Towards a Network Model of Psychopathy: Lack of Empathy is Central to the
Network of Psychopathy Checklist Revised Items in Criminal Offenders
Lena Christ (10204741) Master’s thesis
Under supervision of Arjen Noordhof and Bruno Verschuere University of Amsterdam
Department of Clinical Psychology July 2016
Word count: 9136
Table of Contents
Abstract ... 3
Introduction ... 4
Psychopathy ... 4
The Current State of Psychopathy Research... 5
Emotional deficits. ... 6
Attentional deficits. ... 7
Integrative models... 9
Core symptoms in psychopathy measures and descriptions. ... 9
Conceptualization of Psychopathology... 11
Network Theory ... 12
Current Study ... 14
General Method... 14
Analysis 1: Dutch Sample... 15
Method ... 15
Sample... 15
Outcome measure... 16
Exclusion criteria and score adjustment... 18
Data analysis. ... 19
Results ... 21
Missing data and duplicate cases. ... 21
Participant characteristics. ... 21
Network analysis... 22
Analysis of proposed core symptoms. ... 26
Analysis 2: American Sample ... 27
Method ... 27
Outcome measure... 28
Exclusion criteria, score adjustment, and data analysis. ... 28
Results ... 28
Missing data and duplicate cases. ... 28
Participant characteristics. ... 28
Network analysis... 29
Analysis of proposed core symptoms. ... 33
Discussion ... 34
Key Findings and Theoretical Implications ... 34
Inconsistencies between Samples ... 35
Implications for Psychopathy Assessment Instruments ... 35
Implications for Forensic Practice ... 36
Limitations ... 37 Future Research... 38 Summary ... 38 References ... 39 Tables ... 48 Figures... 53
Abstract
Although psychopathy is well-researched, there remains considerable debate on its core symptoms. Many theoretical models suggest that (a) emotional deficits, (b) attentional deficits, or (c) a combination of the former are hallmark features of psychopathy. Network analysis, a new statistical method to investigate structures of psychological constructs, was applied to visually display PCL-R items and their interaction in a large Dutch sample of violent mentally disordered offenders (N=2,043), and a large North-American offender sample (N =4,221). Items differed substantially in their centrality. In both samples,
‘callousness/lack of empathy’ (emotional deficit) was central to the PCL-R network while ‘shallow affect’ (emotional deficit) was peripheral. ‘Poor behavioral controls’ and
‘impulsivity’ (attentional deficits) were moderately central at most. These results suggest that a core symptom of psychopathy is a lack of empathy, rather than generally blunted
emotionality or attentional deficits.
Keywords: psychopathy, PCL-R, core symptoms, network analysis, empathy, fearlessness, impulsivity, emotional deficits, attentional deficits.
Towards a Network Model of Psychopathy: Lack of Empathy is Central to the Network of Psychopathy Checklist Revised Items in Criminal Offenders
Psychopathy has arguably been one of the most intriguing mental disorders in history, in the scientific world and popular culture alike. In his book, Without Conscience, Hare (1999) wrote that psychopathy is “a dark mystery with staggering implications for society” (p.16). Psychopaths1 are commonly portrayed as the personification of evil, almost inhuman, yet ostensibly charming and well-functioning (Cleckley, 1976; Hare, 1999); with fictional characters such as Hannibal Lector and real life examples such as Charles Manson, Ted Bundy or Mark Dutroux.
Psychopathy
Psychopathy is a severe personality disorder characterized by a lack of emotions, disinhibition and antisocial behavior (Cleckley, 1976; Hare, 1999; Patrick, 2005). High scores on psychopathy measures have been linked to violence, misconduct, and criminality (for a review, see Patrick, 2005; Skeem et al., 2011). According to research findings, as much as 15-30% of criminal offenders exhibit elevated psychopathy symptoms, compared to an estimated 1% of the general population (Hare, 1999; Yildirim & Derksen, 2015). Psychopathy is associated with poor treatment outcome and a higher risk of recidivism in criminal offenders (e.g. Harris & Rice, 2006; Ogloff, Wong, & Greenwood, 1990; but see D’Silva, Duggan, & McCarthy, 2004; Skeem, Monahan, & Mulvey, 2002). Interestingly, research suggests that highly psychopathic offenders are nonetheless more likely to be granted conditional release compared to non-psychopathic offenders (Porter, Brinke, & Wilson, 2009). In sum,
psychopathy causes a lot of physical, emotional and financial harm to society. These findings show the relevance of psychopathy assessments in the forensic setting.
fundamentally different from non-psychopathy. Scientific research, however, is still debating about whether there is a distinct difference between psychopaths and non-psychopaths, other than the degree of psychopathic symptoms: While some studies found a discrete variable underlying psychopathy symptoms (e.g. Sellbom et al., 2015), others suggest that the syndrome is better understood as a set of dimensional traits, much like many other
personality disorders (e.g. Hare & Neumann, 2008; Marcus et al., 2004; Walters et al., 2007). In this paper, we propose a novel view of psychopathy symptoms as components of an interactive network with the hope that a better understanding of the syndrome will aid in the development of assessment tools and treatment options. The structure of the introduction is as follows. First, we briefly summarize the current state of psychopathy research and highlight several influential theories of psychopathy. Second, we review the current conceptual model of psychopathy and other mental disorders. Third, we introduce an alternative, network-based model of psychopathy and discuss possible implications for the assessment and treatment of criminal offenders with psychopathy tendencies.
The Current State of Psychopathy Research
Despite a large body of research on the subject, the debate on the conceptualization of psychopathy and its core features remains unresolved (for a review, see Patrick, Fowles, & Krueger, 2009; Skeem et al., 2011). Among other topics, this debate involves a discussion of whether criminal conduct is essential to psychopathy (e.g. Cooke, Michie, Hart, & Clark, 2004; Cooke & Michie, 2001; Walters, 2004; Yildirim & Derksen, 2015), whether
psychopathy consists of a two-, three- or four-factor structure (Cooke et al., 2004; Hare, 2003; Harpur, Hare, & Hakstian, 1989) and whether the development of psychopathy symptoms can be traced back to genetic predispositions (e.g. Larsson, Andershed, &
Lichtenstein, 2006; Patrick et al., 2009; Taylor, Loney, Bobadilla, Iacono, & McGue, 2003).
Psychopathy theories also speak to which symptoms or symptom clusters are central to the syndrome. In the following paragraphs, we briefly review two of the most influential notions concerning core symptoms in psychopathy: emotional deficits and attentional deficits.
Emotional deficits. Multiple theoretical models suggest that some form of emotional deficit lies at the core of psychopathy symptoms. Most notably, it has been proposed that (1) a lack of fear, (2) a lack of empathy, and (3) generally blunted emotions are core symptoms of psychopathy.
Low fear. One of the earliest and most influential theoretical models of psychopathy is based on Lykken’s research on anxiety and fear: Lykken (1957) found that, compared to controls, psychopathic individuals reported less anxiety, responded less to fear conditioning and performed worse on passive avoidance learning tasks. These findings might be explained with deficient arousal in response to aversive or threatening stimuli (Birbaumer et al., 2005; Fowles, 1980; Hare & Neumann, 2005; Lykken, 1995). Lykken (1957; 1995) argued that fearless individuals would not be motivated to inhibit psychopathic behavior to avoid punishment. Hoppenbrouwers, Bulten, & Brazil (2016) made a distinction between threat detection (i.e. the automatic, physiological response to threatening stimuli) and subjective fear experience. Based on a meta-analysis, they concluded that there is considerable evidence for deficient threat detection in psychopathy but little evidence pointing to abnormal levels of subjective fear (Hoppenbrouwers et al., 2016).
