior, 23, 25–54. doi:
Buffington-Vollum, J., Edens, J. F., Johnson, D. W., & Johnson, J. K. 10.1177/0093854896023001004.
(2002). Psychopathy as a predictor of institutional misbehavior Hare, R. D. (2003). Technical manual for the Revised Psychopathy
among sex offenders: A prospective replication. Criminal Justice Checklist (2nd ed.). North Tonawanda, NY: Multi-Health
and Behavior, 29, 497–511. doi:10.1177/009385402236730. Systems.
Carmel, H., & Hunter, M. (1993). Staff injuries from patient attack: Hare, R. D., & McPherson, L. M. (1984). Violent and aggressive
Five years’ data. The Bulletin of the American Academy of behavior in criminal psychopaths. International Journal of Law
Psychiatry and the Law, 21, 485–493. and Psychiatry, 7, 35–50. doi:10.1016/0160-2527(84)90005-0.
Cornell, D. G., Warren, J., Hawk, G., Stafford, E., Oram, G., & Pine, Hart, S. D., Cox, D., & Hare, R. D. (1996). Manual for the
D. (1996). Psychopathy in reactive and instrumental violent psychopathy checklist: Screening Version. Toronto, ON: Multi-
offenders. Journal of Consulting and Clinical Psychology, 64, Health Systems.
783–790. doi:10.1037/0022-006X.64.4.783. Hart, S. D., Cox, D. N., & Hare, R. D. (1995). Manual for the
Daffern, M., & Howells, K. (2007a). Aggressive behaviour in high- screening version of the psychopathy checklist revised
risk personality disordered inpatients during prison and follow- (PCL:SV). Toronto, ON: Multi-Health Systems.
ing admission to hospital. Psychiatry, Psychology and Law, 14, Harrison, D. (2002). Structure and timing in limited range dependent
26–34. doi:10.1375/pplt.14.1.26. variables: Regression models for predicting if and when. In
Daffern, M., & Howells, K. (2007b). Antecedents for aggression and the Measuring and analyzing behavior in organizations: Advances
function analytic approach to the assessment of aggression and in measurement and data analysis (pp. 446–497). San Francisco:
violence in personality disordered patients within secure settings. Jossey-Bass.
Personality and Mental Health, 1, 126–137. doi:10.1002/pmh.16. Hildebrand, M., de Ruiter, C., & Nijman, H. (2004). PCL-R
Daffern, M., Howells, K., & Ogloff, J. (2007). Interaction between psychopathy predicts disruptive behavior among offenders in a
individual characteristic and the function of aggression in Dutch forensic psychiatric hospital. Journal of Interpersonal
forensic psychiatric inpatients. Psychiatry, Psychology and Violence, 19, 13–29. doi:10.1177/0886260503259047.
Law, 14, 17–25. doi:10.1375/pplt.14.1.17. Hill, C. D., Neumann, C. S., & Rogers, R. (2004). Confirmatory factor
Douglas, K. S., & Skeem, J. L. (2005). Violence risk assessment: analysis of the Psychopathy Checklist: Screening Version in
Getting specific about being dynamic. Psychology, Public Policy, offenders with Axis I disorders. Psychological Assessment, 16,
and Law, 11, 347–383. doi:10.1037/1076-8971.11.3.347. 90–95. doi:10.1037/1040-3590.16.1.90.
Doyle, M., & Dolan, M. (2006). Evaluating the validity of anger Hill, C. D., Rogers, R., & Bickford, M. E. (1996). Predicting
regulation problems, interpersonal style, and disturbed mental aggressive and socially disruptive behavior in a maximum
state for predicting inpatient violence. Behavioral Sciences & the security forensic psychiatric hospital. Journal of Forensic
Law, 24, 783–798. doi:10.1002/bsl.739. Sciences, 41, 56–59.
Duffy, D., Doyle, M., & Ryan, T. (2004). Risk assessment and Hunter, M., & Carmel, H. (1992). The cost of staff injuries from inpatient
management in acute mental health care. In M. Harrison, D. violence. Hospital & Community Psychiatry, 43, 586–588.
Howard & D. Mitchell (Eds.), Acute mental health nursing: Jackson, R. L., Neumann, C. S., & Vitacco, M. J. (2007). Impulsivity,
From acute concerns to the capable practitioner (pp. 144–161). anger, and, psychopathy: The moderating effect of ethnicity.
Thousand Oaks, CA: Sage Publications, Inc. Journal of Personality Disorders, 21, 289–304. doi:
Foley, S. R., Kelly, B. D., Clarke, M., McTigue, O., Gervin, M., 10.1521/pedi.2007.21.3.289.
Kamali, M., et al. (2005). Incidence and clinical correlates of Kockler, T. R., Stanford, M. S., Nelson, C. E., Meloy, J. R., &
aggression and violence at presentation in patients with first Stanford, K. (2006). Characterizing aggressive behavior in a
episode psychosis. Schizophrenia Research, 72, 161–168. doi: forensic population. The American Journal of Orthopsychiatry,
10.1016/j.schres.2004.03.010. 76, 80–85. doi:10.1037/0002-9432.76.1.80.

