Sub-groups of PD in relation to violence
While an association between PD and violence has been well documented [1, 2], the literature highlights the importance of considering particular sub-groups of patients with PD that are characterised by particular patterns of comorbidity and gender. For example, results of a meta-regression analysis  emphasised that the relationship between PD and violence is not straightforward and varies by PD category and by gender. Thus while the overall OR for PD was 3.0, the risk of violence was highest for those with antisocial PD, particularly in women (OR 13.1) compared with men (OR 7.9). In non-forensic clinical samples, co-occurrence of antisocial and borderline PDs is seen more frequently in men than in women [29, 30]. However, in forensic samples, particularly those at the high-severe end of the PD spectrum (e.g., women classified as having “dangerous and severe PD”), antisocial PD comorbid with borderline PD occurs more commonly in women than in men . The higher risk of violence in women with antisocial PD compared with men is therefore likely accounted for by their showing a higher co-occurrence of borderline PD. This can only be surmised, however, since the above-mentioned meta-regression analysis  did not examine comorbidity of PDs in relation to violence.
Subsequent research has confirmed the importance of examining subgroups of PD individuals characterised by particular patterns of comorbidity, as well as gender and ethnicity. A recent study of violence perpetrated by American prison inmates, both male and female and of black and white ethnicity, reported that regardless of ethnicity, those with co-occurring psychopathy and antisocial PD were almost twice as likely, compared with other inmates, to have a history of severe and versatile violent offending . Violent offending was highest in black males and females with comorbid antisocial PD and psychopathy, pointing to the importance of both gender and ethnicity in addition to PD comorbidity in rates of violence among offenders. In patients recruited as part of the McArthur study, borderline PD with co-occurring psychopathic traits was associated with violence during a one-year study period . A triple comorbidity – antisocial PD with co-occurring borderline PD and psychopathy – was found to be associated with the highest rates of severe violent offending among men diagnosed with PD and detained in medium or high security in the UK . The co-occurrence of antisocial personality and borderline PD in a UK household sample was significantly associated with a history of violence, but this was largely, although not entirely, accounted for by co-occurring alcohol dependence, anxiety disorder and severe childhood conduct disorder (CD) . A study that compared non-violent men with violent men who were, or were not, gang members, reported very high levels of psychiatric morbidity (with the exception of depression) in both the latter groups but particularly in gang members . Compared with non-violent men, violent men who were not gang members were more likely to show psychosis (OR 2.9), anxiety (OR 1.8), alcohol dependence (OR 1.6) and antisocial PD (OR 8.8), and to have made use of psychiatric services (ORs 1.9 –2.7). Equivalent ORs for gang members, who showed the highest level of violence, were 4.2, 2.2, 6.5, 57 and 4.3–7.8. A large proportion of violent men who were gang members reported being excited by violence (63 %) and using violence instrumentally (73 %), suggesting that in terms of the quadripartite violence typology (Fig. 1), gang members’ violence was most often of the impulsive/appetitive type.
In short, while a clear relationship appears to exist between personality disorder (e.g., antisocial PD when this co-occurs with other PDs, and particularly with borderline PD) and violent offending in general, rates of violent offending differ by degree of psychiatric morbidity, by gender and by ethnicity. In general, the greater the psychiatric morbidity and comorbidity, the greater is the risk of violence. This suggests that risk of violence may be related to overall severity of psychopathology (p).
Impulsiveness and dysregulated affect
Impulsiveness can broadly be defined as a predisposition to react rapidly and without planning to internal and external stimuli with lack of regard for short-term and long-term consequences for oneself and others . It is considered to be a symptom of many psychiatric disorders including borderline and antisocial PDs, bipolar disorder, attention deficit/hyperactivity disorder, conduct disorder and substance abuse/dependence. Although impulsiveness has been commonly assumed to be linked to violence, this link is questionable, particularly in psychosis where comorbidity is a significant issue [35, 36].