Low empathy. Blair (1992) hypothesized that not only fearlessness may interfere with the inhibition of psychopathic behavior but also a lack of empathy, i.e. unresponsiveness to the distress of others. To test this hypothesis, Blair and colleagues (1997) measured
psychopathic individuals’ physiological response to affective cues; they found that
psychopathy symptoms, i.e. high-scorers on psychopathy measures.
psychopathic individuals did not differ from controls in terms of responsivity to threat cues (e.g. a picture of a gun or an angry face) but that they responded significantly less to distress cues (e.g. pictures of sad or fearful faces). Blair interpreted these findings as indicative of an impaired violence inhibition mechanism in psychopathy (Blair, Colledge, Murray, &
Mitchell, 2001; Blair, 1992, 1995). The violence inhibition mechanism is an instinct seen in animals as well as humans which is thought to moderate violent behavior on sight of an opponent’s submission, e.g. dogs stop attacking each other if the attacked dog offers their throat (Blair et al., 2001; Blair, 1992, 1995).
General emotional blunting. Several studies suggest that not only fear and/or empathy may be deficient but that the emotions of highly psychopathic individuals are generally blunted (Casey, Rogers, Burns, & Yiend, 2013; Ermer, Kahn, Salovey, & Kiehl, 2012; but see Glass & Newman, 2006). An explanation for these generally blunted emotions could be impairments in brain regions that are involved in processing and forming emotions. Research has linked psychopathy to abnormalities in the amygdala, the emotional center of the brain (Boccardi et al., 2011; Pardini, Raine, Erickson, & Loeber, 2014).
In sum, several theoretical models of psychopathy assign a key role to emotional deficits. However, there is no agreement on which of these emotional deficits are central to psychopathy.
Attentional deficits. Attentional deficit models of psychopathy offer an alternative explanation for the perceived emotional deficits in psychopathy. This model suggests that psychopathy symptoms are the result of an impaired ability to shift attention and moderate behavior rather than an insensitivity to emotional stimuli (Hamilton, Hiatt, & Newman, 2015; Hiatt, Schmitt, & Newman, 2004). Newman and colleagues (1993; 2004; 2015) suggested that psychopathic individuals lack the ability to integrate secondary information, i.e. information that would interfere with the ongoing task while engaged in goal-directed
behavior. This would mean that psychopathic individuals have difficulties in inhibiting
behavior due to an over-selective attention (Hamilton et al., 2015; Hiatt et al., 2004; Patterson & Newman, 1993). In line with this, it has been proposed that psychopathic individuals show similar behavioral patterns to patients with frontal lobe lesions, suggesting a dysfunction of these brain regions in psychopathy (Blair, 2001; Devonshire, Howard, & Sellars, 1988; Gorenstein, 1982; Lapierre, Braun, & Hodgins, 1995; Morgan & Lilienfeld, 2000).
Symptoms that are often seen in patients with damage to the frontal lobe, specifically to the orbitofrontal cortex, are impulsivity and a lack of behavioral control, leading to aggressive behavior, sexual promiscuity and irresponsibility (Starkstein & Robinson, 1997). Gorenstein (1982) found that psychopathic individuals performed poorly on tasks that measure executive functioning, e.g. the Wisconsin Card Sorting Test (Gorenstein, 1982; but see Hare, 1984; Sutker & Allain, 1987). The orbitofrontal cortex is also involved in moral decision making and reinforcement learning (Starkstein & Robinson, 1997). Research has repeatedly shown that psychopathic individuals perform poorly on reinforcement learning tasks (Blair, 2013; Blair et al., 2004; Lykken, 1956; Newman, Patterson, Howland, & Nichols, 1990). Recent neuroimaging studies found abnormal neuronal activity in several areas of the brains of highly psychopathic individuals, including increased activity in some frontal lobe regions during tasks that involved processing of emotional stimuli and moral decision-making (for a review, see Seara-Cardoso, Neumann, Roiser, McCrory, & Viding, 2014). Although more research is needed to draw reliable conclusions, these findings suggest that psychopathy is associated with abnormal processing of emotional information, rather than a general dysfunction of frontal lobe regions (Seara-Cardoso et al., 2014).
Integrative models. It is important to stress that emotional deficit models and attentional deficit models of psychopathy are not mutually exclusive and attempts have been made to combine emotional and attentional deficits into a single, integrative model of psychopathy (e.g. Hamilton et al., 2015; Patrick et al., 2009). For example, Patrick and colleagues (2009) proposed a model consisting of three, partly overlapping, phenotypical components: Meanness, defined by a lack of empathy, callousness and instrumental aggression; boldness, defined by fearlessness, high stress resistance and confidence; and disinhibition which is characterized by poor impulse control and reactive aggression.
Core symptoms in psychopathy measures and descriptions. To recap, many theoretical models of psychopathy suggest that (a) emotional deficits, (b) attentional deficits or (c) a combination of the former are core symptoms of psychopathy. As shown in Table 1, common psychopathy descriptions and assessment instruments include both emotional and attentional deficits. Cleckley (1976), a pioneer in psychopathy research, proposed 16 psychopathic characteristics including ‘absence of nervousness or psychoneurotic manifestations’ consistent with the low fear model of psychopathy, ‘lack of remorse and shame’ consistent with the low empathy model of psychopathy, ‘general poverty in major affective reactions’ consistent with the general emotional blunting model, and ‘poor judgment and failure to learn by experience’ consistent with the attentional deficit model of
psychopathy (p. 337-338). Hare’s Psychopathy Checklist Revised (PCL-R; 1991; 2003), a popular rating scale to measure psychopathy in criminal offenders, includes ‘lack of remorse or guilt’ and ‘callousness/lack of empathy’ consistent with the low empathy model of
psychopathy, ‘shallow affect’ consistent with the general emotional blunting model, and ‘impulsivity’ and ‘poor behavioral controls’ consistent with the attentional deficit model of psychopathy. While the PCL-R does not directly measure anxiety or fear, Neumann, Hare, & Johansson (2012) found that all dimensions of the PCL-R correlate with low trait anxiety and
fear (for a full list of PCL-R items, see Table 2). The Psychopathic Personality Inventory (PPI, Lilienfeld & Andrews, 1996), a self-report measure of psychopathy in offender and community samples, contains eight subscales, three of which correspond to emotional deficit models: ‘fearlessness’ and ‘stress immunity’ consistent with the low fear model of
psychopathy; ‘coldheartedness’ consistent with the low empathy model of psychopathy, and ‘carefree nonplanfulness’ consistent with the attentional deficit model of psychopathy. The Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM–5; American
Psychiatric Association, 2013) does not list psychopathy as a diagnosis but proposes an alternative model of antisocial personality disorder with a psychopathic specifier including ‘lack of anxiety or fear’ consistent with the low fear model of psychopathy, ‘callousness’ consistent with the low empathy model of psychopathy, and ‘risk taking’ and ‘impulsivity’ consistent with the attentional deficit model of psychopathy (p. 764-765).
Table 1
Note. Cleckley = The Mask of Sanity (Cleckley, 1976), PCL-R = Psychopathy Checklist Revised (Hare, 1991; 2003), PPI = Psychopathic Personality Inventory (Lilienfeld & Andrews, 1996), DSM-5 = proposed model of antisocial personality disorder with psychopathic features in the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; American Psychiatric Association, 2013).
