Traits of empathy and anger: implications for psychopathy and other disorders associated with aggression

Abstract

Empathy and anger are two social emotions that modulate an individual's risk for aggression. Empathy is an emotional reaction to another individual's emotional state. Anger is an emotional reaction to threat, frustration or social provocation. Reduced empathy, seen in psychopathy, increases the risk for goal-directed aggression. Atypically increased anger (i.e. irritability), seen in conditions like disruptive mood dysregulation disorder and borderline personality disorder, increases the risk for reactive aggression. In this paper, I will outline core neurocognitive functions that correspond to empathy and which are compromised in individuals with psychopathic traits. In addition, I will outline neurocognitive functions involved in either the generation or regulation of anger and which are compromised in psychiatric conditions at increased risk for irritability/reactive aggression. It can be hoped that improved understanding of empathy and anger will lead to better assessment tools and improved interventions to reduce aggression risk.

This article is part of the theme issue ‘Diverse perspectives on diversity: multi-disciplinary approaches to taxonomies of individual differences’.

1. Introduction

Aggression can be defined as behaviour directed towards harming or injuring another living being who is motivated to avoid such treatment. It is a natural and adaptive part of the mammalian social behavioural repertoire. However, it can become maladaptive if it is exaggerated, persistent or expressed out of context [1]. Decreased empathy and increased anger are both associated with maladaptive aggression. Forms of empathy (e.g. the termination of aggression towards an animal displaying submission cues, cf. [2]) and anger/irritability (cf. the work of Amsel [3]) are also part of the mammalian social behavioural repertoire. The goal of this paper is to consider the neurobiological underpinnings of these traits and relate them to disorders associated with aggression. However, before continuing, two different forms of aggression should be distinguished—instrumental and reactive—as empathy and anger have differential impacts on their relative incidences.

(a). Instrumental and reactive aggression

In work on human aggression, a fundamental distinction has been drawn between instrumental (aka proactive/planned) and reactive (aka affective/defensive/impulsive) aggression [4]. Instrumental aggression is planned. The aggressor anticipates that the act will have a positive outcome (increased resources or social status or gratification of a perceived need). It usually occurs in the absence of intense emotion. It will be argued below that empathy dysfunction increases the risk for instrumental aggression.

Reactive aggression is unplanned and is often characterized as impulsive. It usually occurs amidst negative emotions (anger, sadness, frustration and irritation). It will be argued below that increased anger propensity increases the risk for reactive aggression.

It is relatively easy to identify individuals at selective risk for reactive aggression [5]. Indeed, several psychiatric conditions are associated with a selective risk for reactive aggression: e.g. intermittent explosive disorder (IED), disruptive mood dysregulation disorder (DMDD) and borderline personality disorder (BPD) [6–8]. Patients with these conditions show elevated levels of anger [9–11].

By contrast, individuals with a selective risk for instrumental aggression are not commonly seen [5]. Instead, individuals who show high levels of instrumental aggression typically also show high levels of reactive aggression [5]. The only clinical condition associated with an increased risk for instrumental aggression, psychopathy, is also at elevated risk for reactive aggression [12,13]. Notably, the clinical description of psychopathy emphasizes an individual with reduced empathy but intact or possibly exaggerated anger [14].

2. Empathy

Many different definitions of empathy exist (for a review, see [15]); researchers refer to very different neurocognitive functions with the same term. Rather than specify an additional definition, this paper will concentrate on specific neurocognitive processes, which might be considered empathic, that decrease the likelihood of aggression.

The first thing to consider is how might empathy decrease the likelihood of aggression. Perhaps, simply experiencing other individuals’ emotional states motivates aggression avoidance? How then, might one generate the experienced empathic state? Up until recently, an influential view has been that this occurs via representation sharing. In short, observation of another individual's emotion triggers activation of matching neural substrates in the observer allowing the displayer's emotion to be understood [16,17]; i.e. when I see someone who is sad, I activate the same neural substrate that responds when I am sad. This view was developed from data indicating mirror neurons in animals [18] and emerged principally from the studies of witnessed and experienced pain; both conditions have been consistently associated with activity within activation of the same neural substrate—dorsomedial frontal and anterior insula cortices (dmFC and aIC; for a meta-analytic review, see [19]).