123
318 Law Hum Behav (2009) 33:308–319

Leistico, A. R., Salekin, R. T., DeCoster, J., & Rogers, R. (2008). A Pepe, M. S. (2003). The statistical evaluation of medical tests for
large-scale meta-analysis relating the Hare measures of psy- classification and prediction. Oxford: Oxford University Press
chopathy to antisocial conduct. Law and Human Behavior, 32, Quanbeck, C. D., McDermott, B. E., Lam, J., Eisenstark, H., Sokolov,
28–45. doi:10.1007/s10979-007-9096-6. G., & Scott, C. L. (2007). Categorization of aggressive acts
Logan, C. H., & Gaes, G. (1993). Meta-analysis and the rehabilitation committed by chronically assaultive state hospital patients.
of punishment. Justice Quarterly, 10, 245–263. doi: Psychiatric Services (Washington, D.C.), 58, 521–528. doi:
10.1080/07418829300091811. 10.1176/appi.ps.58.4.521.
Martin, T., & Daffern, M. (2006). Clinician perceptions of personal Quinsey, V. L., Jones, G. B., Book, A. S., & Barr, K. N. (2006). The
safety and confidence to manage inpatient aggression in a forensic dynamic prediction of antisocial behavior among forensic
psychiatric setting. Journal of Psychiatric and Mental Health psychiatric patients: A prospective study. Journal of Interpersonal
Nursing, 13, 90–99. doi:10.1111/j.1365-2850.2006.00920.x. Violence, 21, 1539–1565. doi:10.1177/0886260506294238.
Martin, S. L., & Terris, W. (1991). Predicting infrequent behavior: Reiss, D., Grubin, D., & Meux, C. (1999). Institutional performance
Misconceptions surrounding misclassification errors. Forensic of male ‘psychopaths’ in a high-security hospital. Journal of
Reports, 4, 141–148. Forensic Psychiatry, 10, 290–299. doi:
McDermott, B. E., Edens, J. F., Quanbeck, C. D., Busse, D., & Scott, C. 10.1080/09585189908403682.
L. (2008). Examining the role of static and dynamic risk factors in Rogers, R., Salekin, R. T., Hill, C., Sewell, K. W., Murdock, M. E., &
the prediction of inpatient violence: Variable- and person-focused Neumann, C. S. (2000). The Psychopathy Checklist–Screening
analyses. Law and Human Behavior, 32, 325–338. Version: An examination of criteria and subcriteria in three
Meehl, P. E., & Rosen, A. (1955). Antecedent probability and the forensic samples. Assessment, 7, 1–15. doi:10.1177/1073191
efficiency of psychometric signs, patterns, and cutting scores. 10000700101.
Psychological Bulletin, 52, 194–216. doi:10.1037/h0048070. Rogers, R., & Shuman, D. W. (2005). Fundamentals of forensic
Meloy, J. R. (2000). Violence risk and threat assessment: A practical practice: Mental health and criminal law. New York: Springer
guide for mental health and criminal justice professionals. San Science ? Business Media.
Diego, CA: Specialized Training Services. Shafer, A. (2005). Meta-analysis of the Brief Psychiatric Rating Scale
Meloy, J. R. (2006). Empirical basis and forensic application of factor structure. Psychological Assessment, 17, 324–335. doi:
affective and predatory violence. The Australian and New 10.1037/1040-3590.17.3.324.
Zealand Journal of Psychiatry, 40, 539–547. doi:10.1111/j.1440- Simourd, D. J. (2004). Use of dynamic risk/need assessment
1614.2006.01837.x. instruments among long-term incarcerated offenders. Criminal
Miller, H. A. (2006). A dynamic assessment of offender risk, needs, Justice and Behavior, 31, 306–323. doi:10.1177/009385480
and strengths in a sample of pre-release general offender. 3262507.
Behavioral Sciences & the Law, 24, 767–782. doi:10.1002/ Taylor, J. L., DuQueno, L., & Novaco, R. W. (2004). Piloting a Ward
bsl.728. Anger Rating Scale for older adults with mental health problems.
Mills, J. F., & Kroner, D. G. (2003). Anger as a predictor of Behavioural and Cognitive Psychotherapy, 32, 467–479. doi:
institutional misconduct and recidivism in a sample of violent 10.1017/S135246580400164X.
offenders. Journal of Interpersonal Violence, 18, 282–294. doi: Unrau, Y. A., & Coleman, H. (1998). Understanding and interpreting
10.1177/0886260502250085. polytomous logistic regression: Applications to research on
Monahan, J. (1981). Predicting violent behavior: An assessment of social work practice. Research on Social Work Practice, 8, 223–
clinical techniques. Beverly Hills, CA: Sage. 235. doi:10.1177/104973159800800206.