Conceptualization of Psychopathology
In the current study, we aim to explore the role of proposed core symptoms of psychopathy in the PCL-R with the help of a new theoretical framework, network theory. In network theory, psychological constructs are conceptualized as webs of pairwise symptom-interactions (Borsboom, Cramer, Schmittmann, Epskamp, & Waldorp, 2011). Here,
symptoms are not merely passive reflections of an underlying disorder but play an active role in the manifestation of the syndrome.
Traditionally, psychopathy and many other mental health conditions are
Emotional and Attentional Deficits in Popular Psychopathy Measures and Descriptions. Theoretical
model
Cleckley PCL-R PPI DSM-5
low fear ‘absence of nervousness or psychoneurotic manifestations’ ‘fearlessness’ ‘stress immunity’ ‘lack of anxiety or fear’
low empathy ‘lack of remorse and shame’ ‘lack of remorse or guilt’ ‘callousness/lack of empathy’ ‘coldheartedness’ ‘callousness’ general emotional blunting ‘general poverty in major affective reactions’ ‘shallow affect’ attentional deficits ‘poor judgment and failure to learn by experience’ ‘impulsivity’ ‘poor behavioral controls’ ‘carefree nonplanfulness’ ‘risk taking’ ‘impulsivity’
conceptualized as latent variables (Everitt, 1984). In latent variable models, symptoms are defined as passive expressions of an underlying disorder which we cannot directly measure (Borsboom et al., 2011; Borsboom & Cramer, 2013; Everitt, 1984). One assumption of latent variable models is that symptoms are interchangeable or locally independent, which means they co-occur because they have a common cause but do not directly interact with each other (Henning, 1989). While the assumption of local independence of symptoms is generally accepted for medical conditions that can be understood as separate entities (e.g. a brain tumor which exists independently of a patient’s symptoms), there are many practical and conceptual issues with this view regarding mental health conditions (Borsboom & Cramer, 2013; Cramer, Waldorp, van der Maas, & Borsboom, 2010). These issues are best explained by an example: Ruminating thoughts, insomnia and a lack of energy are symptoms of major depressive disorder (American Psychiatric Association, 2013); based on the latent variable model of psychopathology, these symptoms should not directly interact with each other; in practice, however, it is possible that ruminating thoughts cause insomnia and consequently a lack of energy the following day (Schmittmann et al., 2013). Similarly, in the case of
psychopathy, it seems likely that symptoms, e.g. impulsivity and poor behavioral controls, do not merely co-occur because they are both caused by the same underlying disorder but that they also directly interact with each other. In recent years, a new theoretical framework has emerged that provides an addition as well as an alternative to latent variable models that are limited by the assumption of local independence (Borsboom et al., 2011).
Network Theory
A benefit of conceptualizing psychological constructs as network models is that the underlying dynamic of how symptoms relate to each other becomes accessible (Kossakowski et al., 2015). While traditional models and network models of psychopathology are not mutually exclusive, latent variables are not needed to understand psychological constructs
from a network point of view. Network models can also aid in identifying core symptoms, i.e. symptoms that are most relevant to the construct and are strongly associated with other
symptoms (Borsboom & Cramer, 2013). In clinical practice, these are usually symptoms that cause the most psychosocial impairment and predict treatment response and outcome (Fried, Epskamp, Nesse, Tuerlinckx, & Borsboom, 2015). Therefore, these core symptoms of psychopathology should receive special attention in the diagnostic process. In extent, identifying core symptoms may offer new opportunities for the development of more effective treatment options targeted at specific symptoms.
Network analysis is a statistical method to visually display and analyze network models (Epskamp, Rhemtulla, & Waldorp, 2014). In network analysis, instead of trying to determine how well a symptom reflects a latent disorder, we are interested in how central a specific symptom is to the symptom network, i.e. how closely interconnected a symptom is with all other symptoms in the network (Epskamp, Cramer, Waldorp, Schmittmann, & Borsboom, 2012; Opsahl, Agneessens, & Skvoretz, 2010). Network analysis has been applied to a wide range of psychological constructs such as generalized anxiety disorder, depression, quality of life, posttraumatic stress disorder and substance abuse disorder (Cramer et al., 2010; Fried et al., 2015; Kossakowski et al., 2015; McNally et al., 2014; Rhemtulla et al., 2016; Schmittmann et al., 2013). For example, Fried and colleagues (2015) questioned
whether DSM-5 criteria for major depressive disorder are more relevant to the syndrome than symptoms included in non-DSM descriptions of depression. Network analysis showed that DSM symptoms were no more central to the network of depression than non-DSM symptoms (Fried et al., 2015). Furthermore, symptoms differed significantly in terms of centrality, and the most central symptoms were those that previous research had shown to be the most clinically relevant (Fried et al., 2015). Fried and colleagues (2015) concluded that we need to shift our understanding of psychopathology to a more symptom-oriented approach. The
question of symptom relevance has long been a topic of controversy in psychopathy research as well, and we believe that network analysis could be a fruitful addition to traditional
statistical methods in psychopathy research. Current Study
The current study aims to identify the role of proposed core symptoms of psychopathy in a network of PCL-R items. Two criminal offender samples in which the PCL-R was administered were analyzed. Network analysis, a fruitful new method to display complex psychological constructs, might offer some new insights concerning core features of psychopathy. First, based on Lykken´s research one would expect fearlessness to be a core symptom of psychopathy. Fearlessness is not directly measured by the PCL-R and therefore we cannot test this hypothesis in the current study. Second, based on Blair’s low empathy model of psychopathy, one would expect the items ‘callousness/lack of empathy’ and ‘lack of remorse or guilt’ to be most central in the PCL-R network. Third, based on the general
emotional blunting model of psychopathy, one would expect the item ‘shallow affect’ to be most central to the PCL-R network. Forth, based on the attentional deficit model of
psychopathy, one would expect the items ‘impulsivity’ and ‘poor behavioral controls’ to be most central to the PCL-R network. A better understanding of the PCL-R could improve the assessment of psychopathy symptoms and lay the groundwork for treatment approaches targeted at specific core symptoms of psychopathy. Identified core symptoms of psychopathy might also be considered risk factors or early warning signs of psychopathy.
General Method
We computed and interpreted several symptom networks of psychopathy in criminal offenders. First, we analyzed a sample of 2,042 Dutch forensic patients. Second, we repeated the analysis with a sample of 4,221 American prisoners to test whether the network structure was stable across populations.
A number of supplemental materials that might be interesting to the reader but that are not essential to the study can be found at http://tiny.cc/supp_materials. A link to the
supplemental materials is provided where appropriate in the text below. Analysis 1: Dutch Sample
Method
Sample. The sample consisted of 2,042 patients in Detention under Hospital Order (Terbeschikkingstelling or TBS) who were treated at a Dutch forensic psychiatric clinic between 1997 and 2014. Detention under Hospital Order is a uniquely Dutch legal action inflicted upon adult criminal offenders who (1) committed a serious offence that would normally be charged with at least four years of confinement (e.g. manslaughter, murder or aggravated assault), (2) suffer from mental illness (e.g. a personality or psychotic disorder) or retardation and (3) were declared partially or entirely unaccountable for their actions. If a psychiatric examination reveals that an offender poses a risk to the public due to their poor mental health, a court may order them to receive inpatient treatment at a forensic psychiatric clinic to reduce the risk of a future offense. Twelve of the fourteen forensic psychiatric clinics in the Netherlands participated in the current study. The data were gathered as part of the annual risk taxations of patients in Detention under Hospital Order. To estimate the risk of recidivism for these patients, a number of standardized tests and assessments are conducted at least once a year at every forensic psychiatric clinic in the Netherlands. The results of these assessments are documented in the Dutch National Database for Risk Assessment TBS (Landelijke Databank Risicotaxatie TBS2). The dataset did not contain any information about patients’ race or nationality. The Dutch Government states that 90% of TBS patients are of Dutch nationality (TBS Nederland, n.d.). In 2014 72% of TBS patients were born in the Netherlands, 6% in Suriname, 6% in the Dutch Antilles, 4% in Morocco, 2% in Turkey, 1%
in Somalia and 1% in Germany (Gemmert & Schijndel, 2015).
