Language and the Modulation of Impulsive Aggression

C lassification of aggression into premeditated and impulsive subtypes represents an important advancement in the study and treatment of aggression. 1 – 8 Verbal impairments and abnormalities of language processing regions of the brain are prominent features in impulsive aggressors when compared to nonimpulsive aggressors or nonaggressors. 2 , 3 , 9 – 12 Aggression researchers base neurobiological models of impulsive aggression largely on the animal model of defensive rage. 13 , 14 While some of these studies also implicate frontocortical abnormalities that are outside those implicated in the animal models, 11 , 12 , 14 a comprehensive functional anatomy of impulsive aggression involving language processing regions has not been presented. Such a model would have etiological, diagnostic, and therapeutic implications.

Features of Impulsive Aggression

Impulsive aggressive acts are emotionally charged, uncontrollable, and grossly disproportionate “hair-trigger” reactions to stimuli, leading to an agitated state during which verbally or physically aggressive acts are committed. 15 In the agitated state, interpersonal communication is maladaptive, 15 and information processing appears impaired. Perception of contexts and interpersonal interactions are biased toward interpretations of threat, victimization, or disrespect. Even feedback such as an apology may seem to paradoxically fuel escalation. 2 , 3 After an incident, the individual commonly feels remorse, recognizes the loss of control during the agitated state, and recognizes that the aggressive reaction was disproportionate to the evocative trigger. 2 , 3 , 15

Impulsive aggression has been subcategorized into verbal aggression, physical aggression against objects, and physical aggression against people. 16 – 21 These subcategorizations appears warranted as it is not clear that each are the same construct, occur via the same disturbances of self-regulatory functioning, or respond to the same treatment modalities. These three categories derive from the work of Yudofsky and colleagues, 22 who originally devised these categories in the Overt Aggression Scale, an instrument developed for the monitoring and response of aggression to interventions (in psychiatric inpatients) by measuring the frequency and intensity of aggressive acts. Categorization between verbal and physical impulsive aggression appears supported by factor analysis of the construct of impulsivity, consisting of three components: planning (does not plan ahead), cognitive impulsivity (makes up one’s mind quickly), and motor impulsivity (acting without thinking). 23 Although the three factors are interrelated, 23 they may be expressed by different neurobiological pathways.

Pharmacological, neuropsychological, and neurobiological features distinguish impulsive individuals from nonimpulsive (premeditated) aggressors and/or nonaggressive individuals. 2 , 3 , 5 , 24 For example, Houston et al. 8 and Barratt et al. 2 , 3 demonstrated that a selection of predominantly impulsive aggressors results in a group of subjects with robust responsiveness to pharmacological treatment with the anticonvulsant phenytoin, as compared to the response of mixed aggressors (features of both impulsive and premeditated aggression) or predominantly premeditated aggressors. Barratt et al. 2 , 3 demonstrated, using extensive neuropsychological testing, that only tests involving verbal abilities—the Gray Oral Reading Test, verbal memory, verbal IQ, vocabulary, and similarities—differentiated these impulsive aggressors from the nonimpulsive aggressors, with all of the differences favoring nonimpulsive aggressors. In the same study, event-related potential testing demonstrated significantly decreased P300 amplitudes in language processing regions of the brain in impulsive aggressors compared to nonimpulsive aggressors, in response to low frequency, random (oddball) stimuli requiring response inhibition during a go/no-go task. 2 , 3 These event-related potential abnormalities in the impulsive aggressors normalized (augmented) after treatment with phenytoin and were associated with a dramatic reduction in aggressive acts. In the nonimpulsive (premeditated) aggressor group, phenytoin neither significantly improved aggression nor augmented P300 amplitudes in language processing regions as measured by event-related potential testing.