While this view remains influential, it has been extensively challenged [20]. Notably, dmFC and aIC show co-activation to many classes of stimuli with a meta-analysis of over 3500 neuroimaging studies reporting that their activation was among the most frequently observed findings across a wide variety of tasks [21]. Moreover, spatially sensitive multi-voxel pattern analysis techniques reveal that while experienced and witnessed pain evoke shared activity patterns within dmFC and left aIC, these same shared activity patterns are also seen to experienced and witnessed disgust and unfairness [22]. On the basis of these data, if the experience of another's emotional state comes through shared activity patterns, how could the individual distinguish between pain, disgust and fairness? In addition, support vector machine (SVM) classifiers can be used to identify neural patterns associated with experienced (high versus low) and witnessed (high versus low) pain [23]. Notably, these classifiers are at chance in predicting the other experience; i.e. the classifier distinguishing the neural response to high versus low experienced pain is at chance at identifying whether a response to witnessed pain is at high or low intensity. In short: (i) regions (dmFC and aIC) showing a response to seeing another in pain and direct pain experience also respond to witnessed disgust/unfairness and many other task parameters; and (ii) SVM analyses reveal separable neural responses to experienced and witnessed pain [23].

In short, the position here is that another individual's emotional state is not identified by shared representations. Instead, it is assumed that another's emotional state is identified as any other stimulus is identified; i.e. determined through prior learning as a function of sensory cues (e.g. the facial expression of sadness (a nameable cue as the sight of a pencil is a nameable cue)) and semantic information (e.g. the context is a funeral rather than a wedding). Of course, such an identification process would not motivate the individual to decrease aggression. Instead, the argument is that there are three empathic mechanistic processes that all probably decrease the likelihood of aggression.

(a). Empathic reactions to the distress of others inhibit aggression

The idea that a victim's distress cues can inhibit the aggressor's attack was the basis of the cognitive Violence Inhibition Mechanism model [24]. This was a cognitive model that, at the neural level, relied on the specified functional architecture of the amygdala [25]. Considerable literature demonstrates that the amygdala is responsive to distress cues (fear, sadness and pain expressions; for a review, see [26]). Moreover, low-level stimulation of the amygdala by aversive stimuli can, via connections to the hypothalamus and periaqueductal gray (PAG), initiate freezing [27].

(b). Empathic learning about actions that harm others leads to these actions acquiring negative valence and being avoided

Emotional expressions are communicatory displays that provide valence information to observers [28]. In many cases, this valence information initiates stimulus–reinforcement or response–outcome association formation. Valence-based learning on the basis of the emotional expressions of others is well established in the context of observational fear learning [29] or social referencing [30] paradigms. In these paradigms, an infant observes the reaction of a caregiver to a novel object. If the caregiver reacts positively, the infant is significantly more likely to approach the object than if the caregiver shows a negative reaction/distress. Animal work has shown that social referencing is disrupted by amygdala damage [31]. Recent functional magnetic resonance imaging (fMRI) work has also implicated the amygdala in social referencing in humans and shown that the amygdala is sensitive to expression prediction errors (i.e. the degree to which the expression induced by an object deviates from the expected emotional reaction [32]). Prediction error sensitivity is important. Greater learning (a greater change in the value associated with the object) occurs in response to greater prediction errors [33]. Behaviourally, participants are more likely to approach objects associated with others' happiness than objects associated with others’ fear [32]. Interestingly, this vicarious conditioning is increased if participants are asked to enhance their empathic responses to the displayer and/or if the participant is high in trait empathy [34].

(c). Empathy involves deciding about actions that have negative value due to their association with others' distress

While the amygdala is critical for responding to the distress of others and enables learning of the value of representations through stimulus–reinforcement learning, studies examining empathic concern identify regions beyond the amygdala. These include ventral, rostral and dmFC, posterior cingulate and aIC [19,23]. Within these regions, ventromedial prefrontal (vmPFC) and posterior cingulate cortices (pCC) stand out. Activity within these regions is associated with the tendency for an individual to feel tender (rather than distressed) when hearing the biographical details of a suffering other [35]. Moreover, several studies have reported that empathizing with another individual experiencing positive events, including the receipt of money, is associated with activity within vmPFC (for a meta-analytic review of this literature, see [36]). Moreover, vmPFC activity to other individuals’ reward is positively related to the observer's self-reported empathy [37]. By contrast, empathizing with another experiencing negative events is associated with activity within dmFC and aIC [38,39].