Outcome measure. Psychopathy was measured with the Psychopathy Checklist Revised (PCL-R; Hare, 1991; Dutch version: Vertommen, Verheul, De Ruiter, & Hildebrand, 2002). Each of the 20 PCL-R items corresponds to a specific psychopathic trait or behavior (Hare, 1991; 2003). Factor analyses have revealed that items load on at least two factors: Factor 1 covers interpersonal and affective aspects of psychopathy and Factor 2 covers impulsive and antisocial behavioral aspects of psychopathy (Harpur et al., 1988). These findings have been replicated multiple times and seem stable across populations (Hare, Hart, & Harpur, 1991; Harpur et al., 1989; Zwets, Hornsveld, Neumann, Muris, & van Marle, 2015). The two-factor model can be further subdivided into four facets (Hare, 2003): Factor 1 splits into an affective and an interpersonal facet and Factor 2 splits into an impulsive
lifestyle and an antisocial facet (see Table 2). Cooke and Michie (2001) proposed an alternative model excluding seven items that strictly measure antisocial behavior (e.g. ‘criminal versatility’, ‘promiscuous sexual behavior’), arguing that these are not primary symptoms of psychopathy but merely consequences of the disorder. Their three-factor model is consistent with the first three facets of Hare’s (2003) two-factor, four-facet structure (Cooke et al., 2004; Cooke & Michie, 2001).
2
http://www.efp.nl/efp-projecten/landelijke-databank-risicota xat ie-tbs (webpage only available in Dutch)
Table 2
Two-Factor & Four-Facet Model of the PCL-R. Interpersonal-affective factor
(Factor 1)
Impulsive-antisocial lifestyle factor (Factor 2)
Interpersonal facet Affective facet Impulsive lifestyle facet Antisocial facet ‘glibness/superficial charm’ ‘grandiose sense of self-worth’ ‘pathological lying’ ‘conning/manipulative’ ‘lack of remorse or guilt’ ‘shallow affect’ ‘callousness/ lack of empathy ‘failure to accept responsibility for own actions’ ‘need for stimulation/proneness to boredom’ ‘parasitic lifestyle’ ‘lack of realistic, long-term goals’ ‘impulsivity’ ‘irresponsibility’ ‘poor behavioral controls’ ‘early behavioral problems’ ‘juvenile delinquency’ ‘revocation of conditional release’ ‘criminal versatility’
Note. Two items are not assigned to any factor or facet: ‘promiscuous sexual behavior’ and ‘many short-term marital relationships’.
The PCL-R scoring manual (Hare, 2003) lists a number of requirements for assessors, such as work experience in the forensic field and supervised practice scorings. Additionally, an advanced degree in the social, medical, or behavioral sciences is recommended. In the current study, trained clinicians and researchers, employed by the clinic which treated the patient, scored the PCL-R based on semi-structured interviews and additional background information, e.g. criminal records (Psychopathie onderzoek en de LDR-tbs, n.d.). Items are scored on a three-point Likert scale (0 = item does not apply to the individual, 1 = item applies to a certain extent but not to the degree required for a score of 2, 2 = item does apply to the individual). The PCL-R handles a minimum total score of 0 and a maximum total score of 40 points. Total scores correspond to the degree of psychopathic features: The higher a person’s PCL-R total score, the more they resemble the prototypical psychopath (Hare, 2003). In clinical practice, a cutoff score of 26 (Europe) or 30 (US) is commonly handled to diagnose psychopathy (Spreen, Horst, Lutjehuis, & Brand, 2008).
According to the PCL-R manual, intra-class coefficient (ICC) values for total scores range from .86 to .88 (Hare, 2003; but see Edens, Boccaccini, & Johnson, 2010). Research on the PCL-R has also repeatedly shown a high predictive validity in terms of recidivism, violence and response to treatment (e.g. Hare, Clark, Grann, & Thornton, 2000; but see Yang, Wong, & Coid, 2010). The Dutch Association of Psychologists (het Nederlands Instituut van Psychologen or NIP) rated the PCL-R as sufficiently valid and reliable (COTAN, 2010).
Exclusion criteria and score adjustment. As recommended by the National Database for Risk Assessment TBS, only the most recent assessment was included in the current analyses. In case a patient had been assessed multiple times, the most recent
assessment was considered the most reliable because prior risk taxations were always taken into account if a patient was reassessed. However, for some assessments a date was not recorded. These were excluded if there was another assessment with a record of date
available or included if this was the only assessment available for a patient. Additionally, data from participants younger than 18 years was excluded from the analysis since the PCL-R is intended to assess persons aged 18 and above (Hare, 2003).
To score the PCL-R, extensive background knowledge of the assessed person is required. It is therefore not uncommon that interviewers are unable to score all of the 20 items due to insufficient information. To compensate for lower total scores due to missing data, total PCL-R scores were adjusted by prorating according to PCL-R guidelines (Hare, 2003). For example, a raw score of 20 points was adjusted to 21.1 points if one item was not scored, 22.2 points if two items were not scored and 26.7 points if five items were not scored. In case more than five items in total or more than two items of the same factor were not scored, a total score was considered unreliable, and the assessment was excluded from the analysis (Hare, 2003).
Data analysis. Using the R package qgraph by Epskamp, Cramer, Waldorp,
Schmittmann and Borsboom (2012), we plotted multiple networks consisting of nodes (PCL-R items) and edges (associations between items). Our (PCL-R-code is included in the supplemental materials (File 4) to encourage replication with additional samples. Node colors correspond to the two-factor structure of the PCL-R: We displayed Factor 1 items in red, pink or purple and Factor 2 items in green. The items corresponding to proposed core symptoms were additionally color coded to make them stand out. The network edges differ in size and color according to the strength or weight of an association: the thicker an edge, the stronger the association between two nodes. We displayed positive associations in black and negative associations in red. A spring-based algorithm for node placement was applied (Fruchtermann & Rheingold, 1991): Nodes with strong associations to other nodes were plotted in the center of the network while nodes with weak associations were plotted in the peripheral area of the network. Additionally, nodes that were strongly associated were placed in proximity to each other.
Correlation network. In the correlation network, edges corresponded to polychoric correlations between items. Polychoric correlations are a statistical computation method for associations between variables with a small range, e.g. questionnaires with few response options. The R package qgraph (Epskamp et al., 2012) automatically selects the best method for computation of correlations based on data input (Epskamp, Borsboom, & Fried, 2016). To aid the visual identification of nodes that are central to the network, we computed three measures of node centrality for the correlation network: betweenness, closeness, and strength. A higher score on these measures indicates greater centrality or relevance. The betweenness centrality measures how often a node lies on the shortest path between two other nodes. Nodes high in betweenness centrality often form bridges between node clusters. The
one node to all other nodes in the network divided by 1. Consequently, a node that is directly associated with all other nodes in the network would have a closeness centrality score of 1. Nodes that score high on closeness are usually the first ones affected by changes in other nodes. The strength of a node is the sum of all edge weights of a node. Nodes with a high strength score are considered highly influential: If a strong node is activated, many other nodes will be activated as well. Correlation networks are useful to get a first impression but also show spurious correlations, i.e. correlations that are confounded by another variable. This could be because both variables are caused by a third variable and therefore falsely seem related. For example, the items ‘poor behavioral controls’ and ‘criminal versatility’ might be connected by a strong edge in a correlation network, implying that poor behavioral controls directly cause criminal versatility or vice versa. However, it is possible that both of these psychopathy symptoms are caused by impulsivity and do not directly interact with each other. Another possible spurious correlation could be caused by one symptom causing another which, in turn, causes a third symptom. For example, impulsivity might cause poor
behavioral controls which, in turn, might cause criminal versatility. The items ‘impulsivity’ and ‘criminal versatility’ would then be connected to each other in a correlation network, but this connection might disappear after controlling for the item ‘poor behavioral controls’. To avoid drawing wrong conclusions based on spurious correlations, it can be useful to control for them.
