You are on page 1of 16

Aggression and Violent Behavior xxx (xxxx) xxxxxx

Contents lists available at ScienceDirect

Aggression and Violent Behavior


journal homepage: www.elsevier.com/locate/aggviobeh

Facial aect processing in incarcerated violent males: A systematic review



Harriet Chapmana, , Steven M. Gillespieb, Ian J. Mitchella
a
Centre for Forensic and Criminological Psychology, School of Psychology, University of Birmingham, Birmingham, UK
b
Department of Psychological Sciences, Institute of Psychology, Health and Society, University of Liverpool, UK

A R T I C L E I N F O A B S T R A C T

Keywords: Previous reviews exploring facial aect processing among forensic samples have focused on the presence of
Facial aect processing psychopathy and/or have not distinguished on the basis of oence type. In order to develop understandings
Emotion about etiological processes implicated in dierent types of antisocial behavior, the principle aim of this review
Expression was to systematically explore facial aect processing in incarcerated violent oenders, relative to other non-
Violent
violent oenders, sexual oenders, and non-oenders. Following a systematic search of electronic databases and
Oenders
subsequent manual search, eight studies were assessed as meeting inclusion criteria, of which seven obtained a
quality score deemed acceptable for review. These studies examined recognition accuracy, sensitivity and re-
sponse bias for seven emotion categories (including neutral) in incarcerated male oenders with a history of
violence. Findings supported the presence of generally impaired facial aect processing among violent oenders,
including decits in fear, anger, and disgust. Overall the ndings of the review did not support the presence of a
hostile attribution bias among violent oenders. The review also highlights dierences in sample composition,
stimuli, and study designs in emotion recognition research. Recommendations are made for future work on facial
aect processing in clinically relevant groups.

1. Introduction (Gillespie, Rotshtein, Satherley et al., 2015; Loney, Frick, Clements,


Ellis, & Kerlin, 2003). Blair (2001) postulated that aggressive behavior
Evidence suggests that there are six basic emotions that are uni- in antisocial populations may be related to problems in identifying and
versally recognized across cultures: anger, disgust, fear, happiness, responding to social cues, particularly distress cues, such as fear and
sadness and surprise (Ekman, 1972, 1992a, 1992b, 1993; Ekman & sadness. In particular, it is believed that accurate decoding of distress
Friesen, 1971; Elfenbein & Ambady, 2002). According to Keltner (2003) cues is required for evoking aective responses in the decoder such as
emotional facial expressions play a pivotal role in the formation and empathy and remorse that serve to mitigate the likelihood of ag-
regulation of relationships; they provide information about the emo- gression against the sender (Blair, 2001; Marsh & Ambady, 2007; Marsh
tions and motives of the sender, they provoke a response in the receiver, & Blair, 2008). Indeed, in their meta-analysis exploring facial aect
and they provide motivation for desired social behavior (Keltner, recognition in antisocial populations, Marsh and Blair (2008) found
2003). It is perhaps not surprising, then, that breakdowns in social and that individuals who show instrumental aggression have specic de-
emotional responding occur when individuals are impaired in re- cits pertaining to the recognition of fearful, sad, and surprised expres-
cognising others' facial displays of emotion (Gillespie, Rotshtein, sions. Moreover, the impairment in fear recognition was signicantly
Satherley, Beech, & Mitchell, 2015). Indeed, impairments in decoding worse than impairments for sad and surprised expressions. However, it
socio-emotional information, including facial aect, have been ob- is important to note that although impairments in recognising others
served in a number of clinical conditions including anxiety disorder fear have been reported in relation to instrumentally aggressive popu-
(Demenescu, Kortekaas, den Boer, & Aleman, 2010; Easter et al., 2005), lations, these decits are not necessarily indicative of decits in the
attention-decit hyperactivity disorder (Rapport, Friedman, Tzelepis, & subjective experience of fear (Hoppenbrouwers, Bulten, & Brazil, 2016).
Van Voorhis, 2002; Singh et al., 1998), autism (Gross, 2004), depres- Antisocial behavior, broadly dened, covers all behaviors that vio-
sion (Demenescu et al., 2010; Surguladze et al., 2004) and schizo- late social norms and the rights of others (Burt, Mikolajewski, & Larson,
phrenia (Kohler & Brennan, 2004; Trmeau, 2006). 2009; Schnenberg, Mayer, Christian, Louis, & Jusyte, 2015). It in-
Socio-cognitive impairments have also been observed in antisocial cludes aggressive, criminal, and externalising behaviors, and abusive
populations, who exhibit diculties responding to social rules conduct (Marsh & Blair, 2008), and incorporates aggressive and forceful


Corresponding author at: School of Psychology, University of Birmingham, Birmingham, UK.
E-mail address: Harriet-chapman@hotmail.co.uk (H. Chapman).

https://doi.org/10.1016/j.avb.2017.10.006
Received 12 December 2016; Received in revised form 10 October 2017; Accepted 12 October 2017
1359-1789/ 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).

Please cite this article as: Chapman, H., Aggression and Violent Behavior (2017), https://doi.org/10.1016/j.avb.2017.10.006
H. Chapman et al. Aggression and Violent Behavior xxx (xxxx) xxxxxx

contact with a victim (i.e., violent behavior), as well as behaviors that compared to populations of non-violent oenders and/or non-oen-
do not involve such contact. Given the breath of this denition, it is ders. The review also explores how the study of sexual oenders has
perhaps unsurprising that the concept of antisociality appears to consist aected study outcomes.
of at least two distinct and only modestly correlated dimensions: an
aggressive subtype and a rule-breaking subtype (Burt, 2009, 2012; Burt 1.2. Existing reviews
& Neiderhiser, 2009; Tackett, Krueger, Sawyer, & Graetz, 2003). It is
therefore reasonable to propose that there may be fundamentally dif- A scoping exercise to identify the likely volume of studies to be
ferent cognitive mechanisms mediating these dierent subtypes of an- reviewed and any existing reviews was carried out in July 2015. The
tisocial behavior (Gillespie, Rotshtein, Satherley et al., 2015). Indeed, search was conducted using the Cochrane Library, EMBASE, MEDLINE,
the relative inuence of dierent etiological factors diers depending PsycINFO and Web of Knowledge. Over 40 reviews of facial emotion
on the subtype of antisocial behavior (Leist & Dadds, 2009), and dif- recognition were identied. The majority of these papers reviewed
ferent etiological factors have been found to be implicated in violent emotion recognition in neuropsychiatric conditions (n = 33); four pa-
and non-violent behavior. For example, risk factors for violence include pers reviewed abilities in individuals with cognitive impairment
the presence of violent attitudes and aective instability, while these (McCade, Savage, & Naismith, 2012; Moore, 1990; Rojahn, Lederer, &
factors are of lesser importance in predicting non-violent oences. With Tass, 1995; Zaja & Rojahn, 2008); two reviewed abilities in Borderline
this in mind, more recent research has explored facial aect recognition Personality Disorder (BPD) (Domes, Schulze, & Herpertz, 2009;
decits associated with specic types of antisocial behavior. This re- Mitchell, Dickens, & Picchioni, 2014); one reviewed alcoholic patients
search has provided insight into whether, and indeed how, patterns of (Fortunata & de Lima Osrio, 2014); and one reviewed abilities in
socio-cognitive impairment dier between violent and non-violent de- maltreated children (da Silva Ferreira, Crippa, & de Lima Osrio, 2014).
linquency. Of particular relevance to the current review were ve papers that re-
A vast amount of this research has utilized prison samples, due to viewed facial aect processing in antisocial populations, and included
the accessibility of individuals demonstrating violent and non-violent samples of oenders. Two of these reviews, of which one was a meta-
antisocial behavior within incarcerated populations. However, metho- analysis, looked at facial aect processing in antisocial and aggressive
dological variation makes comparisons across studies dicult. For ex- populations more generally (Marsh & Blair, 2008; Mellentin,
ample, some studies have examined violent oenders relative to non- Dervisevic, Stenager, Pilegaard, & Kirk, 2015), while three reviews, of
violent oenders and others relative to non-oenders. For studies that which two were meta-analyses, looked specically at psychopathic
make use of the latter design, it is unclear whether the observed im- populations (Brook, Brieman, & Kosson, 2013; Dawel et al., 2012;
pairments are specic to violent behavior or are associated with rule- Wilson et al., 2011).
breaking behavior more generally. Moreover, there is a lack of con-
sistency in methodological design with regard to the inclusion of sexual 1.2.1. Reviews exploring facial aect processing in antisocial populations
oenders, with some studies including sexual oenders in their sample The meta-analysis of Marsh and Blair (2008) looked at children and
of violent oenders, some studying sexual oenders as a separate adults with antisocial traits or behaviors and included participants
sample, and others altogether excluding sexual oenders from the displaying high levels of violence and/or aggression, of which an un-
sample. Given that a specic set of risk factors is implicated in sexual reported proportion was prisoners. They examined the evidence that
oending (such as self-regulation diculties, sexual preoccupation, and individuals with antisocial behavior showed decits in recognising each
deviant sexual preferences; Hanson & Morton-Bourgon, 2005; Mann, of the six basic expressions, whether the impairment is greatest for fear,
Hanson, & Thornton, 2010), it is reasonable to propose that dierences and whether fear decits are attributable to task diculty. The authors
could extend to socio-cognitive factors, and thus the inclusion and ex- carried out a comprehensive search to identify relevant studies
clusion of sexual oenders may make comparisons across studies pro- (n = 20). Methods included a search of PsycINFO and PubMed, and a
blematic. Indeed, research comparing samples of sexual oenders to search of reference lists, citation reports, and unpublished manuscripts.
other violent or non-violent oenders has indicated dierences in facial The authors concluded that antisocial populations exhibited signicant
aect recognition between these dierent types of oenders (Gillespie, decits in recognising fearful, sad, and surprised expressions, and that
Rotshtein, Satherley et al., 2015; Gery, Miljkovitch, Berthoz, & decits in recognising fear were signicantly greater than other im-
Soussignan, 2009; Hudson et al., 1993). pairments. Moreover, they found that this impairment was not attri-
butable to task diculty. It remains to be investigated to what extent
1.1. Current review their conclusions generalize to populations of violent prisoners speci-
cally, who arguably display more severe aggression and exhibit
While previous reviews and meta-analyses have explored decits in greater antisocial pathology than those continuing to reside in the
emotion recognition in antisocial populations and in relation to psy- community (Pascual-Leone, Bierman, Arnold, & Stasiak, 2011). Fur-
chopathic tendencies (e.g., Dawel, O'Kearney, McKone, & Palermo, thermore, Marsh and Blair (2008) analysed samples of children and
2012; Marsh & Blair, 2008; Wilson, Juodis, & Porter, 2011), including adults together. However, recent research suggests that the facial
samples of oenders, such reviews have not analysed results as a emotion recognition abilities of children dier to those of adults
function of oending status and/or have not distinguished on the basis (Leime, Rique Neto, Alves, & Torro-Alves, 2013), making it dicult to
of the type of oence. Collapsing across violent and non-violent of- draw conclusions about the relationship of antisocial behavior with
fenders precludes learning about dierences in etiological processes emotion recognition in exclusively adult samples. Additionally, Marsh
underlying these subtypes of antisocial behavior. A greater under- and Blair's meta-analysis, published in 2008, only included studies up to
standing of how patterns of socio-cognitive impairment dier among 2005, and many studies have been carried out in the eld since then.
subtypes of oenders could help to inform the development of inter- Mellentin et al. (2015) carried out a systematic review of 15 studies
vention modules that are tailored to the specic needs of dierent types to explore whether anger-prone and aggressive individuals show an
of oender. Indeed, if the behavioral dimensions of rule-breaking and anger bias when perceiving facial expressions in neuropsychological
violent behavior cannot be meaningfully distinguished in their inter- paradigms. Search strategy included the use of EMBASE, PubMed,
personal correlates, then delivering the same intervention protocol to PsycINFO, and Web of Science, as well as a search of references. The
individuals would oer both nancial and resource benets to treat- review included community, forensic and clinical samples of children
ment providers. and adults, and the authors found that anger-prone and aggressive in-
The present review attempts to facilitate understanding by assessing dividuals showed a bias toward perceiving anger and hostility in facial
the literature on facial aect processing in violent oenders as expressions.