A caveat to the interpretation of these study findings is that event-related potential testing has relatively poor localizing value, leaving open the question of which brain regions were functionally normalized by phenytoin treatment. However, volumetric and functional MRI (fMRI) studies of impulsive aggressors have subsequently replicated findings of abnormalities in the language processing regions of the brain as elaborated in the sections “Medial Orbital Prefrontal Cortex” and “Lateral Prefrontal Cortex.” 9 , 10 , 12

Significance of Impulsive Aggression

Impulsive aggression is a problem of substantial magnitude. Impulsive aggressive episodes endanger relationships, occupational functioning, and may lead to legal difficulties. A sizable percentage of veterans seeking help for military-related posttraumatic stress disorder report difficulty with aggression. 25 Siever 26 pointed out that impulsive aggression is a hallmark of personality-disordered individuals, particularly those with borderline personality disorder, but also those with narcissistic personality disorder or antisocial personality disorder. According to DSM-IV-TR criteria, intermittent explosive disorder may be conceptualized as an extreme subset of impulsive aggressors who have severe problems modulating physically aggressive impulses (Matt Stanford, personal communication, 2007), although this has not been systematically investigated.

Current Impulsive Aggression Model

The concept of defensive rage, derived from animal research, is often used to model impulsive aggression. 24 , 27 – 29 This section explains this model and discusses features of impulsive aggression ( Figure 1 ).

Defensive rage is an automatic, reflexive survival response to overwhelming threat. Its explosiveness resembles human impulsive aggression. The pathway mediating defensive rage involves the medial nucleus of the amygdala, the medial hypothalamus, and the dorsal periaqueductal gray. 13 , 24 , 29 Independent stimulation of these regions triggers experimental defensive rage in animals, provided that the pathway is intact. 13 , 24

According to this model, brain regions which inhibit these mediators of defensive rage include the anterior cingulate cortex (ACC) 24 and the orbital prefrontal cortex (OFC), 13 highly confluent frontal regions of the prefrontal cortex. Data supporting this model include that stimulation of the ACC increases latency to aggression in cats. 30

Amygdala

While stimulation of the amygdala triggers defensive rage, the amygdala also modulates defensive rage. 13 , 24 Amygdala lesions may increase or decrease aggression in humans. 13 Impulsive aggressors commonly report in retrospect that when angry, they “lose control,” neither able to modulate anger nor aggressive impulses. This fits well with Stanford et al.’s 4 proposal that impulsive aggressors may have deficits in physiological arousal and that sudden surges in arousal induce emotional states that are difficult to control. Imaging studies of impulsive aggressors demonstrate abnormalities of the amygdala. In a volumetric structural MRI study, van Elst et al. 11 demonstrated severe atrophy of the amygdala, or lesions involving the amygdala, in a subset of individuals with temporal lobe epilepsy and impulsive aggression, compared to individuals with temporal lobe epilepsy without impulsive aggression. In subjects with borderline personality disorder and impulsive aggression, 10 fluoxetine was associated with a reduction in impulsive aggressive acts and increased metabolism in the medial temporal lobes, including the amygdala (Antonia New, M.D., personal communication, 2006). A separate study of borderline personality disorder and impulsive aggression subjects relative to healthy comparison subjects by New et al., 31 utilizing positron emission tomography (PET) imaging coregistered with MRI, found differences in amygdala circuitry in borderline personality disorder and impulsive aggression subjects manifested by the loss of ventral-dorsal anatomic specificity in association with functional “disconnection” of the ventral amygdala with the OFC (Brodmann’s areas 11, 12) and Brodmann’s area 47 and other prefrontal regions. In contrast, comparison subjects did demonstrate ventral-dorsal anatomic specificity of the amygdala in association with functional correlations, suggesting “tight coupling” of the right ventral amygdala with the OFC (Brodmann’s areas 11 and 12), Brodmann’s area 47, and Brodmann’s area 44. Aggression scores in borderline personality disorder and impulsive aggression subjects, as measured by the Buss-Durkee Hostility Inventory (BDHI), negatively correlated with metabolic rate in ventral and dorsal regions of the right amygdala.