Adaptive lasso network. To control for spurious correlations, we plotted a second type of network based on estimated partial correlations. Partial correlations are direct connections; i.e. correlations between two variables that persist even after controlling for all other possible variables. A drawback of this method is that one may overcontrol; if a variable is correlated with two others it may indicate confounding, but might also result from
which connections between items cannot be explained by confounding of other items. The adaptive lasso function, which is implemented in qgraph, was used to estimate a sparse network model based on partial correlations. The adaptive lasso automatically computes one hundred different models that are likely to fit the data and then selects the best fitting model to plot the network (for a more detailed explanation of the adaptive lasso, see Epskamp et al., 2016; Zou, 2006). This means that the adaptive lasso tells us how the true network could look, it does not disregard other possibilities (Epskamp, Kruis, & Marsman, 2016). It is important to keep in mind that the adaptive lasso function assumes that the true network is sparse and will therefore only display the minimum number of edges needed to explain the data (Epskamp, Kruis, & Marsmann, 2016). An advantage of a sparse network is that strong associations are more pronounced which improves interpretability. A disadvantage is that weak associations might be missed because the model disregarded them. Therefore, it is always best to compare the correlation network to the adaptive lasso network and identify similarities between them. To ensure that the adaptive lasso network would not overshadow results from the correlation network we did not compute any centrality measures for nodes in the adaptive lasso network.
Results
Missing data and duplicate cases. The original dataset contained 3,727 cases from 2,042 participants. We excluded 103 cases (3%) due to missing data and 1,643 duplicate cases (44%) from patients who had been assessed multiple times.
Participant characteristics. Nineteen patients (0.01%) did not meet the minimum age requirement of 18 years, leaving 1,962 unique cases (91% male; M age = 38.78 years, SD = 10.04) to be included in the analysis. Mean PCL-R scores are displayed in Table 3. About a quarter of patients in the Dutch sample (N=544, 28%) scored 26 points or higher. On average, older patients scored slightly lower than younger patients did, r(1960) = -.07, p < .05.
Table 3
Participant Characteristics (Sample 1: n = 1,962 ).
Measure M SD Range
Raw total score 20.57 7.15 1-40
Adjusted total score 20.98 7.31 1-40
Factor 1 Score (not adjusted) 8.72 3.43 0-16
Factor 2 Score (not adjusted) 9.67 4.17 0-18
Age (at time of assessment) 38.78 10.05 19-82
Network analysis. We report on the results of the network analysis in the following order: (1) global network structure (i.e. density, connectedness, and formation of clusters), (2) individual node centrality, and (3) associations between node pairs for all PCL-R items.
Global network structure. The correlation and adaptive lasso network are both densely connected (Fig. 1). The structure of the networks corresponds to the two-factor model of psychopathy (Hare, 2003): Items that belong to the same factor are grouped
together in clusters. The adaptive lasso network also shows two sub-clusters within Factor 1 corresponding to the interpersonal and the affective facet of the two-factor, four-facet model of psychopathy (Hare, 2003).
Figure 1. (a) Correlation network and (b) adaptive lasso network of the Dutch dataset. Abbreviations: F1 = Factor 1, LE = low empathy, GEB = general emotional blunting, F2 = Factor 2, AD = attentional deficit. Due to the limited space available here, the networks are displayed on a small scale. For larger images, see File 1 in the supplemental materials.
Node centrality. As apparent in Figure 1, the most central nodes in the correlation network are ‘conning/manipulative’, ‘callousness/lack of empathy’, ‘parasitic lifestyle’ and ‘irresponsibility’. ‘Conning/manipulative’ and ‘callousness/lack of empathy’ are not part of the network’s core anymore in the adaptive lasso network, which could suggest that the centrality of these items is partially based on spurious correlations (Fig. 1b). The measures of node centrality (Fig. 2) reveal that ‘glibness/superficial charm’ and ‘need for
stimulation/proneness to boredom’ are also central to the correlation network; these nodes score especially high on betweenness, indicating that they often lie on the shortest path from one node to another. This is also visible in the network plots (Fig. 1): ‘Glibness/superficial charm’ and ‘need for stimulation/proneness to boredom’ have many edges connected to them. The most peripheral nodes in the correlation network are ‘shallow affect’, ‘promiscuous sexual behavior’ and ‘many short-term marital relationships’ (Fig. 1a and Fig. 2). These nodes also have few (‘shallow affect’) or weak (‘promiscuous sexual behavior’ and ‘many short-term marital relationships’) associations with other nodes in the adaptive lasso network (Fig. 1b).
Figure 2. Measures of node centrality in the correlation network of Dutch PCL-R data. Displayed are betweenness, closeness, and strength for each node. All values are standardized for better interpretability. Items are abbreviated as shown in Table 5.
Node associations. In the correlation network, all edges display positive associations. The strongest edges connect ‘glibness/superficial charm’ with ‘grandiose sense of self-worth’ and ‘lack of remorse or guilt’ with ‘callousness/lack of empathy’ (Fig. 1a). These associations remain strong in the adaptive lasso network (Fig. 1b). Within the Factor 1 clusters, the
strongest edges connect items that belong to the same facet (Fig. 1). For a list of edge weights (correlation coefficients), see File 2 in the supplemental materials.
Analysis of proposed core symptoms. In the following section, we take a closer look at the PCL-R items corresponding to proposed core symptoms of psychopathy in terms of (1) node centrality and (2) associations between nodes.
Emotional deficits. The items corresponding to emotional deficits (‘lack of remorse or guilt’ and ‘callousness/lack of empathy’ consistent with the low empathy model of psychopathy and ‘shallow affect’ consistent with the general emotional blunting model of psychopathy) are placed in proximity to each other (Fig. 1). In the correlation network, these items are part of the Factor 1 cluster; in the adaptive lasso network, they form a sub-cluster together with ‘failure to accept responsibility for own actions’ (Fig. 1).
Node centrality. ‘Callousness/lack of empathy’ (low empathy model) is the most central emotional deficit item in both network types and among the most central items overall in the correlation network (Fig. 1 and Fig. 2). The item scores high on all three measures of node centrality (Fig. 2) which indicates that ‘callousness/lack of empathy’ interacts with many of the other PCL-R items. ‘Lack of remorse or guilt’ (low empathy model) also scores fairly high on the measures of node centrality (Fig. 2) and retains strong associations with other nodes in the adaptive lasso network (Fig. 1b). ‘Shallow affect’ (general emotional blunting model) is the least central of the emotional deficit items and one of the least central nodes overall; the item is located peripherally in the network and scores low on all three measures of node centrality (Fig. 1 and 2).