2
H. Chapman et al. Aggression and Violent Behavior xxx (xxxx) xxxxxx

1.2.2. Reviews exploring facial aect processing in psychopathic 2. Method


populations
The meta-analysis of Wilson et al. (2011) examined the validity of 2.1. Sampling and search procedure
two competing hypotheses that had arisen from models of psychopathy:
that recognition decits are related to amygdala dysfunction (the in- A scoping search was conducted in order to establish the potential
tegrated emotion systems model) and that decits are related to the volume of publications relating to this topic. A thorough and systematic
verbal processing demands of the task (the left hemisphere activation search was then carried out in order to identify relevant studies. The
model). Twenty two studies exploring the relationship between facial following search methods were employed:
aect recognition and psychopathy were reviewed. The papers were 1. Search of electronic databases
identied through searches of PsycINFO, PubMed, Social Work Ab- A comprehensive search of electronic databases was undertaken in
stracts, and ProQuest databases, as well as hare.org/references and order to identify relevant publications. Four bibliographic databases
Google. The review found a small eect size for the relationship of were used:
psychopathy with emotion recognition impairments across all six basic
emotions, with the largest correlations observed for fear and sadness Ovid: PsycINFO (1967 to May Week 1 2016)
(r 0.10 and r = 0.12 respectively). The authors concluded that Ovid: EMBASE (1974 to May 6th 2016)
emotion recognition impairments in psychopathy are dynamic, and are Ovid: MEDLINE (1946 to April Week 4 2016)
dependent on the verbal processing requirements of the response. While ISI Web of Science (all years to May Week 1 2016)
the authors found that the association between psychopathy and re-
cognition decits was not moderated by oending status (i.e., forensic The Cochrane Library and Google Scholar (all years on 8th May
or community samples), it is not clear whether this would hold when 2016) were also searched in order to identify existing reviews in the
looking at specic subtypes of oenders. area.
The meta-analysis carried out by Dawel et al. (2012) included 26 A standardized search strategy was applied to search the databases,
studies evaluating the association between psychopathy and emotion although modications had to be made to meet the specic require-
recognition across visual and auditory modalities (vocal, facial and ments of each database, which therefore introduces some variation. The
postural), in forensic, clinical and community samples. Search strate- search was restricted to English language publications. Book chapters,
gies included the use of PsycINFO, PubMed and Web of Science, and a dissertations, editorials, and comment papers were excluded from the
hand search of reference lists. The authors found that psychopathy was review. Grey literature was not included to ensure that only peer re-
associated with decits for positive as well as negative emotions across viewed articles were analysed.
modalities. Specically, of the six emotion categories explored, decits Search terms:
were found for fear, happiness and surprise for facial and vocal ex- (fac* perception OR fac* expressions OR facial aect recognition OR
pressions, and sadness for facial expressions. Moreover, the authors facial aect decoding OR emotion recognition OR emotion categorisa-
found that, for children, the decit for fear was greater than the decit tion OR emotion processing OR facial aect processing OR emotional
for other emotions. Although this meta-analysis included participants displays OR social cognition OR aective processing)
from forensic settings, results were not analysed as a function of of- AND
fending status and therefore it does not aid understanding of emotion (violen* OR domestic violence OR intimate partner violence OR
recognition decits among violent oenders, and whether or how the violent crime OR perpetrators OR criminals OR prisoner OR oender
pattern of impairment diers from other types of oender. OR incarcerated).
Finally, Brook et al. (2013) reviewed research on emotion proces- Keywords and exploded search terms were used in order to increase
sing to explore whether psychopathy is associated with generalized the likelihood of identifying all of the relevant papers
emotion recognition decits or decits in the recognition of particular 2. Reference lists of reviews related to emotion in oenders
emotions. The authors searched PsycINFO and PubMed databases and 3. Reference lists of papers meeting inclusion criteria and their ci-
included 58 studies in the review. Of these, eleven studies examining tation reports.
emotion recognition in psychopathic oenders were reviewed sepa- 4. Hand-searching journals. Key journals were identied from the
rately and revealed mixed evidence for the specicity of emotion re- electronic database searches and searched for relevant articles. These
cognition decits in oenders with psychopathic features. Nonetheless, included: Journal of Psychiatric Research; International Journal of Law
given that the prevalence of psychopathy in prisoners across England and Psychiatry; Journal of Aggressive Behavior; European Archives of
and Wales has been found to be approximately 7% to 8% (Coid et al., Psychiatry and Clinical Neuroscience; Emotion.
2009), ndings from samples of psychopathic oenders cannot be re-
liably generalized to the vast majority of the prison population, and are 2.2. Study selection
therefore of limited utility for informing forensic practice more broadly.
Titles and abstracts for the identied papers were scanned in order
1.3. Aims and objectives of the current review to eliminate obviously irrelevant studies. Following removal of dupli-
cates, the remaining studies were made subject to the inclusion/ex-
This review aimed to systematically explore research that has ex- clusion criteria detailed below (Table 1). These criteria were informed
amined facial aect processing in violent oenders. Several questions by the initial scoping exercise.
were investigated in relation to this aim: The primary criterion for inclusion in the review was that the study
investigated facial processing (recognition accuracy, sensitivity or re-
1. Is there a consistent pattern of facial aect processing seen in violent sponse/attribution bias) in violent oenders. To ensure that partici-
oenders relative to non-oending control participants? pants' violence suciently deviated from behavior deemed as normal or
2. How do the facial aect processing abilities of violent oenders acceptable, samples of incarcerated oenders were chosen as the po-
compare to those of non-violent oenders? pulation. The violent oence could be current or historical.
3. Do patterns observed dier depending on the inclusion or exclusion Comparators could include a separate sample of sexual or IPV oenders
of sexual oenders? (when they were not included in the violent group), non-violent of-
fenders (such as theft, substance misuse, fraud), or non-oending
controls. Given evidence that diculties in recognising others' facial
expressions of emotion are found in neuropsychiatric conditions such as

3
H. Chapman et al. Aggression and Violent Behavior xxx (xxxx) xxxxxx

Table 1
Inclusion/exclusion criteria.

Inclusion Exclusion

Population Incarcerated violent oenders. Index oence or historical oence. Including Clinical samples of violent oenders (i.e., psychiatric patients). Juvenile/
intimate partner violence (IPV) and sexual violence. Adult males (aged 18 and adolescent oenders (under 18 years). Samples of antisocial oenders
over). whereby oence not specied or analysis as a function of nature of oending
not carried out. Community samples self-reporting violence/ non-convicted
individuals. Females.
Intervention Facial aect processing task using static displays of any combination of the six Other measures of emotional processing not including facial aect e.g., Stroop
basic emotions. Pure emotion or morphed developed using validated and normed tasks. Non-static displays. Studies including context e.g., story or aective
stimuli. stimuli other than face e.g., body posture.
Comparator Non-violent adult male oenders and/or samples of IPV oenders and/or samples Studies that do not distinguish between violent and non-violent oenders.
of sexual oenders, and/or non-oending controls.
Outcome Accuracy of facial expression categorisation or sensitivity to discriminating Studies that ask participants to infer anything other than emotion from facial
emotional expression. Response bias to ambiguous or neutral expressions. aect slides (e.g., thoughts).
Measured via forced-choice or free-response format. Objective criterion for
recognition parameter.
Study type Quasi-experimental Other

schizophrenia (e.g., Trmeau, 2006) and anxiety disorder (e.g., Clear description of measures of predictor variables
Demenescu et al., 2010; Easter et al., 2005), among others, studies were Clear description of outcome measures
excluded if they specically examined clinical populations, in order to
reduce the chance of spurious relationships. Only studies analysing Studies that did not meet these two criteria would not have been
male samples were included due to evidence of sex dierences in facial subsequently subjected to the quality assessment form.
emotion recognition (Robinson et al., 2012). Furthermore, as research 2. Quality assessment forms
has reported dierences in emotion recognition abilities between child, A quality assessment form comprising 20 questions relating to
adult, and elderly samples (Leime et al., 2013; Sullivan, Ruman, & methodological quality was used. The form allowed study biases re-
Hutton, 2007), only adult (18 +) populations were studied in order to lating to selection/sampling, performance, measurement and analysis
ensure maturation of socio-cognitive development. to be identied and assessed in a structured way. Each item pertaining
Only studies examining one or more of the six basic emotions of to these factors was scored on a three-point Likert-scale. This allowed
anger, disgust, fear, happiness, sadness, and surprise were considered for an overall quality score to be calculated. When the item was not
for review. These emotions were selected for consistency with previous applicable it was omitted. Likewise, when there was unclear or in-
reviews and due to evidence of their being universally recognized (e.g., sucient information that could not be claried by authors, the item
Ekman & Friesen, 1971; Elfenbein & Ambady, 2002). This meant that it was omitted.
was not considered necessary to limit investigations to those carried out The scoring system was as follows:
in Western society. Studies had to include a facial aect categorisation Condition not met (N) = 0
task to be included in the review. The stimuli had to be developed from Partially met (P) = 1
appropriately normed or validated images so as to provide an objective Condition fully met (Y) = 2.
criterion for judgement. As the review was focused solely on the in- The overall quality score was calculated by adding all the scores
vestigation of facial expression of emotion, studies that presented af- together; yielding a maximum quality score of 40 if no items were
fective stimuli in addition to facial displays (such as postural or vocal omitted. Scores were converted into a percentage to enable clear
information) were excluded due to their providing accompanying comparison of quality between the studies. A minimum threshold of
emotional information. Similarly, studies utilising contextual informa- 60% quality was set for the study to be included in the review. This was
tion such as stories or simulated scenarios were excluded. Only static regarded as a reasonable threshold to ensure that only good quality
facial stimuli were reviewed (including morphed facial stimuli to allow studies were reviewed, while ensuring that a sucient number of stu-
for a measure of emotion recognition sensitivity and/or attribution dies were reviewed. An independent rater assessed 50% of the studies in
bias). Finally, only articles from peer-reviewed journals were included order to ensure that assessment scores were reliable. No dierence
to ensure a minimum threshold for quality, and non-English studies greater than two points was obtained and, in each case, did not aect
were excluded due to an inability to interpret data. whether the study met the threshold for inclusion.
Of the eight studies that met the inclusion criteria, seven were as-
sessed as being of high enough quality to include in the review. One
2.3. Data extraction study (Hudson et al., 1993) obtained a quality score of 57% and thus
was excluded from the review. The quality of the remaining studies
Information was extracted from the studies that satised the in- reviewed ranged from 65 to 78%. Studies obtaining a score of > 70%
clusion criteria using a pro forma to ensure systematic recording and were considered to be the methodologically stronger studies in the
reporting of information. Specically, information relating to popula- analysis. Fig. 1 depicts the process of the study selection and highlights
tion characteristics, methodology, outcome measures and analyses was the number of studies retained and excluded at each stage of the pro-
extracted and provided sucient information to inform the quality cess.
assessment process. Table 2 highlights key information from each study
reviewed.
3. Results

2.4. Quality assessment 3.1. Sample characterisation

The following steps were taken in order to assess the quality of the 3.1.1. Operationalizing violence
studies meeting the inclusion criteria: Violent oenders were operationalized and compared in the fol-
1. Threshold criteria lowing ways across the seven studies reviewed:

4
Table 2
Characteristics of reviewed studies.

Study and quality score Participants Aim(s) Task variables and recognition parameter(s) Results
studied
H. Chapman et al.