Morrissey, C., Mooney, P., Hogue, T. E., Lindsay, W. R., & Taylor, J. Vitacco, M. J., Neumann, C. S., Caldwell, M. F., Leistico, A., & Van
L. (2007). Predictive validity of the PCL-R for offenders with Rybroek, G. J. (2006). Testing four models of the Psychopathy
intellectual disability in a high security hospital: Treatment Checklist: Youth Version and their association with instrumental
progress. Journal of Intellectual & Developmental Disability, 32, aggression. Journal of Personality Assessment, 87, 74–83. doi:
125–133. doi:10.1080/13668250701383116. 10.1207/s15327752jpa8701_06.
Nolan, K. A., Czobor, P., Roy, B. B., Platt, M. M., Shope, C. B., Vitacco, M. J., Neumann, C. S., & Jackson, R. (2005). Development
Citrome, L. L., et al. (2003). Characteristics of assaultive behavior of a four-factor model of psychopathy: Associations with
among psychiatric inpatients. Psychiatric Services (Washington, ethnicity, gender, violence, and intelligence. Journal of Con-
D.C.), 54, 1012–1016. doi:10.1176/appi.ps.54.7.1012. sulting and Clinical Psychology, 73, 466–476. doi:
Nolan, K. A., Volavka, J., Czobor, P., Sheitman, B., Lindenmayer, J., 10.1037/0022-006X.73.3.466.
Citrome, L. L., et al. (2005). Aggression and psychopathology in Waldheter, E. J., Jones, N. T., Johnson, E. R., & Penn, D. L. (2005).
treatment-resistant inpatients with schizophrenia and schizoaf- Utility of social cognition and insight in the prediction of
fective disorder. Journal of Psychiatric Research, 39, 109–115. inpatient violence among individuals with a severe mental
doi:10.1016/j.jpsychires.2004.04.010. illness. The Journal of Nervous and Mental Disease, 193, 609–
Novaco, R. W. (1994). Anger as a risk factor for violence among the 618. doi:10.1097/01.nmd.0000177788.25357.de.
mentally disordered. In J. Monahan & H. J. Steadman (Eds.), Walters, G. D., Gray, N. S., Jackson, R. L., Sewell, K. W., Rogers, R.,
Violence and mental disorder: Developments in risk assessment Taylor, J., et al. (2007). A taxometric analysis of the Psychop-
(pp. 21–59). Chicago: University of Chicago Press. athy Checklist: Screening Version (PCL:SV): Further evidence
Novaco, R. W., & Taylor, J. L. (2004). Assessment of anger and aggression of dimensionality. Psychological Assessment, 19, 330–339. doi:
in male offenders with developmental disabilities. Psychological 10.1037/1040-3590.19.3.330.
Assessment, 16, 42–50. doi:10.1037/1040-3590.16.1.42. Wang, E. W., & Diamond, P. M. (1999). Empirically identifying
Olver, M. E., Wong, S. C. P., Nicholaichuk, T., & Gordon, A. (2007). factors related to violence risk in corrections. Behavioral
The validity and reliability of the Violence Risk Scale–Sexual Sciences & the Law, 17, 377–389. doi :10.1002/(SICI)1099-
Offender version: Assessing sex offender risk and evaluating 0798(199907/09)17:3\377::AID-BSL351[3.0.CO;2-M.
therapeutic change. Psychological Assessment, 19, 318–329. doi: Ward, T., & Eccleston, L. (2004). Risk, responsivity, and the
10.1037/1040-3590.19.3.318. treatment of offenders: Introduction to the special issue.
Overall, J. E., & Gorham, D. R. (1962). The Brief Psychiatric Rating Psychology, Crime & Law, 10, 223–227. doi:10.1080/106831
Scale. Psychological Reports, 10, 799–812. 60410001662726.

123
Law Hum Behav (2009) 33:308–319 319

Wollert, R. (2006). Low base rates limit expert certainty when current Woodworth, M., & Porter, S. (2002). In cold blood: Characteristics of
actuarials are used to identify sexually violent predators: An criminal homicides as a function of psychopathy. Journal of
application of Bayes’s theorem. Psychology, Public Policy, and Abnormal Psychology, 111, 436–445. doi:10.1037/0021-843X.
Law, 12, 56–85. doi:10.1037/1076-8971.12.1.56. 111.3.436.
Wong, S., & Gordon, A. (2006). The Validity and Reliability of the Yudofsky, S. C., Silver, J., Jackson, W., Endicott, J., & William, D.
Violence Risk Scale: A treatment-friendly violence risk assess- (1986). The Overt Aggression Scale for the objective rating of
ment tool. Psychology, Public Policy, and Law, 12, 279–309. verbal and physical aggression. The American Journal of
doi:10.1037/1076-8971.12.3.279. Psychiatry, 143, 35–39.
Wong, S., & Hare, R. D. (2004). Guidelines for a psychopathy
treatment program. Toronto, ON: Multi-Health Services.

123
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.