Node associations. The items corresponding to emotional deficits are connected to each other by thick edges, which indicates that they strongly interact with each other, even after controlling for all other items (Fig. 1). The edge between ‘lack of remorse or guilt’ and ‘callousness/lack of empathy’ (low empathy model) is especially strong in both networks (Fig. 1). The edge between ‘lack of remorse or guilt’ (low empathy model) and ‘shallow affect’ (general emotional blunting model) is only moderately strong in the adaptive lasso
network (Fig. 1b) which could mean that those items mostly interact with each other via ‘callousness/lack of empathy’ (low empathy model).
Attentional deficits. The items corresponding to attentional deficits (‘poor behavioral controls’ and ‘impulsivity’) are also placed in proximity to each other and are part of the Factor 2 cluster in both network types.
Node centrality. ‘Impulsivity’ is the most central attentional deficit item and located at the center of the Factor 2 cluster in both network types (Fig. 1 and Fig. 2). The item scores low on betweenness and average on closeness and strength (Fig. 2). ‘Poor behavioral
controls’ is located peripherally in both network types (Fig. 1) and scores low to average on all measures of node centrality (Fig. 2)
Node associations. ‘Impulsivity’ and ‘poor behavioral controls’ are connected by a thick edge, which indicates that these items strongly interact with each other, even after controlling for all other items (Fig. 1). Both items have weak connections to emotional deficits in the correlation network (Fig. 1a). In the adaptive lasso network, ‘impulsivity’ is very weakly associated with ‘shallow affect’ (general emotional blunting model) and not associated with empathy deficit items (‘lack of guilt or remorse’ and ‘callousness/lack of empathy’) (Fig. 1b). ‘Poor behavioral controls’ is weakly associated with ‘callousness/lack of empathy’ (low empathy model) but negatively associated with ‘lack of remorse or guilt’ (low empathy model) and not associated with ‘shallow affect’ (general emotional blunting model) (Fig. 1b).
Analysis 2: American Sample Method
Sample. The sample consisted of male offenders confined in Wisconsin state correctional institutes between 2000 and 2013. The majority of participants were recruited
from one of three facilities: Oakhill Correctional Institute, a minimum security facility (N=1,070; 27%); Oshkosh Correctional Institute, a medium-security facility (N=869, 21.9%) and Columbia Correctional Institute, a maximum-security facility (N=786; 19.8%).
Participants consisted of Caucasian (N = 2,356; 59%), African American (N = 1,524; 39%), Hispanic (N = 44; 1%), Native American (N = 20; 0.5%) and Asian (N = 1) offenders. The data have been reported upon in several papers by Joe Newman and colleagues (e.g. Brown et al., 2015; Glass & Newman, 2006).
Outcome measure. As in Analysis 1, psychopathy was measured with the
Psychopathy Checklist Revised (PCL-R; Hare, 2003) which we described in detail above. Participants were assessed by graduate students or research assistants, who received at least two months of training prior to the data collection (Newman, n.d.).
Exclusion criteria, score adjustment, and data analysis. The exclusion criteria, method of score adjustment, and data analysis were equal to those described for Analysis 1. The aim of the second analysis was to test whether the results from Analysis 1 could be replicated in a different sample. Therefore, in this analysis, we focused on identifying similarities and differences between the samples.
Results
Missing data and duplicate cases. The original dataset contained 4,221 unique cases. We had to exclude 250 cases (6%) due to missing values.
Participant characteristics. Eight participants (0.002%) did not meet the minimum age requirement of 18 years, leaving 3,963 cases (100% male; M age = 30.32 years, SD = 7.02) to be included in the analysis. Mean PCL-R scores are displayed in Table 4. Nearly half of participants (N = 1651, 42%) scored 26 points or higher. On average, older participants scored slightly lower compared to younger participants, r(3961) = -.05, p < .01.
Table 4
Network analysis. In the following section, we report upon which results from Analysis 1 could be replicated and which could not, in the following order: (1) global
network structure (i.e. density, connectedness, and formation of clusters), (2) individual node centrality, and (3) associations between node pairs.
Global network structure. Analysis 2 confirmed that the network of PCL-R items is densely connected. In comparison to the networks of the Dutch sample, the networks of the American sample are even more tightly connected, and the factor clusters are less distinct (Fig. 3). While items of the same facet are placed in proximity to each other, associations between items of the same factor or facet are not stronger than associations between items of two different factors or facets (Fig. 3).
Participant Characteristics.
Column1 Dutch sample
n = 1,962
American sample n = 3,963
Measure M SD Range M SD Range
Raw total score 20.57 7.15 1-40 22.98 6.95 2-40
Adjusted total score 20.98 7.31 1-40 23.36 7.05 2.10-40
Factor 1 Score (not adjusted) 8.72 3.43 0-16 8.77 3.40 0-16 Factor 2 Score (not adjusted) 9.67 4.17 0-18 10.83 3.59 0-18
Figure 3. (a) Correlation network and (b) adaptive lasso network of the Dutch (left) and American (right) PCL-R data. Abbreviations: F1 = Factor 1, LE = low empathy, GEB = general emotional blunting, F2 = Factor 2, AD = attentional deficit. Node numbers refer to PCL-R item numbers (see Table 5). Due to the limited space available here, the networks are displayed on a small scale. For larger images, see File 3 in the supplemental materials.
Table 5
PCL-R items and abbreviations.
PCL-R items (Hare, 1991; 2003) Abbreviation 1. Glibness/superficial charm
2. Grandiose sense of self-worth
3. Need for stimulation/proneness to boredom 4. Pathological lying
5. Conning/manipulative 6. Lack of remorse or guilt 7. Shallow affect
8. Callousness/lack of empathy 9. Parasitic lifestyle
10. Poor behavioral controls 11. Promiscuous sexual behavior 12. Early behavioral problems 13. Lack of realistic long-term goals 14. Impulsivity
15. Irresponsibility
16. Failure to accept responsibility for own actions 17. Many short-term marital relationships
18. Juvenile delinquency
19. Revocation of conditional release 20. Criminal versatility glib grndSW bored pathLie con/man guilt shaAff cal/emp paraLS pCtrl prmSex earlPrb ltGoals impuls irresp accResp stRships juvDel release crimVer
Node centrality. The replication analysis confirmed that ‘conning/manipulative’ and ‘callousness/lack of empathy’ are central to the network of PCL-R items, even after
controlling for all other items (Fig. 3 and 4). While ‘parasitic lifestyle’ retains its central position, the item’s associations with other items in the network are weaker compared to Analysis 1 (Fig. 3). This observation is confirmed by the node’s low scores on the measures of centrality (Fig. 4).
Figure 4. Measures of node centrality in the correlation networks of Dutch (red) and
American (green) PCL-R data. Displayed are betweenness, closeness, and strength for each node. All values are standardized for better interpretability. Items are abbreviated as shown in Table 5.
Node associations. In the networks for the American sample, the strongest association was equal to the strongest association in the networks for the Dutch sample (connecting ‘glibness/superficial charm’ with ‘grandiose sense of self-worth’, Fig. 3). The association connecting ‘lack of remorse or guilt’ with ‘callousness/lack of empathy’ is among the strongest in the networks for the American sample as well, but weaker than in the networks for the Dutch sample (Fig. 3). For a list of edge weights (correlation coefficients), see File 2 in the supplemental materials.
Analysis of proposed core symptoms. In the following section, we report upon which results concerning the (1) centrality of and (2) associations between PCL-R items corresponding to proposed core symptoms of psychopathy could be confirmed.