Gillespie, Rotshtein, Sexual oenders (n = 13) and violent non-sex To examine emotion recognition accuracy for the Stimuli: taken from the NimStim Face Stimulus Method of analysis
Satherley et al. oenders (n = 16) recruited from a Therapeutic six basic emotions among sexual and violent Set (Tottenham et al., 2009). Sensitivity was calculated as the discriminability
(2015) Community in England UK. Males aged 2458 years oenders compared with healthy controls. Gender of models: ve male and ve female index (i.e., the dierence between the hit rate and
65% old (mean age of sex oender sample = 50.5, mean To examine the eect of intensity of expression and Emotions studied: happy, sad, angry, fear, false alarm rate). Response bias was calculated as
age of violent oenders = 37.8). sex of model on recognition surprise, disgust, neutral the criterion (with lower values indicating a more
Male non-oending community controls (n = 19). Intensity of expression: low intensity (10% liberal response style). Responses were analysed
Aged 2667 years old (mean age = 48.2). expressive, 90% neutral), moderate intensity using a mixed-model ANOVA. Signicant
Matched on: gender (55% expressive), and high intensity (90% interactions were broken down with further
expressive). ANOVAs.
Number of slides: 180 trials depicting a dierent Sensitivity
stimulus varying in model (10), expression (6) Non-oenders showed greater sensitivity to fearful
and intensity (3). expressions compared to sexual (p = 0.007) and
Presentation delivery: randomized violent (p = 0.015) oenders. Non-oenders also
Response format: face remained on screen until showed greater sensitivity to disgust expressions
participant chose from seven options (six compared to sexual oenders (p = 0.009). The
emotions plus neutral) which emotion was eect of group was non-signicant for angry,
depicted. happy, sad, and surprised expressions (all
Recognition parameters assessed: sensitivity and p > 0.20).
response bias There was a signicant interaction of group and
expression for female faces at a high intensity
(p = 0.031). In particular, sexual oenders showed
reduced sensitivity to female angry expressions
compared to non-oenders (p = 0.014) and violent
oenders (p = 0.021). Non-oenders were more

5
sensitive to female expressions of disgust compared
with sexual oenders (p = 0.005), and were more
sensitive to female expressions of fear compared to
both sexual (p = 0.029) and violent (p = 0.014)
oenders.
Response bias
For moderate intensity male faces, violent oenders
showed a more conservative response style for
labelling faces as disgust or fear relative to angry,
happy or sad (all p < 0.01). For high intensity
male faces, violent oenders showed a more
conservative response style for fear relative to all
other emotions (p < 0.01), together with a lower
criterion for labelling faces as sad compared to
disgust and surprise (p < 0.05).
Hoaken, Allaby, and 20 incarcerated violent* oenders (mean age 34.7) To investigate whether the relationship between Stimuli: taken from Ekman's faces of emotional Method of analysis
Earle (2007) 20 incarcerated non-violent oenders (mean age executive cognitive functioning and aggression aect (Ekman & Friesen, 1976). For accuracy, a one-way ANOVA was computed on
70% 32.9) from a medium security institution in Canada. may be due to impairments in the encoding and Gender: not specied the number of incorrect responses across all trials.
20 university undergraduates and community interpretation levels of social information Emotion: happiness, surprise, sadness, fear, For response bias, a count of the emotions each
controls (mean age 25.2). processing. anger, disgust, and neutral participant attributed to the neutral face was
Matched on: gender Intensity: 100% entered into mixed-model ANOVA to assess for
*Denition of violent oender based on Harris, Rice, Number of slides: 102 group dierences. ANOVAs used for all post-hoc
and Cormier's (2002) denition, i.e., any individual Presentation delivery/response format: analyses.
who was incarcerated for any criminal charge for a Randomized. Face presented for 2000 ms, after Accuracy
violent oence against persons- e.g., assault, assault which participants had to rate which emotion The violent group made a greater number of errors
causing bodily harm, wounding, attempted homicide, was depicted from six options (neutral was not than the non-violent or non-oending groups
homicide, kidnapping, forcible connement, armed an option). (p < 0.001), who did not dier from each other.
robbery and all hands-on sexual oences (p. 383). Recognition parameters assessed: accuracy and However, an analysis of errors as a function of
Non-violent oenders were those without a history of response bias emotion was not conducted.
(continued on next page)
Aggression and Violent Behavior xxx (xxxx) xxxxxx
Table 2 (continued)

Study and quality score Participants Aim(s) Task variables and recognition parameter(s) Results
studied
H. Chapman et al.

the above (thus including non-contact sexual Response bias


oenders) The groups diered in how frequently they labelled
the face as sadness (p < 0.05) and disgust
(p < 0.05). Violent oenders were less likely to
interpret a neutral face as sadness than were non-
oending controls (p < 0.05) and were more
likely to interpret it as disgust than were non-
violent oenders (p < 0.05).
Robinson et al. (2012) Convicted prisoners in Scotland, recruited as part of To explore whether prisoners show an antisocial Stimuli: taken from the Ekman and Friesen Method of analysis
70% an investigation which examined a screening tool for pattern of decits in decoding emotional (1976) stimulus set. Mean dierences between oender and non-
autistic characteristics. 116 prisoners (mean expressions relative to controls. Gender: not specied but 10 models oender groups analysed using t-tests. Repeated
age = 35.6) who either were most likely to have high To examine whether social cognition is related to Emotions: happy, sad, anger, fear, surprise and measures ANOVAs were used to examine
levels of autistic traits or who evidently did not have markers of antisociality and oence history. disgust. recognition accuracy between groups, with the
high levels, were examined in greater detail. Intensity: 100% eect of group explored further using one-way
Participants' were categorized taking into Number of slides: 60 (one expression per emotion ANOVAs.
consideration previous convictions. Two groups: for each model). Accuracy
those who had ever committed violent oences Presentation delivery: randomized. Stimuli shown There was no signicant dierence in emotion
(including sexual oences) and non-violent oences. for 5 s. recognition between or within oenders with a
One participant had dissociative symptoms and one Response format: the names of the six emotions violent conviction and oenders without (p-value
had features suggestive of an organic brain syndrome. were shown on the screen and participants asked not reported).
A community control group (n = 130, mean to select which best described the expression just Prisoners with a sexual index oence were better at
age = 37.2). shown. recognising sadness (p = 0.046) and worse at
Groups matched for age, sex and IQ. Recognition parameter assessed: accuracy recognising surprise (p = 0.006) in comparison to
other prisoners (both violent and non-violent).
Comparison with non-oending controls

6
Prisoners were signicantly less accurate in
recognising sadness, anger, fear (all p < 0.001)
and disgust (p < 0.05) in comparison with
controls.
N.B. Eects held when levels of autistic traits
controlled for.
Schnenberg et al. 44 antisocial violent oenders (mean age = 35.32) To examine facial recognition impairment in Stimuli: digitized colour photographs chosen Method of analysis
(2014) recruited from a German correctional facility. antisocial violent oenders with psychopathic from the Radboud Faces Database (Langner Intensity of emotional expression at time of button
75% Excluded oenders charged with intimate partner personality traits by assessing perceptual et al., 2010) based on accuracy of expression. press for correct responses analysed using a
violence, sexual assault and drug-related crime sensitivity to emotional expressions. Gender: three male models repeated measures ANCOVA with age as covariate.
(reason not provided). None had a history of Emotion: angry, happy, fearful, sad, surprised, Eect of group followed up using t-tests. Repeated-
schizophrenia or mental retardation. disgusted as well as neutral measures ANOVA conducted to examine speed/
43 community controls (mean age = 29.34). Preparation of stimuli: each emotional expression accuracy trade o using analysis of error rates.
Matched for education and gender. depicted by every model was morphed with a Sensitivity
neutral expression in increments of 2%. This Violent oenders exhibited signicantly impaired
produced 51 intensity levels ranging from 0% recognition of fearful (p < 0.01) and surprised
(neutral) to 100% (full emotion). (p = 0.01) expressions relative to non-oending
Number of slides: 72 controls, i.e., they required greater emotional
Delivery: sequence displayed consecutively intensity to correctly detect the emotional
Response: participants had to press a button as expressions.
soon as they were able to identify the emerging Accuracy
expression. The stimulus was then removed and There was a signicant main eect of emotion only
participant had to indicate the emotion that they (p < 0.001), with both groups making more errors
saw in a multiple choice manner. for fearful, disgusted and surprised expressions.
Recognition parameters assessed: sensitivity and
accuracy (the latter to rule out dierential
speed/accuracy trade-os)
Schnenberg and 55 antisocial violent incarcerated oenders, recruited To explore a hostile response bias by assessing Stimuli: selected from the Radboud Faces Method of analysis
Jusyte (2014) 75% from a German correctional facility (mean response styles to ambiguous facial cues in Database (Langner et al., 2010). For response bias, a series of 5 (intensity) 2
(continued on next page)
Aggression and Violent Behavior xxx (xxxx) xxxxxx
Table 2 (continued)

Study and quality score Participants Aim(s) Task variables and recognition parameter(s) Results
studied
H. Chapman et al.

age = 33.35). Exclusion criteria were drug-related antisocial violent oenders compared to matched Gender: three male models (group) 3 (dimension) repeated measures
crime, IPV or sexual assault. All oenders lled the controls. Emotion: angry, happy, fearful ANOVAs were conducted. Signicant eects on
criteria for Antisocial Personality Disorder (ASPD). Preparation of stimuli: faces were morphed with group were followed up with t-tests. Independent t-
Four were also diagnosed with major depression and each other to create three continuous dimensions tests were computed for intensity ratings for the
with dysthymia. None had a history of schizophrenia, (happy-fearful, happy-angry and fearful-angry). three dimensions at 50:50 ratio (i.e., most
BPD, or mental retardation. Each dimension had ve distinct intensity levels ambiguous expression).
55 healthy controls (mean age = 30.38) recruited containing dierent amounts of each blended Response bias
from local vocational schools. Matched for education emotion (intensity rations: 90:10, 70:30, 50:50, A signicant eect of intensity emerged for the
status and age. Control participants did not have a 30:70, and 10:90). happy-fearful dimension (p < 0.001), but neither
history of psychiatric morbidity. Number of slides: 45 group nor the intensity x group interaction reached
Presentation delivery: images were repeated four signicance.
times. Pseudo- randomized across emotions and For the angry-fearful dimension, violent oenders
intensity levels. Presented for 500 ms. made signicantly more angry responses under
Response format: forced-choice. Open-ended time conditions of maximal ambiguity (50:50) than did
frame. non-oenders (p < 0.01)
Participants then rated the intensity level of the For the angry- happy dimension, violent oenders
emotion identied on a scale ranging from 0 (not responded with angry under conditions of
at all present) to 10 (full blown emotion). maximal and high ambiguity (50:50 and 30%
Recognition parameters assessed: response bias angry: 70% happy) (p < 0.05).
Perceived intensity ratings
Violent oenders rated the perceived intensity of
anger in ambiguous angry-happy and angry-fearful
faces signicantly higher than non-oenders (p < .
05 and p < 0.01, respectively). No signicant
group dierence was found for the happy-fearful

7
dimension (p > 0.1).
Schnenberg, Louis, 32 prisoners convicted for repeated grievous bodily To examine identication of threat-related facial Stimuli: digitized colour photographs chosen Method of analysis
Mayer, and Jusyte harm. All met criteria for ASPD. Did not include expressions in violent oenders with ASPD. from the Karolinska Directed Emotional Faces Intensity of correct detection of emotional
(2013) individuals charged with IPV, sexual oences, or database (Lundqvist, Flykt, & hman, 1998). expressions at time of button press analysed using a
78% drug-related oences. Exclusion criteria also included Gender: two male models. repeated measures ANOVA followed by separately
oenders with BPD or schizophrenia. Emotions: angry, happy, fearful, neutral computed t-tests to explore signicant eects
32 controls without a history of oending or Preparation of stimuli: morphing procedure used involving group. Additional analysis using error
psychopathology. to produce stimuli increasing in intensity by rates were used to examine dierential speed/
Matched on age and education. increments of 2%. This produced 51 intensity accuracy trade-os.
levels ranging from 0% (neutral) to 100% (full Sensitivity
emotion). Violent oenders required signicantly higher
Number of slides: Maximum of 2130 (30 intensity levels to detect angry expressions than did
sequences of two models depicting three controls (p = 0.014, n2 = 0.10). There was a trend
emotions at 51 intensity levels plus ve repeated toward violent oenders requiring higher
morphs). emotional intensities for identifying fear
Presentation/response format: each image expressions, although this did not reach statistical
presented for 500 ms, beginning with 0% and signicance (p = 0.068).
progressing successively to 100%. Participants Groups did not dier in their sensitivity to happy
pressed button as soon as they were able to expressions (p = 0.150).
identify the emerging emotion. Sequence then Accuracy
terminated and participant required to indicate Groups did not dier in their recognition accuracy
which emotion they detected (response options of angry, happy and fearful expressions
provided not specied). (p = 0.415).
Recognition parameter assessed: sensitivity and
accuracy (the latter to examine dierential
speed/accuracy trade-os)
Seidel et al. (2013) 30 incarcerated violent oenders (mean age 35.6) To test the three stage model of empathy in violent Stimuli: colour photographs of Caucasian faces Method of analysis
75% and 30 non-oenders (mean age 34.8) matched for oenders compared to matched controls (i.e., were taken from a standardized stimulus set (Gur Accuracy data were analysed using repeated-
age, sex (males), education and intelligence. et al., 2002). measures ANOVAs. Signicant eects on group
(continued on next page)
Aggression and Violent Behavior xxx (xxxx) xxxxxx
H. Chapman et al. Aggression and Violent Behavior xxx (xxxx) xxxxxx

1. Violent oenders (including sex oenders) compared to non-violent

A signicant eect of group was found (p = 0.026)

controls. A signicant emotion by group interaction


was found (p = 0.049). Post-hoc tests revealed that
sexually-violent oenders and non-sexually violent

(p = 0.001). There were no signicant dierences


There were no signicant dierences for accuracy

with oenders impaired relative to non-oending

controls outperformed oenders for disgust only


oenders and non-oenders (Hoaken et al., 2007; Robinson et al.,

when the violent group was divided up into


2012)
2. Violent oenders (including sex oenders) compared to non-oen-

between groups for all other emotions


were explored using a series of t-tests.

ders (Seidel et al., 2013)


3. Violent oenders (excluding sex oenders) compared to non-oen-
ders (Schnenberg et al., 2013; Schnenberg et al., 2014;
Schnenberg & Jusyte, 2014).
oenders (p > 0.177).