The variable findings in the field are likely accounted for by the heterogeneity of both violence and impulsiveness. First, as discussed above, not all violence is impulsive, and not all impulsive violence is motivated in the same way (see Fig. 1). Secondly, impulsiveness is multifaceted, incorporating a number of dimensions, including a tendency to act rashly and intemperately under the pressure of positive or negative emotions . When behaving in an emotionally impulsive way, the individual responds to a stimulus or event on the basis of an immediate emotional reaction such as desire or anger, with little if any checking of long-term consequences . It is apparent that impulsive violence as defined in the typology outlined in Fig. 1 is related specifically to emotional impulsiveness rather than to other aspects of impulsiveness such as a tendency to think or act rashly. Measures of impulsiveness, both self-report and behavioural, are limited in the degree to which they tap emotional impulsiveness. For example, a commonly used self-report measure of impulsiveness, the Barratt Impulsivity Scale (BIS) , does not include an explicitly emotional component.
The UPPS model and measures of impulsive behaviour
The UPPS is “a promising measure and model of impulsivity because it conceptualises and assesses impulsivity as a multifaceted construct that includes various, separable, and distinct pathways to impulsive behaviour….” ( p.4). UPPS includes a scale, negative Urgency, which reflects “a tendency to experience strong impulses, frequently under conditions of negative affect” ( p. 685). Subsequently UPPS was revised to include a positive Urgency scale to reflect impulsive behaviour occurring in the context of positive affect. Positive and negative Urgency were found to correlate highly and were considered as a unitary scale in a study that examined relationships between UPPS scales (Urgency, [lack of] Perseverence, [lack of] Premeditation, and Sensation Seeking), and DSM-5 PDs assessed both categorically (DSM-5 section 2) and by traits (DSM-5 section 3) . Urgency correlated most strongly with PD traits – with 3 of the 5 trait domains (Negative Affectivity, Antagonism and Disinhibition), and with 14 of 25 lower-order traits. This lack of discriminant validity suggests that, rather than reflecting specific PD types, Urgency reflects overall PD severity. Both Urgency and (lack of) Premeditation correlated with antisocial and borderline PDs, while Sensation Seeking correlated only with Histrionic and Narcissistic PDs. The authors suggested that (lack of) Premeditation and Urgency may help explain the high rates of externalising behaviours associated with antisocial and borderline PDs. Supporting this, it was found that the incidence of serious physical violence committed by psychiatric inpatients was increased threefold in those who scored high on Urgency, and was nearly two times higher in those with PD (specific types of PD were not examined in this study) . Another study  found that a composite measure of serious violence comprising serious violence in the criminal record, early onset of violent behaviour and serious institutional violence correlated significantly with both Urgency and (lack of) Premeditation. A subsequent closer examination of the data from this study (unpublished observations, Howard and Khalifa) showed that Urgency correlated significantly (p < .01) and positively with dimensional scores of 5 out of 10 PD categories (paranoid, antisocial, borderline, histrionic and dependent) and with overall PD severity defined by the combination of internalizing and externalizing PD traits. Exceptionally, Urgency correlated significantly (p < .05) and negatively with schizoid PD. Unlike antisocial PD, which was significantly associated with all UPPS facets, borderline PD did not correlate significantly with Sensation Seeking. This suggests that borderline PD is distinguishable from antisocial PD in lacking the excitement seeking facet of Externalizing that characterises antisocial PD .
Emotion dysregulation, also known as affective instability or emotional lability, is a key feature of borderline PD and is a lower-order trait within the domain of Negative Affectivity in the DSM-5 section 3 trait-based typology. However, the high overlap beween DSM-5 domains has been noted above, and high and significant correlations were reported between emotional lability and many traits across other domains . Affective instability overlaps both conceptually and empirically with affective impulsiveness or Urgency  and correlates significantly and positively with FFM Neuroticism and negatively with both Agreeableness and Conscientiousness . Evidence suggests that emotion dysregulation mediates the link between borderline PD and violence [43, 44]. The latter study found a relationship between emotional dysregulation and perpetration of physical assault that was mediated by emotion dysregulation but not by trait impulsiveness. Notably, however, trait impulsiveness was measured using an FFM-based metric that predominantly tapped non-emotional facets of impulsiveness such as self-discipline and deliberation.