These features are interesting to note because all these brain regions (ventral, rostral and dmFC, pCC and aIC) have been implicated in the representation of expected value/reinforcement expectancies [40–43]. These regions are implicated in using expected value in response selection (particularly ventral/rostral frontal cortex and pCC [40,41]) and organizing avoidance responses [42,43]. Moreover, they are consistently implicated in moral judgement tasks (for a meta-analytic review, see [44]). In short, empathic concern and related moral decision-making may involve representation of the expected value of the others' experience; i.e. vmPFC would code the expected value corresponding to the ‘feelings of tenderness’ when hearing the biographical details of a suffering other (cf. [35]) or a potential moral act (cf. [44]). This valence information may guide the individual to choose to avoid causing harm to another or guide them to choose a helping response option (see also, §3).

(d). To what extent are these functions compromised in psychopathy?

(i). Evidence for reduced responding to distress cues in psychopathy

Individuals with psychopathy have long been reported to show deficient empathy [45]. It has been suggested that this deficient empathy reflects a reduced responsiveness to the distress cues of other individuals [24,46]. The individual with psychopathy is thus more likely to proceed with an aggressive act than an individual who finds the distress of other individuals aversive.

In line with this view, both youth and adults with psychopathic traits show deficits in expression recognition—particularly for fearful and sad expressions—while the recognition of disgusted and angry expressions remains intact [47,48]. They further show reduced autonomic reactivity [49–51] and amygdala responses [52–57] to expressions of fear, sadness and pain. Given the argument made here that empathy dysfunction particularly increases the risk for instrumental aggression, it is worth noting that the reduced amygdala response to other's fear mediates the observed relationship between psychopathic traits and level of instrumental aggression [58].

(ii). Evidence for reduced vicarious aversive conditioning in psychopathy

Currently, there are no clear data that youth and adults with psychopathic traits show impaired vicarious aversive conditioning. However, it would be surprising if they do not. This is because studies have shown that individuals with psychopathy show impaired aversive conditioning generally (e.g. [59]) and reduced amygdala responses to conditioned stimuli relative to comparison individuals [60]. Given this evidence of generally impaired stimulus–reinforcement association formation, it would be surprising if they were selectively intact for stimulus–reinforcement associations where the reinforcement was social (i.e. others' distress) particularly given their reduced responsiveness to others’ distress (see above).

(iii). Evidence for reduced empathic decision-making in psychopathy

A series of studies have indicated the importance of vmPFC for empathic concern (see above). Studies investigating empathic concern in youth and adults with psychopathic traits have consistently revealed reduced vmPFC/rostral medial frontal cortex activity relative to comparison individuals [56,61–63]. Moreover, youth and adults with psychopathic traits show impairment in moral judgement [24], including abnormally utilitarian moral judgements [64] and reduced endorsement of care-based norms [65,66]. It has been argued that individuals show impairment in moral judgement because a core component of moral judgement is representing the expected value of the (im)moral action to be considered [67]. As the salience of the victim increases (i.e. the expected value of the action diminished because of representation of the cost of the victim), the action is more likely to be judged as ‘bad’. However, it should be noted that it is the role of the amygdala in moral judgement, rather than that of vmPFC, that appears to be particularly disrupted in individuals with psychopathic traits (e.g. [68,69]).

(iv). Summary

Core ‘empathic’ processes involve: (i) the inhibition of aggression in response to distress cues; (ii) the learning of the negative value of actions such as aggression that are associated with others' distress; and (iii) reasoning about actions that are associated with others’ distress. If these are compromised, as they appear to be in individuals with psychopathic traits, the individual is at significantly increased risk for instrumental aggression. The individual will be significantly more likely to commit acts like aggression for personal gain even if they are associated with significant suffering for others due to reduced responsiveness to this suffering.

3. Anger

Anger is a species-typical response to either: (i) perceived threat [70], (ii) frustration [3,71] or (iii) social provocation [72]. For example, unfair resource allocations elicit retaliation punishments and the extent to which this occurs relates to the level of anger the punisher feels towards the other individual [73].