4. Violent oenders compared to sexually-violent oenders and non-


oenders (Gillespie, Rotshtein, Satherley et al., 2015).

(p > 0.164).
3.1.2. Dening violent oender
Accuracy

In the majority of studies, oenders were classied into relevant


Results

participant groups (i.e., violent or non-violent) on the basis of the


nature of their index oence. However, two studies looked at previous
anger, fear, disgust and neutral) until a response
Emotions: happiness, sadness, anger, fear, disgust

convictions in assigning participants to oender groups (Hoaken et al.,


Gender: gender-balanced but unknown number

presented for 5 s and remained on screen with


Task variables and recognition parameter(s)

2007; Robinson et al., 2012).


six response categories (happiness, sadness,
Presentation/response format: stimuli were

Hoaken et al. (2007) categorized participants into groups based on


Recognition parameter assessed: accuracy

the Harris et al. (2002) denition of a violent oence (see Table 2).
Other studies did not specify how violence was dened, although
Gillespie, Rotshtein, Satherley et al. (2015) provided examples of some
violent oences included (murder and wounding with intent to do
grievous bodily harm) and Schnenberg et al. (2014) and Schnenberg
Number of slides: 36

and Jusyte (2014) specied oences that were excluded (reported


Intensity: 100%

below). Finally, Schnenberg et al. (2013) specically sampled violent


and neutral.

oenders who had committed repeated grievous bodily harm.


was given.
of models
studied

3.1.3. Control groups and sample characteristics


Note: The samples reported in Schnenberg et al. (2014), Schnenberg and Jusyte (2014), and Schnenberg et al. (2013) did not overlap.

All seven studies compared violent oenders to a non-oending


control group. These were recruited from the community, including
undergraduate students (Hoaken et al., 2007). Three studies matched
emotion recognition, perspective taking and

experimental and control groups for age and education level


(Schnenberg et al., 2013; Schnenberg et al., 2014; Schnenberg &
Jusyte, 2014), one study matched groups on age and intelligence
(Robinson et al., 2012), and another also matched on education (Seidel
et al., 2013). Two studies did not match groups on these variables
aective responsiveness).

(Gillespie, Rotshtein, Satherley et al., 2015; Hoaken et al., 2007). All


studies compared male samples in analyses. Sample sizes of violent
oenders ranged from 16 (Gillespie, Rotshtein, Satherley et al., 2015)
to 87 (Robinson et al., 2012), with ages of oenders ranging from 18 to
62.
Aim(s)

Studies varied in the sampling of psychiatric diagnoses.


Schnenberg et al. (2014) only sampled oenders with psychopathic
tendencies. In the violent sample of Schnenberg et al. (2013) and
disorders. Some had a history of alcohol (n = 10) or

Schnenberg and Jusyte (2014), all participants lled criteria for An-
The non-oending control group were recruited by
drug (n = 3) dependence. Their mean PCL-R score

advertisements. They had no history of psychiatric


Most oenders (n = 22) had cluster B personality

illness, neurological illness, or substance abuse in

tisocial Personality Disorder (ASPD). In the former sample, two men


also met the criteria for historical major depressive disorder, while in
themselves and their rst degree relatives.

the latter sample, four men were diagnosed with major depression and
two met criteria for dysthymia. None had a history of schizophrenia
was 21.5 (medium to high scorers).

(Schnenberg et al., 2013; Schnenberg et al., 2014; Schnenberg &


Jusyte, 2014), mental retardation (Schnenberg et al., 2014;
Schnenberg & Jusyte, 2014), or BPD (Schnenberg et al., 2013;
Schnenberg & Jusyte, 2014). In the oender sample of Seidel et al.
(2013), 13 participants had a history of alcohol/drug dependence and
22 were diagnosed with Cluster B personality disorders. In addition, the
Participants

sample scored medium-high on the PCL-R (mean = 21.5). In the of-


fender sample of Robinson et al. (2012), one participant had dis-
sociative symptoms and one had features suggestive of an organic brain
syndrome.
Study and quality score

In terms of control groups, the samples of Schnenberg et al. (2013)


and Schnenberg and Jusyte (2014) had no current or historical psy-
Table 2 (continued)

chiatric morbidity, and the sample of Seidel et al. (2013) did not have a
history of psychiatric/neurological illness or substance abuse. Neither
Gillespie, Rotshtein, Satherley et al.'s (2015) nor Hoaken et al.'s (2007)
studies contained information pertaining to psychological disorders in
either experimental or control samples.

8
H. Chapman et al. Aggression and Violent Behavior xxx (xxxx) xxxxxx

Fig. 1. Process of study selection.


Total hits from electronic databases
n = 430

PsycInfo = 272
EMBASE = 84
MEDLINE = 64
Web of Science = 10

Duplicates excluded n = 63

Excluded following search


of abstracts n = 345

Could not obtain n = 1

Papers retrieved for detailed


assessment n = 21
Excluded n = 14

No non-sexual violent group


n=4

Studies identified from Additional contextual


reference lists n = 1 information n = 3

No measure of facial affect


processing n = 2

Insufficient detail pertaining


to offending status n = 2

No control group n = 1

Psychiatric population n = 1

Non-English paper n = 1
Studies meeting inclusion criteria
n=8

Excluded on the basis of


quality assessment n = 1

Studies included in review


n=7

3.1.4. Exclusion criteria an IQ < 70 and/or those scoring above the diagnostic cut o for Au-
Three studies excluded from their group of violent oenders in- tistic Spectrum Disorders (Robinson et al., 2012). Finally, one study
dividuals charged with domestic violence, sexual assault or drug-re- excluded those who were denying their oence or appealing their
lated crime (Schnenberg et al., 2013; Schnenberg et al., 2014; conviction (Hoaken et al., 2007).
Schnenberg & Jusyte, 2014), as well as individuals with inadequate
knowledge of the German language (Schnenberg et al., 2014;
Schnenberg & Jusyte, 2014). Another study excluded prisoners with

9
H. Chapman et al. Aggression and Violent Behavior xxx (xxxx) xxxxxx

3.2. Task characterisation expressions. Robinson et al. (2012) found that violent oenders were
signicantly less accurate in recognising sadness, anger, fear and
Two studies (Hoaken et al., 2007; Robinson et al., 2012) used the disgust in comparison to non-oenders, and Hoaken et al. (2007)
Pictures of Facial Aect stimulus set (Ekman & Friesen, 1976), two found that, overall, violent oenders made signicantly more errors
studies (Schnenberg et al., 2014; Schnenberg & Jusyte, 2014) used than did the non-oenders. However, an analysis of between group
stimuli from the Radboud Faces Database (Langner et al., 2010), one dierences across the six emotions was not conducted.
study (Schnenberg et al., 2013) selected stimuli from the Karolinska
Directed Emotional Faces database (Lundqvist et al., 1998), one study 3.4.1.2. Sensitivity. Gillespie, Rotshtein, Satherley et al., 2015 found
(Gillespie, Rotshtein, Satherley et al., 2015) used stimuli from NimStim that violent oenders were signicantly less sensitive to discriminating
Face Stimulus Set (Tottenham et al., 2009), and one study (Seidel et al., fearful expressions compared to non-oenders. Moreover, the violent
2013) used a stimulus set developed and validated by Gur et al. (2002). oenders were signicantly less sensitive to female fearful expressions
Three of these studies used male models only (Schnenberg et al., 2013; at high intensities compared to non-oenders. Finally, sexually-violent
Schnenberg et al., 2014; Schnenberg & Jusyte, 2014) and two did not oenders showed signicantly reduced sensitivity to disgust
specify the gender of the models (Hoaken et al., 2007; Robinson et al., expressions, and to female angry and disgust expressions at high
2012). intensities, relative to non-oenders (Gillespie, Rotshtein, Satherley
Two studies investigated the six basic emotions plus a neutral ex- et al., 2015).
pression (Hoaken et al., 2007; Schnenberg et al., 2014), two studies Using male models only, Schnenberg et al. (2013) found that a
investigated the six basic emotions (Gillespie, Rotshtein, Satherley sample of non-sexually violent oenders required signicantly higher
et al., 2015; Robinson et al., 2012), one study investigated ve emo- levels of intensity to detect anger in comparison to non-oenders, and
tions (excluding surprise) plus a neutral expression (Seidel et al., 2013), there was a trend toward the same for the identication of fear ex-
and two studies investigated anger, happiness and fear (Schnenberg pressions. The two groups did not dier in their sensitivity to happy
et al., 2013; Schnenberg & Jusyte, 2014). expressions. Using a similar sample and procedure but examining sen-
All stimuli were presented supraliminary. Three studies displayed sitivity to six emotion categories, Schnenberg et al. (2014) found that
the stimulus for a limited duration (Hoaken et al., 2007; Robinson et al., violent oenders showed signicantly impaired sensitivity to fearful
2012; Schnenberg et al., 2013; Schnenberg & Jusyte, 2014). For the and surprised expressions compared to non-oenders.
remaining studies, the stimulus remained on screen until the participant
was either ready to make (Schnenberg et al., 2014) or made a response 3.4.1.3. Response bias. Hoaken et al. (2007) found that violent
(Gillespie, Rotshtein, Satherley et al., 2015; Seidel et al., 2013). Six oenders (including sexual oenders) were signicantly less likely to
studies used a forced-choice response format with an open-ended time interpret a neutral face as sad compared to non-oenders. Schnenberg
frame. The response format of Schnenberg et al. (2013) is not clear. and Jusyte (2014) found that violent oenders made signicantly more
angry responses to ambiguous displays containing anger in
3.3. Outcomes measured comparison to controls. Specically, violent oenders were more
likely to interpret anger under conditions of maximal ambiguity
Facial aect processing was measured using three parameters: (50:50) for an angry-fearful dimension, and under all conditions of
ambiguity for an angry-happy dimension (50:50 and 30:70).
1. Accuracy - measured by the number of correct responses (hits) to Furthermore, the violent oenders rated the perceived intensity of
emotional expressions (assessed by Hoaken et al., 2007; Robinson anger in these two dimensions signicantly greater than did non-
et al., 2012; Seidel et al., 2013). oending controls. No signicant dierence was found between groups
2. Sensitivity to emotional expression - including both perceptual sen- for a happy-fearful dimension.
sitivity (assessed by using morphing techniques to create emotional Analyses of the Criterion index, or how conservative participants
expressions of various intensities; Schnenberg et al., 2013; were in labelling faces as a particular emotion, by Gillespie, Rotshtein,
Schnenberg et al., 2014) and discriminability index, also known as Satherley et al., 2015 showed that, for moderate intensity male faces,
d (assessed by Gillespie, Rotshtein, Satherley et al., 2015 as the violent oenders were more conservative in labelling faces as disgust or
dierence between the normalized hit rate and false alarm rate for a fear relative to angry, happy, or sad. For high intensity male faces,
given expression). violent oenders showed a higher criterion for labelling fear relative to
3. Response bias - measured by responses made to ambiguous expres- all other emotions, but were more liberal when labelling faces as sad
sions (created using morphing techniques; Schnenberg & Jusyte, compared with disgust and surprise. The same interaction between
2014) or when forced to attribute emotion to a neutral expression emotion, sex and intensity of expression was not observed for either
(Hoaken et al., 2007). Response bias was also measured using a non-oenders or sex oenders.
measure of criterion to assess the extent to which a conservative or
liberal response style was adopted when labelling emotional faces 3.4.2. How do the facial aect processing abilities of violent oenders
(Gillespie, Rotshtein, Satherley et al., 2015). compare to those of non-violent oenders?
3.4.2.1. Accuracy. Hoaken et al. (2007) found that, overall, violent
3.4. Outcomes of systematic review oenders (inclusive of sex oenders) made signicantly more errors
than did non-violent oenders. However, the authors did not analyse
The results of the reviewed studies were examined to identify if the number of errors as a function of the expressed emotion. In contrast,
there is a consistent pattern of impairment in facial aect processing Robinson et al. (2012) did not nd a signicant dierence in emotion
associated with violent oending, and whether this is distinguishable recognition accuracy between oenders with a violent conviction
from that of non-violent oenders. The results are reported as a func- (including sexual) and those without.
tion of recognition parameter in answering the review questions.
3.4.2.2. Sensitivity. None of the studies reviewed compared sensitivity
3.4.1. Is there a consistent pattern of facial aect processing seen in violent between violent oenders and non-violent oenders.
oenders relative to non-oending control participants?
3.4.1.1. Accuracy. Seidel et al. (2013) found that violent oenders 3.4.2.3. Response bias. The only study to look at response bias among
showed decits in the recognition of disgust expressions only. However, samples of violent and non-violent oenders found that violent
it is noted that this study did not examine recognition for surprise oenders (including sexual oenders) were more likely to interpret a