Emotional deficits. Node centrality. The centrality of emotional deficit items in the networks for the American sample is very similar to the centrality of emotional deficit items in the networks for the Dutch sample: ‘Callousness/lack of empathy’ (low empathy model) is the most central emotional deficit item, ‘lack of remorse or guilt’ (low empathy model) the second most central and ‘shallow affect’ (general emotional blunting model) the least central emotional deficit item in the PCL-R network (Fig. 3 and 4).
Node associations. In both samples, ‘callousness/lack of empathy’ (low empathy model) has strong associations with the other emotional deficit items (‘lack of remorse or guilt’ consistent with the low empathy model of psychopathy and ‘shallow affect’ consistent with the general emotional blunting model of psychopathy) (Fig. 3). The association between ‘lack of remorse or guilt’ (low empathy model) and ‘shallow affect’ (general emotional blunting model) is weaker compared to the associations with ‘callousness/lack of empathy’ (low empathy model), which supports the idea that ‘lack of remorse or guilt’ and ‘shallow affect’ mostly interact with each other via ‘callousness/lack of empathy’ (Fig. 3).
Attentional deficits. Node centrality. The centrality of ‘impulsivity’ is similar in both samples while the centrality of ‘poor behavioral controls’ differs. In the networks of
American PCL-R data, ‘poor behavioral controls’ is more central than ‘impulsivity’ (Fig. 3 and 4). Both attentional deficit items score low on ‘betweenness’ and average to moderately high on ‘closeness’ and ‘strength’ (Fig. 4).
Node associations. In contrast to the Dutch sample, the attentional deficit items are only weakly associated in the American sample (Fig. 3 and 4); both items have stronger associations to ‘callousness/lack of empathy’ (low empathy model) than to each other (Fig.
3). The association between ‘poor behavioral controls’ (attentional deficit model) and ‘callousness/lack of empathy’ (low empathy model) belongs to the strongest associations overall in both network types for the American sample (Fig. 3). The weak association between ‘impulsivity’ (attentional deficit model) and ‘shallow affect’ (general emotional blunting model) in the adaptive lasso network for the Dutch sample could be replicated.
Discussion
The current study aimed to identify the role of proposed core symptoms of psychopathy in the network of PCL-R items. Based on multiple theoretical models of psychopathy, we expected that (a) emotional deficits, (b) attentional deficits, or (c) a combination of the former would be central to the PCL-R network. We analyzed two large offender samples in which the PCL-R was administered to test whether results were stable. We found that PCL-R items differed substantially in their centrality and that ‘callousness/lack of empathy’ was among the most central item in both samples. The results suggest that lack of empathy is a core symptom of psychopathy as defined by the PCL-R.
Key Findings and Theoretical Implications
The PCL-R networks for the Dutch and the American sample differed from each other in many aspects. Nonetheless, the centrality of emotional deficits was fairly stable across samples. The network analyses revealed that ‘callousness/lack of empathy’ and ‘lack of remorse or guilt’, but not ‘shallow affect’, were central emotional deficit items in the PCL-R network. Furthermore, ‘callousness/lack of empathy’ was among the most central items overall and ‘shallow affect’ among the least central items overall in both samples. The results support Blair’s low empathy model of psychopathy (Blair et al., 2001; Blair, Jones, Clark, & Smith, 1997; Blair, 1992, 1995).
Results concerning items corresponding to the attentional deficit model of
suggest that attentional deficit items ‘impulsivity’ and ‘poor behavioral controls’ are at most moderately central to the network of PCL-R items.
Inconsistencies between Samples
As stated above, we identified a couple of differences in the networks for the Dutch and the American sample. For example, the networks for the American sample were more tightly connected and items belonging to the same factor of Hare’s (1991; 2003) two-factor structure were less clustered compared to the networks for the Dutch sample. There were also inconsistencies concerning the centrality of attentional deficit items: In the Dutch sample, ‘impulsivity’ was more central than ‘poor behavioral controls’ and the two items were
strongly associated; in the American sample, ‘poor behavioral controls’ was more central than ‘impulsivity’ and the two items were only moderately associated with each other. ‘Poor behavioral controls’ was also strongly associated with ‘callousness/lack of empathy’ in the American sample but not in the Dutch sample. We offer two possible explanations for these inconsistencies. First, the scoring range for PCL-R items is very limited which might result in big changes in the network due to small changes in the data. Second, it is possible that the PCL-R network differs between populations (e.g. high vs. low scorers, older vs. younger persons) or between psychopathy subtypes (e.g. primary and secondary psychopathy (Yildirim & Derksen, 2015)).
Implications for Psychopathy Assessment Instruments
Symptoms that are central to a symptom network are symptoms that are most likely to activate many other symptoms in the network (Fried et al., 2015; Schmittmann et al., 2013). This means that a person who scores high on a central item such as ‘callousness/lack of empathy’ is more likely to develop or exhibit other psychopathy symptoms than a person who scores high on a peripheral item such as ‘many short-term marital relationships’. Therefore, a lack of empathy should receive special attention in the diagnostic process.
Our results also have implications for the use of psychopathy measures for risk assessment purposes. Psychopathy is associated with violence and criminality, and the PCL-R is commonly used for risk assessment purposes in forensic settings (Porter et al., 2009; Skeem et al., 2011; Yildirim & Derksen, 2015). However, research suggests that the PCL-R performs quite poorly as a risk assessment tool compared to purpose-built instruments (Singh, Grann, & Fazel, 2011; Walters, 2004; Yang et al., 2010). Furthermore, a
comprehensive meta-analysis of risk assessment tools showed that PCL-R Factor 1 items, including low empathy items, predict violence only at chance level (Yang et al., 2010). If lack of empathy is a core symptom of psychopathy and psychopathy predicts violent criminality, we would expect that lack of empathy also predicts violent criminality. Therefore, we propose that we need to reassess the role of empathy in risk assessment.
Implications for Forensic Practice
Kossakowski and colleagues (2015) suggested that core symptoms hold important implications for the treatment of mental disorders. Treatment targeted at central symptoms is more likely to improve a wider range of symptoms than treatment targeted at peripheral symptoms (Kossakowski et al., 2015). Teaching someone to experience an emotion, such as empathy, might prove difficult, if not impossible, especially if there are neurological or genetic causes to the deficit like some research suggests (e.g. Seara-Cardoso et al., 2014). Nonetheless, if a lack of empathy can be identified as a core symptom of psychopathy, this knowledge could be used to improve the treatment and rehabilitation of criminal offenders with psychopathic tendencies. For example, one might try a more pragmatic approach and emphasize the negative consequences that a future offense will have for the offender, instead of focusing on the victims suffering or other moral aspects of the crime. Assuming that someone with a psychopathic personality will act upon egoistic motives, regardless of others, the best prevention of criminal and violent behavior in psychopathic individuals might be to
appeal to their selfishness. On the other hand, research suggests that psychopathic individuals can be empathic if they actively choose to be but that they lack the spontaneous empathy most people experience (Arbuckle & Cunningham, 2012; Meffert, Gazzola, Den Boer, Bartels, & Keysers, 2013). In this case, psychopathy treatment should be targeted at increasing motivation for activating empathy.