It may be concluded from these studies that emotional impulsiveness – the tendency to respond rashly and to act out under pressure of high emotional arousal - is the facet of impulsiveness that is most relevant to the link between PD (particularly borderline PD) and violence. However, it appears likely that emotional impulsiveness contributes importantly to overall PD severity. Therefore it is considered likely that it is PD severity, rather than emotional impulsiveness or emotion dysregulation per se, that accounts for the high degree of violence associated with borderline PD. Small wonder, then, that a physiological correlate of emotional impulsiveness was successful in identifying those mentally disordered offenders who were at high risk of re-offending violently upon their release from secure care into the community . Evidence that this physiological correlate is changeable, and is therefore a dynamic rather than static risk factor, offers hope that intervention pre-release can potentially reduce the risk of violent re-offending .
Psychopathy as conceptualised and measured by the currently most used instrument, the Psychopathy Checklist (PCL) is a much-debated construct  and is not accepted as a valid psychiatric condition by all forensic psychiatrists (e.g., ). Nonetheless, it is commonly regarded as a personality disorder and has entered the nomenclature of DSM-5 under the label “Antisocial/Psychopathic Personality Disorder”  where, in section 3, a psychopathic features specifier has been included to designate the classically low-anxious, socially assertive variant of antisocial personality described in the adult psychopathy literature [49–52].
The most recent, 20-item revision of the PCL (PCL-R) consists largely of items assessing dispositions such as impulsiveness and behaviours such as pathological lying and criminal versatility . It yields, in addition to a total score, scores on two factors – selfish, callous and remorseless use of others (F1), and chronically unstable and antisocial lifestyle (F2). Each factor subsumes two facets, so that F1 subsumes the facets Interpersonal and Affective, while F2 subsumes the facets Lifestyle and Antisocial. Some authors have warned against reliance on the total PCL-R score to “diagnose” psychopathy. According to Lilienfeld and colleagues reliance on total PCL-R scores “…is no longer defensible given that the subdimensions of most psychopathy measures are associated with substantially different personality correlates” (, p.30). Howard & Duggan suggested that given the high heterogeneity within the class of high PCL-R scorers, “a high score on the PCL may tell us little more about the individual than that he or she is a high PCL scorer – one may just as well call him or her ‘a bastard’..” (, p. 284). It is not surprising, therefore, that the PCL-R factors are differentially related to violence, with Factor 2 showing a considerably stronger relationship than Factor 1, as reviewed in . These authors conclude: “…. it appears to be the combination of criminal behaviour and poor behavioural controls (irritability, aggression and inadequate control of anger) that, among traits tapped by the PCL, accurately predicts future violent and non‐violent offending” (p.284).
Studies concur in showing that PCL-R psychopathy maps onto normal personality traits measured using the FFM, and that central to it is (lack of) Agreeableness. PCL psychopathy appears to be largely an admixture of low Agreeableness and low Conscientiousness, with varying contributions made to the PCL-R factors by Neuroticism and Extroversion [54, 56]. Similarly, psychopathy appears to be embedded within, and to extend across, several categories of PD, in particular antisocial and narcissistic [57, 58]. We have noted above the relationship of the general psychopathology factor (p) with low Agreeableness, low Conscientiousness and high Neuroticism. We must consider therefore the possibility that this particularly toxic combination of personality traits reflects overall severity of psychopathology, and that “psychopathy”, rather than being a unitary disorder, represents a variable constellation of traits that contribute importantly to overall severity of psychopathology. Supporting this, PCL-R scores were reported to be significantly associated with a measure of overall PD severity, obtained by summing across individual PD criteria . A measure of psychopathy derived from an assessment of PD, labelled “acting out” , also correlated with PD severity and, in regression analysis, predicted a high degree of severe violence in the criminal records of personality disordered offenders .