10
H. Chapman et al. Aggression and Violent Behavior xxx (xxxx) xxxxxx

neutral face as disgust relative to non-violent oenders (Hoaken et al., In examining recognition accuracy, Seidel et al. (2013) did not nd
2007). a dierence between sexually-violent oenders and violent non-sex
oenders across the ve emotions studied (anger, disgust, fear, happi-
3.4.3. Do patterns observed dier depending on the inclusion or exclusion of ness and sadness).
sexual oenders?
3.4.3.1. Studies that included sex oenders in the violent oender 4. Discussion
sample. One study found that the violent/sexually-violent group was
less accurate at recognising disgust compared to non-oenders (Seidel This review set out to explore impairments in facial aect proces-
et al., 2013), while another found decits in disgust, sadness, anger and sing among violent oenders, and to examine whether the pattern of
fear in violent/sexually-violent oenders compared to non-oenders performance diers from that of non-violent oenders and non-oen-
(Robinson et al., 2012). Robinson et al. (2012) did not nd any ders. We also looked at whether the inclusion or exclusion of sexually-
dierences between the mixed violent oender group and the non- violent oenders aected the pattern of results observed in the violent
violent oenders. However, a third study found that, overall, the sample. In total we reviewed seven published articles of facial aect
violent/sexually-violent group made a greater number of errors than processing in violent oenders. Each of these studies reported some
both a non-violent and a non-oending group, but did not analyse form of recognition impairment among violent oenders relative to
results by emotion (Hoaken et al., 2007). Furthermore, Hoaken et al. non-oenders. While these impairments were limited to negative
(2007) found that the violent/sexually-violent oenders were more emotional expressions, the particular emotion/s that impairments were
likely to interpret a neutral face as disgust relative to non-violent observed for varied between studies. Overall, the studies reviewed
oenders, and were less likely to interpret a neutral face as sadness varied widely in the combination of comparator and outcome variables
relative to non-oenders. It is noted that, in this latter study, an used, making it dicult to examine trends across the various studies.
individual who had committed a non-contact sexual oence was Further still, the studies varied in their sampling of psychiatric diag-
analysed as a non-violent oender. Thus, the non-violent sample also noses, their matching of groups on demographic variables, and the
included non-contact sex oenders. stimulus set employed. With this in mind, the aggregation of these data
provides only a coarse estimate of recognition decits. Below we recap
3.4.3.2. Studies that excluded sex oenders from their sample of violent briey on the pattern of results observed across these studies.
oenders. Three studies excluded sex oenders from their sample of
violent oenders (as well as excluding IPV oenders and oenders with 4.1. Comparisons of violent oenders and non-oenders
a history of drug-related crime) and examined comparisons with non-
oenders. Schnenberg et al. (2013) and Schnenberg et al. (2014) When comparing accuracy and sensitivity among violent oenders
found that violent oenders were less sensitive to recognising angry and non-oenders the evidence suggests that violent oenders were less
(Schnenberg et al., 2013) and fearful and surprised expressions able to recognize negative emotions relative to non-oenders.
(Schnenberg et al., 2014), that is, they required greater emotional Diculties in emotion recognition were observed across dierent sti-
intensity to correctly detect the expression. Schnenberg and Jusyte mulus sets, duration of stimulus presentation, and response format (that
(2014) found that the violent oenders made signicantly more angry is, including or excluding a neutral option). Furthermore, decits for
responses to 50:50 angry-fearful faces than did non-oenders. Likewise, accuracy were found in a sample containing medium-high scorers on
the oenders made more angry responses to 30% angry: 70% happy psychopathy, in which 72% had cluster B personality disorders (Seidel
and 50: 50 than did non-oenders. As none of these studies compared et al., 2013), as well as in samples where no such pathology was re-
violent oenders to a sample of non-violent oenders, it is not clear to ported (Hoaken et al., 2007; Robinson et al., 2012). The impairments
what extent the ndings are related to antisocial pathology as opposed that were most consistently reported were reduced accuracy for disgust
to violent behavior more specically. (Robinson et al., 2012; Seidel et al., 2013), and reduced sensitivity to
Overall, both types of sampling method revealed dierent patterns fear (Gillespie, Rotshtein, Satherley et al., 2015; Schnenberg et al.,
of facial aect processing in the violent sample relative to non-of- 2013; Schnenberg et al., 2014). The observation of impairments for
fending controls. The absence of studies comparing violent non-sex disgust was noteworthy given research showing specic decits in
oenders to non-violent oenders means that it is not possible to review disgust recognition in a sample of psychopathic inmates (Kosson,
whether dierences in sampling method aected whether decits were Suchy, Mayer, & Libby, 2002).
observed relative to non-violent oenders. Although tentative, there is Although error rates were similar among violent oenders in com-
some indication that decits in the processing of disgust expressions are parison to non-oending controls in two studies that excluded sex of-
only found when the violent sample includes sex oenders. fenders from the violent oender sample (Schnenberg et al., 2013;
Schnenberg et al., 2014), in both of these studies violent oenders
3.4.3.3. Studies that analysed sex oenders separately. Two studies required a greater intensity of emotional expression information in
compared sex oenders to violent non-sex oenders. Gillespie, order to make accurate judgements. Thus, violent oenders in these
Rotshtein, Satherley et al., 2015 compared sex oenders to violent two studies demonstrated reduced sensitivity to emotional expressions.
non-sex oenders and non-oenders. They found that both groups of Two separate studies of violent non-sex oenders (Gillespie, Rotshtein,
oenders were less sensitive to recognising fear compared to non- Satherley et al., 2015; Schnenberg et al., 2014) suggested that violent
oenders, and that sex oenders were less sensitive to recognising oenders show impaired sensitivity for fearful expressions in particular
disgust compared to non-oenders. Gillespie, Rotshtein, Satherley compared with non-oenders. A similar trend was also reported by
et al., 2015 also found that emotion processing abilities may be Schnenberg et al. (2013). These ndings are consistent with the meta-
dependent on the sex and the intensity of the emotional expression. analysis of Marsh and Blair (2008) that found decits in fear recogni-
Specically, sex oenders and violent oenders were less sensitive to tion among antisocial populations.
female fearful expressions at high intensities compared to non- There was contrasting evidence as to whether violent oenders
oenders. Sex oenders were also less sensitive to high intensity show a hostile attribution bias (Hoaken et al., 2007; Schnenberg &
female angry expressions compared to violent oenders and non- Jusyte, 2014), providing limited support for theories of aggression that
oenders, and less sensitive to high intensity female faces depicting cite a tendency to attribute hostile intentions to others (Crick & Dodge,
disgust relative to non-oenders. For moderate intensity male faces, 1996; Dodge, Price, Bachorowski, & Newman, 1990; Mellentin et al.,
violent oenders showed a more conservative response style for fear 2015). Although there was some evidence that violent oenders are less
compared with other emotions. likely to interpret fear in morphed or ambiguous facial displays

11
H. Chapman et al. Aggression and Violent Behavior xxx (xxxx) xxxxxx

(Gillespie, Rotshtein, Satherley et al., 2015; Schnenberg & Jusyte, depression and borderline traits have been observed among groups of
2014), overall the ndings raise questions about the consistency with IPV perpetrators (Holtzworth-Munroe, Meehan, Herron, Rehman, &
which biases are observed among violent oenders. Future research Stuart, 2000; Maiuro, Cahn, Vitaliano, Wagner, & Zegree, 1988; Ruiz-
should focus on the conditions under which these biases may be ob- Hernndez, Garca-Jimnez, Llor-Esteban, & Godoy-Fernndez, 2015).
served in carefully selected samples that are well characterized in terms Such ndings provide a basis upon which to explore how the person-
of psychopathology. ality proles of dierent types of oenders are related to social-cogni-
tive abilities.
4.2. Comparisons of oenders with and without a history of violence The nding that violent oenders show generally impaired re-
cognition of emotional expressions is consistent with the ndings from
Of two studies that compared violent oenders to non-violent of- meta-analyses of emotion recognition impairments in psychopathy
fenders (Hoaken et al., 2007; Robinson et al., 2012), one found that, (Dawel et al., 2012; Wilson et al., 2011). In their meta-analysis Dawel
overall, violent oenders generally made more errors (Hoaken et al., et al. (2012) concluded that psychopathy-related impairments in emo-
2007). However, dierences in sample composition were noted, with tion recognition were pervasive across dierent emotions, and also
Hoaken et al. (2007) including sexually-violent oenders in the violent across modalities (e.g., vocal tones). Recent evidence suggests that
oender sample. None of the studies reviewed here examined sensi- psychopathic traits are associated with dierences in the allocation of
tivity between violent and non-violent oenders, while only one study attention to emotionally-salient aspects of the face in both develop-
looked at response bias and found that violent oenders were more mental (Dadds, El Masry, Wimalaweera, & Guastella, 2008) and adult
likely to interpret a neutral face as disgust (Hoaken et al., 2007). samples of violent oenders (Gillespie, Rotshtein, Beech, & Mitchell,
2017) and non-oenders (Gillespie, Rotshtein, Wells, Beech, & Mitchell,
4.3. How does the study of sex oenders aect patterns of performance 2015). Consistent with this, instructing children with callous-unemo-
observed in violent oenders? tional traits to look at the eye region of facial displays temporarily
improves emotion recognition impairments (Dadds et al., 2006).
Due to considerable variability in comparison groups and outcome A better understanding of how attention and emotion interact
measures employed, it is dicult to draw conclusions as to whether the among oender groups may have implications for improving emotion
inclusion or exclusion of sex oenders in violent oender samples af- recognition among violent oenders. A recent study showed that a
fects the observed pattern of results. More specic decits in accuracy training approach encouraging juvenile oenders to attend to the
for disgust were consistently reported in samples that included sexual salient features of happy, sad, angry and fear expressions was successful
oenders (Robinson et al., 2012; Seidel et al., 2013), although reduced in improving recognition of these particular expressions (Hubble,
sensitivity to fear was consistently reported for violent oender samples Bowen, Moore, & Van Goozen, 2015). However, improvements did not
that excluded sexual oenders (Gillespie, Rotshtein, Satherley et al., generalize to the recognition of disgust, a non-trained emotion. Relative
2015; Schnenberg et al., 2014). to treatment as usual, juveniles who undertook this training showed a
The most informative studies in answering this question are those signicant reduction in the severity of oending behavior over the
that compared sexual oenders and violent non-sex oenders to non- following six months (Hubble et al., 2015). Similar results have also
oending controls. Using this design, one study found that sexual of- been reported following emotion recognition training in a develop-
fenders showed more pervasive decits in comparison to violent non- mental sample referred for emotional/behavioral problems (Dadds,
sex oenders (Gillespie, Rotshtein, Satherley et al., 2015). Specically, Cauchi, Wimalaweera, Hawes, & Brennan, 2012). However, it remains
both sexually-violent and violent non-sex oenders were less sensitive unclear how such improvements are attained (Hubble et al., 2015).
to fearful expressions compared to non-oenders, while sexual oen-
ders also showed reduced sensitivity to disgust relative to non-oen- 4.5. Limitations
ders, and to female anger expressions at high intensities relative to both
non-oenders and violent oenders. Overall, the ndings of this review Although this review highlights consistent ndings of impaired re-
suggest that both sexual and violent oenders show impaired facial cognition of others emotional expressions among violent oenders, the
aect recognition, and that any dierences between these groups are studies included in the review varied widely in terms of methodology
likely to be subtle. and sample characteristics. In this section we will discuss the im-
portance of these methodological dierences, and propose a research
4.4. General discussion agenda for future studies of emotion recognition in forensic and clinical
samples.
At present, a lack of understanding about the mechanisms under- The rst limiting factor in comparing across the studies was varia-
pinning emotion recognition impairments in relation to violent and bility in the inclusion criteria for the dierent samples. Several studies
sexually-violent oending precludes understandings about whether, included sexual oenders as part of the violent oender sample, others
and indeed why, decits may be associated with particular forms of excluded sexual oenders, and one study found evidence for subtle
oending. As suggested by Gillespie, Rotshtein, Satherley et al. (2015), dierences when comparing sexual and violent oenders (Gillespie,
it is possible that dierences in social cognition relate to dierences in Rotshtein, Satherley et al., 2015). Sampling dierences are emphasized
personality characteristics. Indeed, a wide body of literature has de- further when considering the inclusion of non-contact sexual oenders.
monstrated a link between personality variables and emotion proces- For example, one study included non-contact sexual oenders in a
sing; including psychopathic traits (Dawel et al., 2012; Wilson et al., sample of non-violent oenders (Hoaken et al., 2007), while a separate
2011), borderline features (Bland, Williams, Scharer, & Manning, 2004; study categorized all sexual oenders, contact and non-contact, as
Daros, Zakzanis, & Ruocco, 2013; Domes et al., 2009; Dyck et al., 2009; violent (Robinson et al., 2012). Although dierences between contact
Lynch et al., 2006), depression (Brotman et al., 2004; Demenescu et al., and non-contact sexual oenders have been identied in the literature
2010) and anxiety (Buckner, Maner, & Schmidt, 2010; Easter et al., (Babchishin, Hanson, & VanZuylen, 2015), the extent to which these
2005). The extent to which these traits are characteristic of dierent groups dier in social-cognitive abilities remains unknown. Indeed, it
types of oender may help to account for the pattern of emotion re- may be hypothesized that intact aective and social-cognitive abilities
cognition impairments observed. For example, sex oenders with child (including facial emotion recognition) represents one potential barrier
victims have been found to have high levels of social phobia in com- to contact oending among men with a history of online-only of-
parison to other groups of oenders (McElroy et al., 1999; Raymond, fending. Carefully dierentiating between these groups may allow for a
Coleman, Ohlerking, Christenson, & Miner, 1999), while high levels of more nuanced understanding of potential protective factors that exist