Zagrodzka et al. 32 found that in cats, unilateral lesions of the central nucleus of the amygdala increased expression of defensive rage. Seigel et al. 33 found that enkephalinergic projections from the central nucleus of the amygdala to the dorsal periaqueductal gray produce potent inhibition of defensive rage via mu receptors in cats, providing a mechanism by which lesions of the amygdala increase aggression. 29 Lesions of excitatory glutamatergic prefrontal cortical neurons, which inhibit the amygdala via synapses with inhibitory interneurons, 27 , 34 may also increase aggression.

Accurate identification of emotion in facial expressions and social cues relies on intact amygdala and orbitofrontal cortical functioning. 35 – 37 In a study of subjects with intermittent explosive disorder compared to subjects without, Best et al. 38 found that subjects with the disorder inaccurately identified emotion in pictures of facial expressions, which is consistent with reports of misperception of intent during impulsive aggressive episodes. Intermittent explosive disorder subjects were significantly more prone to label neutral expressions as fear and disgust and demonstrated impaired recognition of facial expressions of surprise and anger. These authors concluded that the bias to interpret neutral expressions as negative, and the inability to recognize cues of impending threat which would inhibit aggression in healthy subjects, might contribute to provocation of aggression in individuals with intermittent explosive disorder.

Affective startle appears to involve frontal-amygdala circuitry. 39 Recently, Hazlett et al. 39 reported that borderline personality disorder patients characterized by impulsivity and aggression show greater-than-normal affective startle eyeblink amplitude during processing of unpleasant words, indicating heightened emotional processing. Yet on self-report, borderline personality disorder patients rated the unpleasant words as less unpleasant than healthy comparison subjects, suggesting that individuals with impulsivity and aggression have deficits in labeling their own emotions. While it is not clear from this study that the borderline personality disorder patients would meet criteria for impulsive aggression, the findings are consistent with abnormalities in emotional control in a population with a high incidence of impulsive aggression. In summary, dysfunction of key regions of the amygdala (or its prefrontal connections with the orbital prefrontal cortex) that modulate arousal and defensive rage, or that contribute to networks involved in the discrimination of emotional cues, may explain the importance of amygdala abnormalities in neuroimaging studies of impulsive aggressors. The complexity of the amygdala, however, precludes a straightforward relationship to emotional control and, by extension, aggression. For example, Kim and Hamonn 40 demonstrated amygdala activation to both negative stimuli (e.g., traffic accidents, vermin, domestic violence, bodily injury) and positive stimuli (celebrations, sporting events, romantic couples) and during regulation of positive emotions, but not during regulation of negative emotions.

Medial Orbital Prefrontal Cortex

It has been proposed that abnormalities of the anterior cingulate cortex (ACC) and orbital prefrontal cortex (OFC) regions of the medial prefrontal cortex, alone or in combination with abnormalities of the amygdala, underlie the hyperarousal/dyscontrol states seen in impulsive aggressors. 9 , 10 , 13 , 14 , 38 Blair 13 and Davidson et al. 14 proposed that the ACC and OFC are normally activated during anger arousal via serotonergic mechanisms and exert inhibitory influence over aggressive emotional responding via mechanisms including inhibition of the amygdala, hypothalamus, and brainstem periaqueductal gray. Healthy comparison subjects have been shown to activate the OFC when angry. 13 Studies of impulsive aggressors have found hypoactivation of the ACC and OFC regions of the medial prefrontal cortex. 9 , 10 , 38 The OFC carries out low-level appraisals of punishment/reward values of behavioral responses (in contrast to complex punishment/reward appraisals performed by lateral prefrontal cortical regions) and of the appropriateness of behaviors in accordance with social cues.