Given the heterogeneity of violence outlined above, we must still ask whether particular variants, or components, of psychopathy might contribute to different types of violence. Unfortunately, studies looking at violence through the lens of QVT in different subtypes of psychopath are still to be done. Evidence reviewed above suggests that violence to satisfy a lust for excitement is common in delinquent youth and in violent gang members, 86 % of whom qualified for antisocial PD . Narcissistic PD has been reported to be strongly related to causing pain and suffering to others, and this relationship was significant even when other Cluster B personality disorders were controlled . However, the motivation here is unclear, and case reports of men with narcissistic PD suggest their violence is triggered by a slight or insult and is motivated by a desire for vengeance (see for example Case 4 in ). In terms of the violence typology shown in Fig. 1, this would clearly correspond to the controlled/aversive violence type, but this type of violence may be more characteristic of the vulnerable narcissist than the grandiose narcissist. While both grandiose and vulnerable sub-types are characterized by low FFM Agreeableness, the vulnerable subtype is additionally associated with prominent Neuroticism traits (e.g., shame, need for admiration) and low Extraversion, while the grandiose sub-type is associated with high Agentic Extraversion (e.g., exhibitionism, authoritativeness); the DSM construct of narcissistic PD captures a mixture of these two sub-types . Borderline and antisocial features of PD may be closely linked to aversively motivated violence, both in its controlled, premeditated form where the motivation is revenge, and in its impulsive form where it is motivated by removal of an immediate interpersonal threat. Violence when it (rarely) occurs in the classic Cleckleyan  manifestation of psychopathy may be associated with avarice – greed for material objects or social dominance.
Delusional ideation in PD patients
Despite deficits in the cognitive domain being a core area of deficit in the PDs, delusional thinking has been relatively neglected as a possible mediator of violence in personality disorders, notwithstanding its well-documented presence in PD. In borderline PD a “quasi-psychotic” thought disturbance is common, for example a delusional belief in imminent abandonment by a romantic partner or health professional . Recent evidence suggests that delusional ideation is related to overall severity, rather than type, of PD, and that severe violence in forensic patients is related to both severity of PD and degree of delusional thinking . This is consistent with evidence that severity of PD is related to metacognitive deficits that include an impaired ability to recognize the subjective nature of one’s thoughts and to achieve a critical distance when considering one’s beliefs . This would necessarily result in idiosyncratic interpretations of external reality, and would likely result in the types of deficit in social cognition seen in BPD patients, namely: a tendency to misinterpret neutral situations, to feel socially rejected during normative inclusion conditions, and to have difficulty restoring cooperation after experiencing disappointment . A bias towards interpreting neutral or ambiguous social encounters as threatening, as demonstrated for example in , would impact negatively on borderline patients’ everyday social interactions and predispose them to react to interpersonal stress with aggression and violence. They would be particularly susceptible to the impulsive/aversive type of violence (Fig. 1) that is associated with interpersonal threat.
Recent evidence suggests that delusions implying threat or harm to the individual are associated with angry affect, and that angry affect due to delusions mediates the latter’s relationship with serious violence . It therefore seems likely that angry affect resulting from threat-related delusional thinking, and in particular the inability to regulate that affect, may be a critical link in the pathway leading from psychosis to serious violence. The degree of conviction with which delusions are held may be a factor in determining the emotional response .
Separate pathways from internalizing and externalizing to violence
As noted above, a study of forensic psychiatric patients with PD found a relationship between severity of delusional ideation and severe violence . Regression analysis revealed that severe violence was predicted by high scores on two trans-diagnostic dimensions of PD, “acting out” (equivalent to psychopathy) and “anxious-inhibited” (equivalent to neurotic introversion). However, while “acting out” independently predicted severe violence, the effect of “anxious-inhibited” on violence appeared to be mediated by delusional ideation. This result mirrors those from the New Zealand cohort study  referred to above, where a general psychopathology factor (p) linked to delusional thinking and an Externalizing factor emerged as independently associated with violent offending – both correlated significantly and positively with violent conviction. Internalizing was inversely associated with violence when p was taken into account. Externalizing was positively associated with FFM Extraversion but not associated with Neuroticism, while p was strongly associated with Neuroticism and inversely associated with Agreeableness and Conscientiousness.
Taken together, these results suggest that separate developmental pathways might link Externalising and Thought Disorder aspects of PD with violence. One pathway, operating via delusional thinking and general psychopathology (p), would result from extremely high and/or inflexible levels of negative emotionality, empathy and rumination experienced in the context of intense and/or chronic interpersonal stress during adolescence, particularly in females . Another, externalising pathway would lead from severe conduct disorder in early childhood, particularly when this co-occurs with callous-unemotional traits , to adult antisocial personality, via effects of alcohol misuse in early adolescence on adolescent brain development [10, 71, 72]. These pathways would not, of course, be mutually exclusive. Indeed, both pathways would be expected to operate conjunctively in those with an end-point of severe PD (e.g., those with co-occurring antisocial and borderline PD).