12
H. Chapman et al. Aggression and Violent Behavior xxx (xxxx) xxxxxx

among men who may be motivated to sexually abuse young people, yet et al. (2012) may have diminished any between group dierences that
appear able to desist from contact sexual abuse (see the motivation- would have been observed under more restricted viewing times.
facilitation model of sexual oending; Seto, 2017). Finally, there was high variability in the number of participants
Understanding dierences based on oence type also has implica- recruited, with small sample sizes a common factor. Although recruit-
tions for assessment and treatment of individuals in the criminal justice ment of forensic samples is often dicult, it is important to recognize
system. The allocation of individuals to treatment programs is often that low sample sizes have implications for estimates of eect size, and
based on the index oence, and this allows for the tailoring of treatment the chances of observing eects that are real, or not real (Button et al.,
modules to the needs of specic types of oender. However, collapsing 2013; Khberger, Fritz, & Scherndl, 2014). For example, a negative
across sexual and violent oenders clouds any judgments about real correlation of sample size with eect size has been found in the psy-
dierences in social-cognitive and aective abilities between these chological literature (Khberger et al., 2014), meaning that even where
groups. The logic for allocating oenders with dierent oence types to a true eect is detected, estimates of the magnitude of the eect are
specialized treatment programs, for example the old Sex Oender likely to be exaggerated in small, underpowered studies (Button et al.,
Treatment Program oered by the Prison and Probation Service for 2013). Moreover, a recent article from the Open Science Collaboration
England and Wales, is based on the assumption that these groups can be (2015) has highlighted problems in the replicability of psychological
distinguished in terms of criminogenic needs, that is, those needs that research, although the conclusions reached have been contested
when treated will be associated with a reduced risk of reoending (Gilbert, King, Pettigrew, & Wilson, 2016). Given the low statistical
(Andrews & Bonta, 2010; Carter & Mann, 2016; Mann et al., 2010). power in the neurosciences literature, recommendations for improving
However, if such dierences do not exist, at least in some domains, then research practices in psychology and the neurosciences have been put
placing individuals on the same program may have benets for treat- forward by Button et al. (2013). These include: performing a priori
ment providers, both in terms of nance and resources. Notably, this power calculations; full disclosure of sample size, data exclusions,
approach has been taken in a redesign of treatment programs for high manipulations, and measures; preregistration of study protocols and
risk and moderate risk oenders in England and Wales, with more data analysis plans; and making study materials and data openly ac-
specialized modules (e.g., the Healthy Sex Program) oered to those cessible (Button et al., 2013). Such measures can also help to increase
with more specic needs. the replicability of research in forensic and clinical psychology, and
Beyond sample composition, some samples were also better char- increase the chances that results detected represent true eects.
acterized than others. For example, not all studies assessed for the The validity of the conclusions drawn from this review is contingent
presence of clinical or antisocial pathology in the sampling of partici- upon a representative sample of all research conducted in this area
pants, and some samples included a number of psychological disorders, being examined. The inclusion and exclusion criteria applied means
including personality disorder, depression, and psychopathy. The in- that only a subset of studies investigating emotion recognition in vio-
consistent sampling of clinical pathologies renders it dicult to reliably lent oenders was examined. The review did not include participants
aggregate and explore ndings across studies. Moreover, it is dicult to from psychiatric populations or female samples, which limits the gen-
determine to what extent the ndings are attributable to psycho- eralizability of the ndings beyond men incarcerated for a violent of-
pathology, violent oending, or both. The failure of some studies to fence. The review specically explored emotion recognition using static
assess for psychopathy (Hoaken et al., 2007; Robinson et al., 2012) is displays of facial aect. While facial expressions convey important in-
particularly concerning given that psychopathic traits are associated formation about emotional experience and are central to social com-
with impaired facial expression recognition (Blair et al., 2004; Dawel munication (Mellentin et al., 2015), facial aect recognition is just one
et al., 2012; Dolan & Fullam, 2006; Gillespie, Mitchell, Satherley, part of social information processing and thus the ndings should not
Beech, & Rotshtein, 2015; Kosson et al., 2002; Montagne et al., 2005). be generalized to social-cognitive abilities more generally. Further-
Previous research has also demonstrated how the nature of the task more, despite eorts to identify relevant studies, some may have been
aects the processing of facial aective information (Smith & Merlusca, missed. Finally, the inclusion of only published studies means that the
2014), and methodological dierences were also noted in terms of the ndings may have been aected by publication bias or a le drawer
stimuli selected, and the presentation of the stimuli on screen. For ex- eect, given that papers reporting positive results are more often
ample, studies varied in the selection of stimuli across the dierent published.
emotional expressions (e.g., anger, disgust, fear, happy, sad, surprise),
the intensity of the expression, and the sex of the model portraying the 5. Conclusion
expression. These features have been shown to aect the processing of
emotional expressions, including accuracy, response times, and eye In conclusion, this review found evidence for generally impaired
movements (Gillespie et al., 2017; Guo, 2012; Kret & de Gelder, 2013; facial aect processing among violent oenders. Decits for sensitivity
Kret, Pichon, Grzes, & de Gelder, 2011; Schurgin et al., 2014; Wells, and accuracy were found in comparison to both non-violent oenders
Gillespie, & Rotshtein, 2016). Studies also varied in the analysis and non-oenders and were most reliably reported for disgust and fear.
strategy, and while the majority of studies analysed dierences as a These decits cannot be explained by task diculty since outcomes are
function of the emotion expressed, one study reported eects collapsed reported relative to the performance of controls on each expression.
across the dierent expressions (Hoaken et al., 2007). Perhaps surprisingly, there was no consistent evidence that violent of-
In terms of dierences in presentation, Schnenberg et al. (2014) fenders show a hostile attribution bias. The review found some evi-
sequentially presented stimuli of increasing intensity in increments of dence based on a limited number of studies that the exclusion of sex
2%, rather than presenting individual stimuli in isolation. Although this oenders from violent samples could aect the pattern of results, al-
paradigm allowed for a sophisticated analysis of the eects of intensity, though violent samples including and excluding sexual oenders
participants may have become more attuned to the emotional stimuli, tended to show some pattern of impairment. The review also high-
masking any potential dierences in accuracy. Dierences in pre- lighted that greater consistency between studies is required to allow for
sentation times are also observed, with stimuli presented for 5 s more meaningful comparisons. Given that impairments appear to be
(Robinson et al., 2012), 2 s (Hoaken et al., 2007), or remaining onsc- generalized across emotions, we would recommend that future studies
reen while a response was made (Gillespie, Rotshtein, Satherley et al., include the six basic emotions, and also examine the eects of intensity
2015). Previous research has indicated that the duration of stimulus and sex given that emotion recognition varies with these parameters.
presentation aects emotion recognition, with better accuracy observed More consistent reporting of psychopathology and oence history
at longer presentation times (Fenske et al., 2015; Neath & Itier, 2014). would also benet future comparisons. Finally, many studies included
Thus, it is possible that the longer viewing times employed by Robinson small sample sizes and this is likely to have resulted in inated eect