Within the current model of defensive rage, conditions of overwhelming threat may “release” defensive rage from inhibitory modulation. It remains unclear as to how this “release” occurs in humans. One hypothesis consistent with the current model ( Figure 1 ) is that the ACC and OFC regions in impulsive aggressors may lack sufficient levels of activation, with resultant insufficient inhibition of rage responses. New et al.’s 9 findings of blunted OFC (Brodmann’s area 10) and ACC (Brodmann’s area 32) activation in response to a serotonergic stimulus utilizing PET imaging support this hypothesis, as do findings 10 that fluoxetine increased metabolism in the medial OFC, in association with decreased aggression. An alternative hypothesis within this model is that defensive rage may deactivate prefrontal cortical regions that modulate its expression, thus releasing defensive rage responses from prefrontal inhibition. Garcia et al.’s 41 findings that increasing degrees of conditioned fear led to the deactivation of medial prefrontal cortex in mice support this hypothesis.

Thalamus

Abnormalities of the thalamus, a region critical to integrated prefrontal cortical functioning, might also play a role in impulsive aggression. This hypothesis is supported by the findings of Stanford and colleagues 4 demonstrating impaired sensory gating in impulsive aggressors that normalized with phenytoin, in association with a reduction in impulsive aggressive acts. The thalamus is critical to sensory gating. 42 Abnormalities of the serotonin system have been shown to disrupt sensory gating in rats. 43 Raine et al.’s 44 PET study finding of lower left than right thalamic metabolism in impulsive aggressors provides supportive evidence of thalamic abnormalities in this condition. Abnormalities of the serotonin system affecting the thalamus and other brain regions might be postulated to disrupt sensory gating and integrated prefrontal functioning of regions important to the modulation of impulsive aggression. Frankle et al. 45 found decreased concentration of serotonin transporter proteins in the ACC in a group of 10 impulsive aggressors relative to 10 age- and sex-matched comparison subjects. The serotonin transporter has been proposed as a marker for the integrity of the serotonin system. 46

Lateral Prefrontal Cortex

During the 1990s, growing evidence of neurobiological differences between impulsive and premeditated aggression inspired functional neuroimaging studies of aggressive individuals using a variety of descriptions that share similarities with more recent characterization of impulsive aggression. Many of these studies employed [ ¹⁸ F]fluorodeoxyglucose (FDG) PET. For example, Volkow et al. 47 compared eight psychiatric patients with “purposeless, repetitive violent behavior” with eight healthy comparison subjects and found significantly lower metabolic rates in bilateral medial temporal and prefrontal cortices. Raine et al. 44 compared nine “affective” murderers, 15 “predatory” murderers, and 41 age- and sex-matched healthy comparison subjects while they performed a Continuous Performance Test. They found that bilateral medial and lateral prefrontal cortical hypometabolism significantly distinguished “affective” murderers from both “predatory” murderers and healthy comparison subjects. “Affective” murders, compared to both “predatory” murderers and healthy comparison subjects also had higher right subcortical (thalamus, hippocampus, midbrain, amygdala) metabolism and lower right hemisphere prefrontal cortical/subcortical ratios. Raine et al., 44 studying 41 murderers categorized as “not guilty by reason of insanity,” analyzed 41 healthy comparison subjects and found that “affective” murderers, relative to comparison subjects, had lower right and left medial and lateral prefrontal hypometabolism, and abnormal subcortical asymmetries (left hemisphere lower than right) of metabolic activity in the thalamus, amygdala, and medial temporal lobe. Goyer et al. 48 studied 17 personality-disordered individuals compared to 43 healthy comparison subjects and found an inverse correlation between a life history of aggressive impulse difficulties and regional metabolism in the lateral orbital prefrontal cortex, including Brodmann’s area 47.

In considering these findings in the context of their functional significance, it should be noted that variation exists in the classification of Broca’s area. Some authors include Brodmann’s areas 44, 45, and 47, 49 , 50 including the pars orbitalis of Brodmann’s area 47, 51 as Broca’s area, while others include Brodmann’s areas 44 and 45 only, but classify Brodmann’s area 47 as a language processing region important to semantic retrieval, decision-making, 52 , 53 and syntactic and phonological processing. 50 , 53 – 55 Brodmann’s area 47 extends ventromedially from the lateral aspect of the inferior frontal gyrus into the orbitofrontal cortex, including the orbital aspect of the inferior frontal gyrus, and is therefore also referred to as orbitofrontal cortex. 56 Edmund Rolls suggests that Brodmann’s area 47 is a language processing region with connections to contiguous regions of the orbitofrontal cortex (personal communication, 2006). For the purposes of this article, we classify Brodmann’s area 47 as part of Broca’s area. However, for clarity, where studies reviewed in this article find significant changes in Brodmann’s area 47 unaccompanied by changes in Brodmann’s areas 44 or 45, the region will be referred to simply as Brodmann’s area 47.