Like other emotional expressions, displays of anger are communicative displays designed to impart valence information to observers [28,74]. In the case of anger, the informational content is to make one desist from the current behaviour [28]. As communicative displays, demonstrations of anger are governed by display rules; the child is socialized so that they learn when it is appropriate to display the emotion [75]. Some social circumstances permit the display of anger, while others do not.

While anger is a species-typical response, it can also be shown atypically (e.g. too frequently or too intensely). This leads to a definition of irritability as ‘an increased propensity to exhibit anger relative to one's peers’ ([76], p. 277). Particularly high levels of anger may result in the display of reactive aggression [71]. As such, to understand how anger might increase the risk for maladaptive aggression, one needs to understand the neurobiology of anger (cf. [77]) and then how dysfunction of this neurobiology might lead to irritability (cf. [76]).

It can be argued that the behavioural expression of anger is a modulated form of reactive aggression [70,77]. As such, systems involved in mediating reactive aggression should be implicated in the expression of anger. Considerable animal work has specified the neural systems necessary for reactive aggression. These are the amygdala, hypothalamus and PAG [27]. Notably, human fMRI work has shown that these systems respond to all three elicitors of anger: approaching threat [78–80], frustration [81] and as a function of the level of retaliation to social provocations (unfair resource allocations [82,83]).

Of course, if the amygdala, hypothalamus and PAG are critically involved in anger, then individuals showing irritability should show increased responsiveness of these regions to threat, frustration and social provocation (cf. [77]). This suggestion receives support with respect to threat and social provocation. Work with healthy participants has reported that a predisposition to anger is positively associated with the amygdala responding to masked fearful expressions [84]. Moreover, patients with psychiatric disorders at increased risk for reactive aggression and irritability (BPD, IED and DMDD) show hyper-amygdala responsiveness to threat [85–87]. Others at heightened risk for reactive aggression also show increased amygdala responses to threat [88,89]. With respect to social provocation, relative to comparison individuals, patients with BPD show heightened amygdala responses to provocation [90], and reactively aggressive youth show increased amygdala and PAG responsiveness as a function of retaliation level to socially provoking unfair resource allocations [83].

Support for the idea that individuals with irritability should show heightened amygdala, hypothalamus and PAG activity to frustrating events is not well supported, however. Frustrating events involving goal-blocking activate the amygdala/hypothalamus/PAG in healthy individuals [81], and one study reported that a predisposition to anger is positively associated with the amygdala response to the word ‘no’, a word associated with goal frustration [91]. However, a series of studies have not reported increased amygdala (or hypothalamus/PAG) activation to frustration induced by rigged feedback (reward loss for ‘too slow’ responding) in irritable youth [92–94]. One even reported that frustration leads to increased amygdala deactivation in irritable youth relative to comparison youth [93].

Frustration occurs when an individual acts in expectation of a reward but does not receive the reward [71]. A considerable body of fMRI work reports that a failure to receive an expected reward, a prediction error, is associated with a decline in activity within the striatum (for a review, see [95]). This prediction error signal has been considered to be the basis of frustration [96]. From this it might be argued that irritability, the heightened propensity for anger, results from an exaggerated response to not receiving expected reward (i.e. the frustration cue for anger is stronger). Some work supports this suggestion. Three studies have examined the relationship between irritability and sensitivity to reward losses with a methodology (fMRI) suitable to examine striatal responding. Two reported exaggerated declines in activity within the striatum following losses in irritable children/adolescents relative to comparison children/adolescents [93,94] though the third reported the opposite [97]. There was also indication that the greater the reduction in striatal response to loss, the greater the parent-reported irritability in the child [94].

The three core triggers for human anger—approaching threat, frustration (whether indexed as goal-blocking or prediction errors) and social provocation—all initiate activity within dmFC and lateral frontal/aIC [79–83]. Interestingly, there has been some work suggesting that, among healthy adults, anger proneness is positively related to activation of these regions following prediction errors [98]. Similarly, there have been reports using infrared spectroscopy (fNIRS) in healthy younger participants (aged 3–7) that increased distress following frustration relates to increased recruitment of the lateral frontal cortex [99,100]. By contrast, children with clinical levels of irritability show a decreased capacity to recruit the lateral frontal cortex [99]. This has led to the suggestion that the lateral frontal cortex regulates irritability. It is hypothesized that within healthy individuals, the lateral frontal cortex increases activity in response to frustration distress to regulate anger. However, it is suggested that for patients with clinical irritability its operation is compromised such that anger is more likely to be displayed [99,100].