13
H. Chapman et al. Aggression and Violent Behavior xxx (xxxx) xxxxxx

sizes and reduced power. Based on the nding of generalized impair- dx.doi.org/10.1002/ab.20324.
ments in face aect recognition, we would suggest that this may be Burt, S., & Neiderhiser, J. (2009). Aggressive versus nonaggressive antisocial behavior:
Distinctive etiological moderation by age. Developmental Psychology, 45(4),
considered a potential treatment target for violent oenders. However, 11641172. http://dx.doi.org/10.1037/a0016130.
a greater understanding of the potential mechanisms underlying these Button, K. S., Ioannidis, J. P., Mokrysz, C., Nosek, B. A., Flint, J., Robinson, E. S., &
impairments is necessary to form an understanding of how these im- Munaf, M. R. (2013). Power failure: why small sample size undermines the relia-
bility of neuroscience. Nature Reviews Neuroscience, 14(5), 365376. http://dx.doi.
pairments may be modied, and to inform the development of suc- org/10.1038/nrn3475.
cessful training-initiatives (Hubble et al., 2015). Carter, A. J., & Mann, R. E. (2016). Organizing principles for an integrated model of
change for the treatment of sexual oending. In A. R. Beech, & T. Ward (Vol. Eds.),
The Wiley Blackwell Handbook on assessment, treatment and theories of sexual oending.
Funding Vol. 1Chichester: Wiley. http://dx.doi.org/10.1002/9781118574003.wattso017.
Coid, J., Yang, M., Ullrich, S., Roberts, A., Moran, P., Bebbington, P., ... Singleton, N.
This work was supported by a grant from the Economic and Social (2009). Psychopathy among prisoners in England and Wales. International Journal of
Law and Psychiatry, 32(3), 134141. http://dx.doi.org/10.1016/j.ijlp.2009.02.008.
Research Council [ES/L002337/1].
Crick, N. R., & Dodge, K. A. (1996). Social information-processing mechanisms in reactive
and proactive aggression. Child Development, 67(3), 9931002. http://dx.doi.org/10.
References1 1111/j.1467-8624.1996.tb01778.
Dadds, M. R., Cauchi, A. J., Wimalaweera, S., Hawes, D. J., & Brennan, J. (2012).
Outcomes, moderators, and mediators of empathic-emotion recognition training for
*Gillespie, S. M., Rotshtein, P., Satherley, R. M., Beech, A. R., & Mitchell, I. J. (2015). complex conduct problems in childhood. Psychiatry Research, 199(3), 201207.
Emotional expression recognition and attribution bias among sexual and violent of- Dadds, M. R., El Masry, Y., Wimalaweera, S., & Guastella, A. J. (2008). Reduced eye gaze
fenders: A signal detection analysis. Frontiers in Psychology, 6. http://dx.doi.org/10. explains fear blindness in childhood psychopathic traits. Journal of the American
3389/fpsyg.2015.00595. Academy of Child and Adolescent Psychiatry, 47, 455463. http://dx.doi.org/10.1097/
*Hoaken, P. N., Allaby, D. B., & Earle, J. (2007). Executive cognitive functioning and the CHI.0b013e31816407f1.
recognition of facial expressions of emotion in incarcerated violent oenders, non- Dadds, M. R., Perry, Y., Hawes, D. J., Merz, S., Riddell, A. C., Haines, D. J., ...
violent oenders, and controls. Aggressive Behavior, 33(5), 412421. http://dx.doi. Abeygunawardane, A. I. (2006). Attention to the eyes and fear-recognition decits in
org/10.1002/ab.20194. child psychopathy. British Journal of Psychiatry, 189, 280281. http://dx.doi.org/10.
*Robinson, L., Spencer, M. D., Thomson, L. D., Sprengelmeyer, R., Owens, D. G., 1192/bjp.bp.105.018150.
Staneld, A. C., ... Johnstone, E. C. (2012). Facial emotion recognition in Scottish Daros, A. R., Zakzanis, K. K., & Ruocco, A. C. (2013). Facial emotion recognition in
prisoners. International Journal of Law and Psychiatry, 35(1), 5761. http://dx.doi. borderline personality disorder. Psychological Medicine, 43(9), 19531963. http://dx.
org/10.1016/j.ijlp.2011.11.009. doi.org/10.1017/S0033291712002607.
*Schnenberg, M., Christian, S., Gauer, A. K., Mayer, S. V., Hautzinger, M., & Jusyte, A. Dawel, A., O'Kearney, R., McKone, E., & Palermo, R. (2012). Not just fear and sadness:
(2014). Addressing perceptual insensitivity to facial aect in violent oenders: First Meta-analytic evidence of pervasive emotion recognition decits for facial and vocal
evidence for the ecacy of a novel implicit training approach. Psychological Medicine, expressions in psychopathy. Neuroscience & Biobehavioral Reviews, 36(10),
44(5), 10431052. http://dx.doi.org/10.1017/S0033291713001517. 22882304. http://dx.doi.org/10.1016/j.neubiorev.2012.08.006.
*Schnenberg, M., & Jusyte, A. (2014). Investigation of the hostile attribution bias toward Demenescu, L. R., Kortekaas, R., den Boer, J. A., & Aleman, A. (2010). Impaired attri-
ambiguous facial cues in antisocial violent oenders. European Archives of Psychiatry bution of emotion to facial expressions in anxiety and major depression. PLoS One,
and Clinical Neuroscience, 264(1), 6169. http://dx.doi.org/10.1007/s00406-013- 5(12), e15058.
0440-1. Dodge, K. A., Price, J. M., Bachorowski, J. A., & Newman, J. P. (1990). Hostile attribu-
*Schnenberg, M., Louis, K., Mayer, S., & Jusyte, A. (2013). Impaired identication of tional biases in severely aggressive adolescents. Journal of Abnormal Psychology, 99,
threat-related social information in male delinquents with antisocial personality 385392. http://dx.doi.org/10.1037/0021-843X.99.4.385.
disorder. Journal of Personality Disorders, 27(4), 496505. http://dx.doi.org/10.1521/ Dolan, M., & Fullam, R. (2006). Face aect recognition decits in personality disordered
pedi_2013_27_100. oenders: Association with psychopathy. Journal of Psychological Medicine, 36,
*Seidel, E. M., Pfabigan, D. M., Keckeis, K., Wucherer, A. M., Jahn, T., Lamm, C., & Derntl, 15631570. http://dx.doi.org/10.1017/S0033291706008634.
B. (2013). Empathic competencies in violent oenders. Psychiatry Research, 210(3), Domes, G., Schulze, L., & Herpertz, S. C. (2009). Emotion recognition in borderline
11681175. http://dx.doi.org/10.1016/j.psychres.2013.08.027. personality disorder: A review of the literature. Journal of Personality Disorders, 23,
Andrews, D. A., & Bonta, J. (2010). The psychology of criminal conduct (5th ed.). New 619. http://dx.doi.org/10.1521/pedi.2009.23.1.6.
Providence, NJ: LexisNexis Group. Dyck, M., Habel, U., Slodczyk, J., Schlummer, J., Backes, V., Schneider, F., & Reske, M.
Babchishin, K. M., Hanson, R. K., & VanZuylen, H. (2015). Online child pornography (2009). Negative bias in fast emotion discrimination in borderline personality dis-
oenders are dierent: A meta-analysis of the characteristics of online and oine sex order. Psychological Medicine, 39(05), 855864. http://dx.doi.org/10.1017/
oenders against children. Archives of Sexual Behavior, 44(1), 4566. http://dx.doi. S0033291708004273.
org/10.1007/s10508-014-0270. Easter, J., McClure, E. B., Monk, C. S., Dhanani, M., Hodgdon, H., Leibenluft, E., ... Ernst,
Blair, R., Mitchell, D., Peschardt, K., Colledge, E., Leonard, R., Shine, J., ... Perrett, D. M. (2005). Emotion recognition decits in pediatric anxiety disorders: Implications
(2004). Reduced sensitivity to others' fearful expressions in psychopathic individuals. for amygdale research. Journal of Child and Adolescent Psychopharmacology, 15,
Personality and Individual Dierences, 37, 11111122. http://dx.doi.org/10.1016/j. 563570. http://dx.doi.org/10.1089/cap.2005.15.563.
paid.2003.10.008. Ekman, P. (1972). Universals and cultural dierences in facial expressions of emotion. In
Blair, R. J. R. (2001). Neurocognitive models of aggression, the antisocial personality J. Cole (Ed.). Nebraska symposium on motivation (pp. 207282). Lincoln: University of
disorders and the psychopath. Journal of Neurology, Neurosurgery & Psychiatry, 71, Nebraska Press.
727731. http://dx.doi.org/10.1136/jnnp.71.6.727. Ekman, P. (1992a). Facial expressions of emotion: New ndings, new questions.
Bland, A. R., Williams, C. A., Scharer, K., & Manning, S. (2004). Emotion processing in Psychological Science, 3, 3438. http://dx.doi.org/10.1111/j.1467-9280.1992.
borderline personality disorders. Issues in Mental Health Nursing, 25, 655672. http:// tb00253.
dx.doi.org/10.1080/01612840490486692. Ekman, P. (1992b). Are there basic emotions? Psychological Review, 99, 550553. http://
Brook, M., Brieman, C. L., & Kosson, D. S. (2013). Emotion processing in Psychopathy dx.doi.org/10.1037/0033-295X.99.3.550.
Checklist- assessed psychopathy: A review of the literature. Clinical Psychology Ekman, P. (1993). Facial expression and emotion. American Psychologist, 48, 384392.
Review, 33(8), 979995. http://dx.doi.org/10.1016/j.cpr.2013.07.008. http://dx.doi.org/10.1037/0003-066X.48.4.384.
Brotman, M., Rich, B., Schmajuk, M., Reising, M., Monk, C., Dickstein, D., ... Leibenluft, E. Ekman, P., & Friesen, W. V. (1971). Constants across cultures in the face and emotion.
(2004). Attention bias to threat faces in children with bipolar disorder and comorbid Journal of Personality and Social Psychology, 17, 124129. http://dx.doi.org/10.1037/
lifetime anxiety disorders. Biological Psychiatry, 61(6), 819821. http://dx.doi.org/ h0030377.
10.1016/j.biopsych.2006.08.021. Ekman, P., & Friesen, W. V. (1976). Pictures of facial aect. San Francisco: University of
Buckner, J. D., Maner, J. K., & Schmidt, N. B. (2010). Diculty disengaging attention California.
from social threat in social anxiety. Cognitive Therapy and Research, 34, 99105. Elfenbein, H. A., & Ambady, N. (2002). On the universality and cultural specicity of
http://dx.doi.org/10.1007/s10608-008-9205. emotion recognition: A meta-analysis. Psychological Bulletin, 128, 203235. http://dx.
Burt, S. (2009). Are there meaningful etiological dierences within antisocial behavior? doi.org/10.1037/0033-2909.128.2.
Results of a meta-analysis. Clinical Psychology Review, 29(2), 163178. http://dx.doi. Fenske, S., Lis, S., Liebke, L., Niedtfeld, I., Kirsch, P., & Mier, D. (2015). Emotion re-
org/10.1016/j.cpr.2008.12.004. cognition in borderline personality disorder: Eects of emotional information on
Burt, S. (2012). Do etiological inuences on aggression overlap with those on rule negative bias. Borderline Personality Disorder and Emotion Dysregulation, 2(1), 112.
breaking? A meta-analysis. Psychological Medicine, 42(1), 112. http://dx.doi.org/10. http://dx.doi.org/10.1186/s40479-015-0031.
1017/S0033291712001894. Fortunata, M., & de Lima Osrio, F. (2014). Recognition of facial expressions by alcoholic
Burt, S., Mikolajewski, A., & Larson, C. (2009). Do aggression and rule-breaking have patients: A systematic literature review. Neuropsychiatric Disease and Treatment, 10,
dierent interpersonal correlates? A study of antisocial behavior subtypes, negative 16551663. http://dx.doi.org/10.2147/NDT.S65376.
aect, and hostile perceptions of others. Aggressive Behavior, 35(6), 453461. http:// Gery, I., Miljkovitch, R., Berthoz, S., & Soussignan, R. (2009). Empathy and recognition of
facial expressions of emotion in sex oenders, non-sex oenders and normal controls.
Psychiatry Research, 165(3), 252262. http://dx.doi.org/10.1016/j.psychres.2007.
1
References marked with an asterisk indicate studies included in the systematic re- 11.006.
view. Gilbert, D. T., King, G., Pettigrew, S., & Wilson, T. D. (2016). Comment on Estimating the