Davidson et al. 14 proposed that abnormalities of the amygdala, ACC, OFC and “other” regions of the prefrontal cortex with which the OFC is connected constitute the critical nodes of networks that modulate impulsive aggressive outbursts. By extension, we propose that these “other” regions of the prefrontal cortex include the premotor cortex and Broca’s area. Recent neuroimaging studies utilizing more rigorous criteria and nomenclature for impulsive aggression and more detailed structural and functional methods demonstrate hypofunctioning (by functional neuroimaging) or volumetric decreases (by structural MRI) in the inferior frontal gyrus (Brodmann’s areas 44, 45, 47), 9 , 10 which overlaps extensively with Broca’s area (Brodmann’s areas 44, 45, 47), and the left premotor cortex. 12

New et al. 9 found that blunted activation in bilateral Broca’s area (left Brodmann’s area 45 and right Brodmann’s area 47) accompanied the lack of ACC and OFC activation in response to metachlorophenylpiperazine (m-CPP), a serotonin agonist, with blunted prefrontal response much greater on the left. Hazlett et al. 57 found that borderline personality disorder subjects with greater white matter volume in Brodmann’s areas 44 and 47 had greater irritability-assaultiveness subscale scores on the Buss-Durkee Hostility Inventory (BDHI). They hypothesized that greater white matter volume may be a marker of inefficient white matter processing and/or connections. New et al.’s 10 study of subjects with borderline personality disorder and impulsive aggression demonstrated that fluoxetine treatment of impulsive aggression was associated with changes in Brodmann’s area 47 that correlated with changes in temporal lobe neocortex, including Brodmann’s area 22 (middle temporal gyrus) on the right, and with Brodmann’s areas 20, 21 (superior temporal gyrus), and 22 (middle temporal gyrus) on the left, along with the aforementioned improved activation of the OFC and medial temporal regions including the amygdala. Brodmann’s areas 20, 21, and 22 are language processing regions of temporal lobe neocortex. While not specifically investigating impulsive aggression, Hazlett et al. 57 showed that borderline personality disorder patients, who frequently manifest both impulsivity and aggression, demonstrated marked reductions in anterior cingulate gyrus volume. Woermann et al. 12 found significant decreases in gray matter involving left inferior frontal cortex and left premotor cortex by volumetric MRI in impulsive aggressors. They also found that subjects with temporal lobe epilepsy and impulsive aggression had increases in gray matter volume in the left temporal lobe neocortex, which includes Brodmann’s areas 21 and 22, in contrast to temporal lobe epilepsy subjects without impulsive aggression who demonstrated bilateral gray matter increases in the temporal lobe neocortex without lateralized preponderance. These investigators proposed that the gray matter increases represented gliosis, although this remains to be directly demonstrated.

Dimensions of Language Processing

The following sections discuss which dimensions of language processing may be relevant to the modulation of impulsive aggression ( Figure 2 ).