Limitations
This study is not without its limitations. First, although the sample sizes were large, we cannot conclude that our findings are generalizable to the forensic population as a whole, especially considering the inconsistencies between samples. Moreover, our results are limited to psychopathy as defined by the PCL-R. To draw reliable inferences, we first need to test whether our results can be replicated with the use of other psychopathy assessment
instruments. Second, the low fear model of psychopathy could not be tested because the PCL-R does not measure fearlessness. It is possible that fearlessness is a more central psychopathy symptom than lack of empathy which would mean that assessment and treatment should be targeted more at fearlessness. Third, the selection of items that are most consistent with the core symptoms of proposed theoretical models of psychopathy was partly subjective, and not all items can be clearly assigned to a certain theoretical model. For example, Gorenstein (1982) originally proposed that impulsivity and irresponsibility are core symptoms of the frontal lobe impairments thought to cause psychopathy symptoms. Based on the item descriptions in the PCL-R manual (Hare, 2003), we argue that ‘poor behavioral controls’ provides a better fit for the attentional deficit model of psychopathy. The item
‘irresponsibility’ on the other hand, “describes an individual who habitually fails to fulfill or honor obligations and commitments to others” (Hare, 2003, p. 43). While such behavior can certainly result from attentional deficits, it might also result from a general disregard for others (low empathy).
Future Research
Interesting questions for future research could be whether the network of PCL-R items differs for younger and older persons, men and women, or high and low psychopathic individuals. We suggest that future research should compute separate networks for each of these groups and compare them to identify similarities and differences between groups. It would also be interesting to compare the networks for individual persons to see if assessment and treatment of psychopathy should take individual differences into account. As stated above, it is important to test whether our results can be replicated using additional psychopathy measures, e.g. the PPI (Lilienfeld & Andrews, 1996). There are two big advantages of using the PPI instead of the PCL-R as an outcome measure: First, the PPI directly measures fearlessness, which the PCL-R does not; second, the PPI is not limited to the use with criminal offenders and can thus be used to measure psychopathy in community samples. Thus, a PPI network could be useful to compare the centrality of low fear items to the centrality of low empathy items and to compare community samples to offender samples. Summary
The current study offers some interesting new insights concerning core symptoms of psychopathy as defined by the PCL-R. The results suggest that a lack of empathy is a core symptom of psychopathy. However, it is yet too early to judge the scope and extent of these findings. In addition, not all findings are consistent across samples. Nonetheless, network analysis seems to be a promising addition to traditional approaches to psychopathy and provides many opportunities for future research.
References
American Psychiatric Association. (2013). Diagnostic and Statistical Manual of Mental Disorders: DSM-5. ManMag.
Arbuckle, N. L., & Cunningham, W. A. (2012). Understanding everyday psychopathy: Shared group identity leads to increased concern for others among undergraduates higher in psychopathy. Social Cognition, 30(5), 564–583.
http://doi.org/10.1521/soco.2012.30.5.564
Birbaumer, N., Veit, R., Lotze, M., Erb, M., Hermann, C., Grodd, W., & Flor, H. (2005). Deficient fear conditioning in psychopathy: a functional magnetic resonance imaging study. Archives of General Psychiatry, 62(7), 799–805.
Blair, R. J. R. (1995). A cognitive developmental approach to morality: investigating the psychopath. Cognition, 57(1), 1–29. http://doi.org/10.1016/0010-0277(95)00676-P Blair, R. J. R. (2013). Psychopathy: cognitive and neural dysfunction. Dialogues in Clinical
Neuroscience, 15(2), 181–190. Retrieved from
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3811089/
Blair, R. J. R. (1992). The development of morality. University of London.
Blair, R. J. R., Colledge, E., Murray, L., & Mitchell, D. G. V. (2001). A selective impairment in the processing of sad and fearful expressions in children with psychopathic
tendencies. Journal of abnormal child psychology, 29(6), 491-498.
Blair, R. J. R., Jones, L., Clark, F., & Smith, M. (1997). The psychopathic individual: A lack of responsiveness to distress cues? Psychophysiology, 34(2), 192–198.
http://doi.org/10.1111/j.1469-8986.1997.tb02131.x
Blair, R. J. R., Mitchell, D. G. V, Leonard, A., Budhani, S., Peschardt, K. S., & Newman, C. (2004). Passive avoidance learning in individuals with psychopathy: Modulation by reward but not by punishment. Personality and Individual Differences, 37(6), 1179– 1192. http://doi.org/10.1016/j.paid.2003.12.001
Boccardi, M., Frisoni, G., Hare, R., Cavedo, E., Najt, P., Pievani, M., … Tiihonen, J. (2011). Cortex and amygdala morphology in psychopathy. Psychiatry Research -
Neuroimaging, 193(2), 85–92. http://doi.org/10.1016/j.pscychresns.2010.12.013 Borsboom, D., Cramer, A. O., Schmittmann, V. D., Epskamp, S., & Waldorp, L. J. (2011).
The small world of psychopathology. PloS one, 6(11), e27407.
Borsboom, D., & Cramer, A.O. (2013). Network analysis: an integrative approach to the structure of psychopathology. Annual review of clinical psychology, 9, 91-121. Brown, A. R., Dargis, M. A., Mattern, A. C., Tsonis, M. A., & Newman, J. P. (2015). Elevated
psychopathy scores among mixed sexual offenders: replication and extension. Criminal Justice and Behavior, 42(10), 1032–1044.
http://doi.org/10.1177/0093854815575389
Campbell, W. M., Dagli, C. K., & Weinstein, C. J. (2013). Social network analysis with content and graphs. Lincoln Laboratory Journal, 20(1), 20.
Casey, H., Rogers, R. D., Burns, T., & Yiend, J. (2013). Emotion regulation in psychopathy. Biological Psychology, 92(3), 541–548.
http://doi.org/10.1016/j.biopsycho.2012.06.011
Cleckley, H. (1976). The mask of sanity (5th ed.). St. Louis, MO: Mosby.
Cooke, D. J., & Michie, C. (2001). Refining the construct of psychopathy: Towards a hierarchical model. Psychological Assessment, 13(2), 171.
Cooke, D. J., Michie, C., Hart, S. D., & Clark, D. A. (2004). Reconstructing psychopathy: Clarifying the significance of antisocial and socially deviant behavior in the diagnosis of psychopathic personality disorder. Journal of Personality Disorders, 18(4), 337. Cramer, A.O.J., Borsboom, D., Aggen, S.H., & Kendler, K.S. (2011). The pathoplasticity of
dysphoric episodes: Differential impact of stressful life events on the pattern of depressive symptom inter-correlations. Psychological Medicine.
doi:10.1017/S003329171100211X
Cramer, A. O., Waldorp, L. J., van der Maas, H. L., & Borsboom, D. (2010). Comorbidity: A network perspective. Behavioral and Brain Sciences, 33(2-3), 137-150.
D’Silva, K., Duggan, C., & McCarthy, L. (2004). Does treatment really make psychopaths worse? A review of the evidence. Journal of Personality Disorders, 18(2), 163–177. http://doi.org/10.1521/pedi.18.2.163.32775
Devonshire, P. A., Howard, R. C., & Sellars, C. (1988). Frontal lobe functions and personality in mentally abnormal offenders. Personality and Individual Differences, 9(2), 339– 344. http://doi.org/10.1016/0191-8869(88)90096-7
Edens, J. F., Boccaccini, M. T., & Johnson, D. W. (2010). Inter-rater reliability of the PCL-R total and factor scores among psychopathic sex offenders: Are personality features more prone to disagreement than behavioral features? Behavioral Sciences and the Law, 28(1), 106–119. http://doi.org/10.1002/bsl.918
Egberink, I.J.L., Janssen, N.A.M., & Vermeulen, C.S.M. (2015). COTAN beoordeling 2010, Nederlandstalige bewerking van de Hare Psychopathy Checklist-Revised van R.D. Hare (1991) [COTAN review 2010, Hare Psychopathy Checklist-Revised]. Retrieved from www.cotandocumentatie.nl