14
H. Chapman et al. Aggression and Violent Behavior xxx (xxxx) xxxxxx

reproducibility of psychological science. Science, 351(6277), 1037-1037. http://dx. Mann, R. E., Hanson, R. K., & Thornton, D. (2010). Assessing risk for sexual recidivism:
doi.org/10.1126/science.aad7243. Some proposals on the nature of psychologically meaningful risk factors. Sexual
Gillespie, S. M., Mitchell, I. J., Satherley, R. M., Beech, A. R., & Rotshtein, P. (2015). Abuse, 22(2), 191217. http://dx.doi.org/10.1177/1079063210366039.
Relations of distinct psychopathic personality traits with anxiety and fear: Findings Marsh, A. A., & Ambady, N. (2007). The inuence of the fear facial expression on pro-
from oenders and non-oenders. PLoS One, 10(11), e0143120. social responding. Cognition and Emotion, 21, 225247. http://dx.doi.org/10.1080/
Gillespie, S. M., Rotshtein, P., Beech, A. R., & Mitchell, I. J. (2017). Boldness psychopathic 02699930600652234.
traits predict reduced gaze toward fearful eyes in men with a history of violence. Marsh, A. A., & Blair, R. J. R. (2008). Decits in facial aect recognition among antisocial
Biological Psychology, 128, 2938. http://dx.doi.org/10.1016/j.biopsycho.2017.07. populations: A meta-analysis. Neuroscience & Biobehavioral Reviews, 32(3), 454465.
003. http://dx.doi.org/10.1016/j.neubiorev.2007.08.003.
Gillespie, S. M., Rotshtein, P., Wells, L. J., Beech, A. R., & Mitchell, I. J. (2015). McCade, D., Savage, G., & Naismith, S. L. (2012). Review of emotion recognition in mild
Psychopathic traits are associated with reduced attention to the eyes of emotional cognitive impairment. Dementia and Geriatric Cognitive Disorders, 32(4), 257266.
faces among adult male non-oenders. Frontiers in Human Neuroscience, 9, 111. http://dx.doi.org/10.1159/000335009.
http://dx.doi.org/10.3389/fnhum.2015.00552. McElroy, S. L., Soutullo, C. A., Taylor, P., Nelson, E. B., Beckman, D. A., Brusman, L. A., ...
Gross, T. F. (2004). The perception of our basic emotions in human and non human faces Keck, P. M. (1999). Psychiatric features of 36 men convicted of sexual oenses. The
by children with autism and other developmental disabilities. Journal of Abnormal Journal of Clinical Psychiatry, 60, 414420. http://dx.doi.org/10.4088/JCP.
Child Psychology, 32, 469480. http://dx.doi.org/10.1023/B:JACP.0000037777. v60n0613.
17698.01. Mellentin, A. I., Dervisevic, A., Stenager, E., Pilegaard, M., & Kirk, U. (2015). Seeing
Guo, K. (2012). Holistic gaze strategy to categorize facial expression of varying in- Enemies? A systematic review of anger bias in the perception of facial expressions
tensities. PLoS One, 7(8), e42585. among anger-prone and aggressive populations. Aggression and Violent Behavior, 25,
Gur, R. C., Sara, R., Hagendoorn, M., Marom, O., Hughett, P., Macy, L., ... Gur, R. E. 373383. http://dx.doi.org/10.1016/j.avb.2015.09.001.
(2002). A method for obtaining 3-dimensional facial expressions and its standardi- Mitchell, A. E., Dickens, G. L., & Picchioni, M. M. (2014). Facial emotion processing in
zation for use in neurocognitive studies. Journal of Neuroscience Methods, 115(2), borderline personality disorder: A systematic review and meta-analysis.
137143. http://dx.doi.org/10.1016/S0165-0270(02)00006-7. Neuropsychology Review, 24(2), 166184. http://dx.doi.org/10.1007/s11065-014-
Hanson, R. K., & Morton-Bourgon, K. E. (2005). The characteristics of persistent sexual 9254-9.
oenders: A meta-analysis of recidivism studies. Journal of Consulting and Clinical Montagne, B., van Honk, J., Kessels, R. P., Frigerio, E., Burt, M., van Zandvoort, M. J., ...
Psychology, 73(6), 11541163. http://dx.doi.org/10.1037/0022-006X.73.6.1154. de Haan, E. H. (2005). Reduced eciency in recognising fear in subjects scoring high
Harris, G. T., Rice, M. E., & Cormier, C. A. (2002). Prospective replication of the Violence on psychopathic personality characteristics. Personality and Individual Dierences, 38,
Risk Appraisal Guide in predicting violent recidivism among forensic patients. Law 511. http://dx.doi.org/10.1016/j.paid.2004.02.008.
and Human Behavior, 26, 377394. http://dx.doi.org/10.1023/A:1016347320889. Moore, B. S. (1990). The origins and development of empathy. Motivation and Emotion,
Holtzworth-Munroe, A., Meehan, J. C., Herron, K., Rehman, U., & Stuart, G. L. (2000). 14, 7580. http://dx.doi.org/10.1007/BF00991636.
Testing the Holtzworth-Munroe and Stuart (1994) batterer typology. Journal of Neath, K., & Itier, R. (2014). Impact of task demands and xation to features on the time
Consulting and Clinical Psychology, 68(6), 10001019. http://dx.doi.org/10.1037/ course of facial emotion processing. Journal of Vision, 14(10), 1397-1397. http://dx.
0022-006X.68.6.1000. doi.org/10.1167/14.10.1397.
Hoppenbrouwers, S. S., Bulten, B. H., & Brazil, I. A. (2016). Parsing fear: A reassessment Open Science Collaboration. (2015). Estimating the reproducibility of psychological sci-
of the evidence for fear decits in psychopathy. Psychological Bulletin, 142(6), ence. Science, 349(6251), aac4716. http://dx.doi.org/10.1126/science.aac4716.
573600. http://dx.doi.org/10.1037/bul0000040. Pascual-Leone, A., Bierman, R., Arnold, R., & Stasiak, E. (2011). Emotion-focused therapy
Hubble, K., Bowen, K. L., Moore, S. C., & Van Goozen, S. H. (2015). Improving negative for incarcerated oenders of intimate partner violence: A 3-year outcome using a new
emotion recognition in young oenders reduces subsequent crime. PLoS One, 10(6), whole-sample matching method. Psychotherapy Research, 21(3), 331347. http://dx.
113 e0132035. doi.org/10.1080/10503307.2011.572092.
Hudson, S. M., Marshall, W. L., Wales, D., McDonald, E., Bakker, L. W., & McLean, A. Rapport, L., Friedman, S., Tzelepis, A., & Van Voorhis, A. (2002). Experienced emotion
(1993). Emotional recognition skills of sex oenders. Sexual Abuse, 6(3), 199-21. and aect recognition in adult attention- decit hyperactivity disorder.
http://dx.doi.org/10.1177/107906329300600303. Neuropsychology, 16, 102110. http://dx.doi.org/10.1037//0894-4105.16.1.10.
Keltner, D. (2003). Expression and the course of life: Studies of emotion, personality, and Raymond, N. C., Coleman, E., Ohlerking, F., Christenson, G. A., & Miner, M. (1999).
psychopathology form a social-functional perspective. Annals of the New York Psychiatric comorbidity in pedophilic sex oenders. The American Journal of
Academy of Sciences, 1000, 222243. http://dx.doi.org/10.1196/annals.1280.011. Psychiatry, 156, 786788. http://dx.doi.org/10.1176/ajp.156.5.786.
Kohler, C. G., & Brennan, A. (2004). Recognition of facial emotions in schizophrenia. Rojahn, J., Lederer, M., & Tass, M. J. (1995). Facial emotion recognition by persons with
Current Opinion in Psychiatry, 17, 8186. mental retardation: A review of the experimental literature. Research in Developmental
Kosson, D. S., Suchy, Y., Mayer, A. R., & Libby, J. (2002). Facial aect recognition in Disabilities, 16(5), 393414. http://dx.doi.org/10.1016/0891-4222(95)00019-J.
criminal psychopaths. Emotion, 2, 398411. http://dx.doi.org/10.1037/1528-3542.2. Ruiz-Hernndez, J. A., Garca-Jimnez, J. J., Llor-Esteban, B., & Godoy-Fernndez, C.
4.398. (2015). Risk factors for intimate partner violence in prison inmates. The European
Kret, M. E., & de Gelder, B. (2013). When a smile becomes a st: The perception of facial Journal of Psychology Applied to Legal Context, 7(1), 4149. http://dx.doi.org/10.
and bodily expressions of emotion in violent oenders. Experimental Brain Research, 1016/j.ejpal.2014.11.003.
228, 399410. http://dx.doi.org/10.1007/s00221-013-3557-6. Schnenberg, M., Mayer, S. V., Christian, S., Louis, K., & Jusyte, A. (2015). Facial aect
Kret, M. E., Pichon, S., Grzes, J., & de Gelder, B. (2011). Men fear other men most: recognition in violent and nonviolent antisocial behavior subtypes. Journal of
Gender specic brain activations in perceiving threat from dynamic faces and bodies Personality Disorders, 30(5), 708719. http://dx.doi.org/10.1521/pedi_2015_29_217.
An fMRI study. Frontiers in Psychology, 2(3), http://dx.doi.org/10.3389/fpsyg.2011. Schurgin, M. W., Nelson, J., Iida, S., Ohira, H., Chiao, J. Y., & Franconeri, S. L. (2014). Eye
00003. movements during emotion recognition in faces. Journal of Vision, 14(13), 14-14.
Khberger, A., Fritz, A., & Scherndl, T. (2014). Publication bias in psychology: a diagnosis http://dx.doi.org/10.1167/14.13.14.
based on the correlation between eect size and sample size. PLoS One, 9(9), Seto, M. C. (2017). The motivation-facilitation model of sexual oending. Sexual Abuse
e105825. (doi:1079063217720919).
Langner, O., Dotsch, R., Bijlstra, G., Wigboldus, D. H. J., Hawk, S. T., & Van Knippenberg, da Silva Ferreira, G. C., Crippa, J. A., & de Lima Osrio, F. (2014). Facial emotion pro-
A. (2010). Presentation and validation of the Radboud Faces Database. Cognition and cessing and recognition among maltreated children: A systematic literature review.
Emotion, 24, 13771388. http://dx.doi.org/10.1080/02699930903485076. Frontiers in Psychology, 5, 110. http://dx.doi.org/10.3389/fpsyg.2014.01460.
Leime, J. L., Rique Neto, J., Alves, S. M., & Torro-Alves, N. (2013). Recognition of facial Singh, S. D., Ellis, C. R., Winton, A. S., Singh, N. N., Leung, J. P., & Oswald, D. P. (1998).
expressions in children, young adults and elderly people. Estudos de Psicologia Recognition of facial expressions of emotion by children with attention-decit hy-
(Campinas), 30(2), 161167. http://dx.doi.org/10.1590/S0103- peractivity disorder. Behavior Modication, 22(2), 128142. http://dx.doi.org/10.
166X2013000200002. 1177/01454455980222002.
Leist, T., & Dadds, M. R. (2009). Adolescents' ability to read dierent emotional faces Smith, M. L., & Merlusca, C. (2014). How task shapes the use of information during facial
relates to their history of maltreatment and type of psychopathology. Clinical Child expression categorizations. Emotion, 14(3), 478487. http://dx.doi.org/10.1037/
Psychology and Psychiatry, 14, 237250. http://dx.doi.org/10.1177/ a0035588.
1359104508100887. Sullivan, S., Ruman, T., & Hutton, S. B. (2007). Age dierences in emotion recognition
Loney, B. R., Frick, P. J., Clements, C. B., Ellis, M. L., & Kerlin, K. (2003). Callous-un- skills and the visual scanning of emotion faces. The Journals of Gerontology. Series B,
emotional traits, impulsivity, and emotional processing in adolescents with antisocial Psychological Sciences and Social Sciences, 62, 5360. http://dx.doi.org/10.1093/
behavior problems. Journal of Clinical Child and Adolescent Psychology, 32(1), 6680. geronb/62.1.P53.
http://dx.doi.org/10.1207/S15374424JCCP3201_07. Surguladze, S. A., Young, A. W., Senior, C., Brebion, G., Travis, M., & Phillips, M. (2004).
Lundqvist, D., Flykt, A., & hman, A. (1998). The Karolinska Directed Emotional Faces - Recognition accuracy and response bias to happy and sad facial expressions in pa-
KDEF, CD ROM from Department of Clinical Neuroscience, Psychology section, Karolinska tients with major depression. Neuropsychology, 18, 212218. http://dx.doi.org/10.
Institutet. 91-630-7164-9. 1037/0894-4105.18.2.212.
Lynch, T. R., Rosenthal, M. Z., Kosson, D. S., Cheavens, J. S., Lejuez, C. W., & Blair, R. J. R. Tackett, J. L., Krueger, R. F., Sawyer, M. G., & Graetz, B. W. (2003). Subfactors of DSM-IV
(2006). Heightened sensitivity to facial expressions of emotion in borderline per- conduct disorder: Evidence and connections with syndromes from the Child Behavior
sonality disorder. Emotion, 6(4), 647655. http://dx.doi.org/10.1037/1528-3542.6. Checklist. Journal of Abnormal Child Psychology, 31(6), 647654 (doi:0091-0627/03/
4.647. 1200-0647/0).
Maiuro, R. D., Cahn, T. S., Vitaliano, P. P., Wagner, B. C., & Zegree, J. B. (1988). Anger, Tottenham, N., Tanaka, J. W., Leon, A. C., McCarry, T., Nurse, M., Hare, T. A., ... Nelson,
hostility, and depression in domestically violent versus generally assaultive men and C. (2009). The NimStim set of facial expressions: Judgements from untrained re-
nonviolent control subjects. Journal of Consulting and Clinical Psychology, 56(1), search participants. Psychiatry Research, 168, 242249. http://dx.doi.org/10.1016/j.
1723. http://dx.doi.org/10.1037/0022-006X.56.1.17. psychres.2008.05.006.

15
H. Chapman et al. Aggression and Violent Behavior xxx (xxxx) xxxxxx

Trmeau, F. (2006). A review of emotion decits in schizophrenia. Dialogues in Clinical analytic investigation of the facial aect recognition decit. Criminal Justice and
Neuroscience, 8, 5970. Behavior, 38(7), 659668. http://dx.doi.org/10.1177/0093854811404120.
Wells, L. J., Gillespie, S. M., & Rotshtein, P. (2016). Identication of emotional facial Zaja, R. H., & Rojahn, J. (2008). Facial emotion recognition in intellectual disabilities.
expressions: eects of expression, intensity, and sex on eye gaze. PLoS One, 11(12), Current Opinion in Psychiatry, 21(5), 441444. http://dx.doi.org/10.1097/YCO.
e0168307. 0b013e328305e5fd.
Wilson, K., Juodis, M., & Porter, S. (2011). Fear and loathing in psychopaths: A meta-

16

You might also like