Syntactic and Semantic Dimensions of Language Processing

Villemarette-Pittman et al. 5 provide evidence directly linking the relationship between syntactic and semantic language processing and executive functioning to impulsive aggressors (N=40) by utilizing a battery of verbal measures requiring increasing degrees of spontaneous organization (Peabody Picture Vocabulary Test, WAIS-III Picture Arrangement, self-produced oral narratives to a picture [e.g., discourse analysis]). They found that impulsive aggressors differed significantly from nonaggressive comparison subjects on discourse analysis variables of “syntactical well-formedness” (e.g., proper grammar) and “semantic accuracy” (e.g., accuracy of information garnered from picture), whereas the groups performed similarly on measures of vocabulary and word-picture matching (e.g., Peabody Picture Vocabulary Test). These impairments occurred despite the absence of significant differences in education or general intellectual ability between the impulsive aggressors and nonaggressive comparison subjects. The authors provide a conceptually appealing conclusion that impulsive aggressors have language processing impairments that involve a high degree of executive functioning. 5 While limitations of this study include the use of discourse analysis itself, which lacks normative values and is not validated as a measure of executive functioning, their notion of executive dysfunction is consistent with others. 38 The precise relationship between executive functioning and language and the relationship between these cognitive functions and behavior are not yet known.

The left inferior frontal cortex (regions of Brodmann’s areas 44, 45, and 47) controls semantic accuracy, semantic selection and retrieval, and mediates syntactic processing. 53 It has been hypothesized that the left Brodmann’s area 47 subserves syntactic/semantic processing within a construct that is broader than, but which encompasses, language. Levitin and Menon 53 found activation of the ACC, the pars orbitalis region of left Brodmann’s area 47 (and its right homologue to a lesser extent), while healthy control subjects compared music (activation) to counterpart versions scrambled in order (inactivation), with timbre, pitch, and volume held constant. Levitin 58 proposed that the pars orbitalis of left Brodmann’s area 47 processes the syntax and semantic meaning of sequences of stimuli, independent of modality (e.g., linguistic stimuli, signed language, music), which unfold over time.

Semantic meaning relies on temporal ordering of sequences of stimuli inherent in syntactic processing. All languages (spoken and signed) and music rely on a linear order of constituent components to convey meaning as opposed to random order. 53 Levitin and Menon 53 proposed that Brodmann’s area 47 and the adjoining anterior insula organize temporal sequences of stimuli in the perceptual stream according to preexisting, learned expectancies of “how things go together.” These investigators proposed that these regions also process stimulus sequences according to whether the sequences are consistent with, or violate (as in oddball tasks), such expectancies. These functions potentially relate with the putative function of mirror neurons that are believed to originate in Broca’s area. 59 Researchers propose that mirror neurons contribute to the ability to infer the motives underlying the sequences of actions of others, as when someone reaches for a glass of water. 59 Levitin and Menon 53 pointed out that damage to the lateral left inferior frontal cortex has been shown to impair the ability to generate conscious representations of speech and behavioral sequences. 60 These findings, taken together with the findings of Villemarette-Pittman et al. 5 in impulsive aggressors, support a relationship between syntactic and semantic processing disturbances and impulsive aggression. Error detection and inhibition of errant behavioral sequences potentially may not occur if the behavioral sequences do not first exist as conscious mental representations. Perceptual processing disturbances involving speech and behavioral sequences which violate or conform to social norms and learned expectancies may lead to misinterpretations and hence inappropriate (e.g., aggressive) responses. Syntactic and semantic processing disturbances may reinforce aggressive responding in the face of an inability to negotiate conflict and anger with the use of words.

Verbal Working Memory

The left premotor cortex and left inferior frontal cortex (Brodmann’s areas 44, 45, 47) are important to verbal working memory. 61 Researchers propose that verbal working memory is critical to the executive, hierarchic control of behavior. 62 , 63 This proposal has implications for potential mechanisms underlying modulation of impulsive aggressive behavior. Such mechanisms may include cognitive constraint (verbal self-instruction), verbal (including silent) rehearsal, reflective functioning, and deductive reasoning, all of which rely on these language regions 53 , 63 – 65 and upon the left inferior frontal cortex’s role of maintaining verbal material in verbal working memory. 53 , 63 Deductive reasoning is also highly contingent upon intact syntactic processing. Intriguing findings by Matsubara et al. 66 implicate left Brodmann’s area 47 as critical to the intentional, hierarchic capacity to override habitual, reinforced behaviors. In this functional MRI (fMRI) study, activation of primarily left Brodmann’s area 47 occurred during a task requiring inhibition of habitual, manual responses associated with positive reinforcement. Subjects were instructed to intentionally lose a rock-paper-scissors game rather than allow themselves to win (the habitual, reinforced behavior). Stronger activation of left Brodmann’s area 47 was associated with lower error rates (of inhibition). These authors cited other researchers’ findings 67 of the activation of left Brodmann’s area 47 in association with similar response inhibition (during the go/no-go task). These findings support the proposal that left Brodmann’s area 47 is critical to the executive capacity to intentionally override habitual or automatic behaviors, providing a potential link between verbal working memory and overriding reflexive reactions such as aggressive responding.