Another region probably involved in the regulation of angry responses/reactive aggression is vmPFC. Anger displays and retaliation responses can be instrumental behavioural choices that are governed by display rules. Moreover, the classic measures of reactive aggression, the Taylor Aggression Paradigm [101] and the Point Subtraction Aggression Paradigm [102] involve clear instrumental components. In these paradigms, participants receive provocations (e.g. aversive thermal stimulation or money loss) from task competitors and then decide on the intensity of the retaliatory response via a choice of response button. While the motivation for the aggression may come from activation of the threat system (amygdala, hypothalamus, PAG) by threat or social provocation, the form of response (nothing, anger displayed or reactive aggression) is selected like any other instrumental response (cf. [103]). VmPFC is critically involved in response selection via its role in representing reinforcment expectancies [40].

Increasing levels of retaliatory behaviour are associated with increasing activity within the PAG and decreasing activity in vmPFC [82,104]. This decreasing vmPFC activity is thought to represent the increasing costs to the aggressor of increasing retaliation [103]. Of course, a healthy individual may still decide on a measured retaliatory response because of the scale of the provocation. However, an individual whose ability to represent the costs/benefits of their actions is compromised is less likely to show a measured retaliatory response. They will be impaired in representing the costs/benefits of their response and thus more likely to engender a disproportionate retaliation. Aggressive individuals and patients with BPD show disproportionate retaliations to provocation [83,90], and this appears to relate to their compromised vmPFC activity [83,90]. Indeed, neurological lesions of vmPFC lead to dysregulated expression of anger and reactive aggression [105,106].

(a). Summary

Irritability (i.e. dysregulated anger) and an increased risk for reactive aggression appear to be associated with: (i) heightened responsiveness of the amygdala and PAG in response to threat and social provocation; (ii) heightened responsiveness of the striatum to negative prediction errors (the unexpected absence of reward) and (iii) dysfunction in potential anger regulatory roles of the lateral frontal cortex and vmPFC.

4. Conclusion

The individual's capacity to show two core social emotions, empathy and anger, significantly modulates his or her risk for aggression. Core ‘empathic’ processes involve: (i) the inhibition of aggression in response to distress cues; (ii) the learning of the negative value of actions such as aggression that are associated with others' distress; and (iii) reasoning about actions that are associated with others’ distress. The greater the degree to which these empathic processes are compromised, as is seen in psychopathy, the greater the likelihood that an individual will engage in instrumental aggression. As such, individuals with psychopathy will be less influenced by the negative valence typically associated with actions that involve harm to others and thus more likely to choose those actions to achieve their goals.

Irritability (i.e. dysregulated anger) is associated with heightened responsiveness of the amygdala and PAG to threat/social provocation and of the striatum to negative prediction errors and/or with dysfunction in potential anger regulatory roles of the lateral frontal cortex and vmPFC. Patients with DMDD, BPD and IED show these forms of atypical responding that increase the risk for anger and reactive aggression. Heightened threat/frustration responding probably means that the individual responds to provocation with aggression rather than withdrawal. Compromised regulatory functioning, particularly of vmPFC, probably means that a more aggressive form of retaliation, one less influenced by the costs to the self and others, is more likely to be selected. It appears that many individuals with psychopathy are compromised in this function of vmPFC and that this may underpin their increased risk for reactive aggression even though their response to threat is not heightened [83].

This cognitive neuroscience approach provides potential assessment tools and treatment targets for aggressive individuals. It stresses the importance for more individualized care for patients with psychopathy, DMDD, BPD or IED. While the functioning of empathic and anger-related functions is probably correlated in the individual, it is plausible that there will be individual differences. Some patients with DMDD may show heightened responsiveness of the amygdala and PAG to threat/social provocation but intact vmPFC functioning, while others may show the opposite pattern. Interventions will need to be targeted to the specific weaknesses of the individual to achieve optimized efficacy. Translating the information above into assessment strategies and using this information to improve current interventions are the next challenges.

Data accessibility

This article has no additional data.

Competing interests

I declare I have no competing interests.

Funding

The study was funded by US Department of Health and Human Services, National Institutes of Health, National Institute of Mental Health K22 MH109558-01.