Konishi et al. 68 , 69 found bilateral but predominantly right hemispheric inferior prefrontal cortical activation (including right Brodmann’s areas 44, 45) in association with motoric inhibition (no-go) on a go/no-go task, suggesting a role for right homologues of the left inferior frontal cortex in hierarchic control of behavior and response inhibition. A PET study of healthy subjects by Pietrini et al. 70 supports a connection between right homologues of left language regions and the cognitive constraint of aggression associated with defensive rage. This PET imaging study, performed while healthy male subjects imagined not reacting aggressively while their mother was assaulted (cognitive restraint or inhibition of aggression), demonstrated increased regional cerebral blood flow (rCBF) in right Brodmann’s area 47, bilateral OFC (Brodmann’s area 10 and 11), right supplementary motor cortex (Brodmann’s area 8), and right ACC (Brodmann’s area 32). When subjects imagined responding aggressively to the assailants in an unrestrained manner, even killing them, the PET imaging demonstrated deactivation in right Brodmann’s area 47, associated with decreased rCBF in Brodmann’s areas 10 and 11 of the OFC.

The above research findings support a link between syntactic processing, verbal working memory, and behavioral control (motoric inhibition). Barratt et al. 2 , 3 found low P300 amplitudes in a go/no-go paradigm only in response to “no-go” stimuli (low frequency, random, “oddball” stimuli requiring motoric inhibition) in impulsive aggressors. Barratt et al. 2 proposed that the low P300 amplitudes in the language processing regions represented diminished cognitive resources in those regions. This proposal supports that diminished verbal working memory, and/or diminution in cognitive processes employing syntactic processing (detection of violation of expectancies), may contribute to compromised executive control of behavior, including reduced capacity for inhibition of impulsive aggression.

While the rostral-ventral affective division of the ACC (Brodmann’s areas 24, 25, 32) modulates mediators of emotional experience and behavior, 71 the ACC is also important to language processing and speech. 71 , 72 The ACC exerts executive control over the constituents of working memory, 61 appears to have connections with Broca’s area, 72 and is therefore important to verbal working memory, cognitive control, reflective functioning, 65 and goal-directed behavior. 63 , 71 Osaka et al.’s 73 fMRI findings suggest that the ACC relies on interaction with the lateral prefrontal cortical regions, including regions of the inferior frontal gyrus (Brodmann’s areas 44, 45 in this study) in its role of allocating cognitive resources and coordinating attentional control for the performance of cognitive tasks. Devinsky et al. 71 proposed that the ACC is uniquely positioned to integrate modulation of affect and behavior with intellect, with intellect here specifically referring to ideas held and manipulated in verbal working memory. This supports the idea put forth by Hariri et al. 34 that the ACC relies on interaction with more lateral cortical regions so that modulation of affect and behavior may be guided by sophisticated appraisals and reasoning for flexible and adaptive functioning.

Nonverbal Dimensions of Language Processing (Prosody)

Given the inability to recognize emotional intent of others during an impulsive aggressive episode, impairments in prosodic dimensions of language processing leading to impairments in recognition of emotional subtexts may contribute to impulsive aggression. Activation patterns of processing emotional voice prosody have been reported. In an fMRI study of healthy subjects, Mitchell et al. 74 found that the overall processing of emotional voice prosody was relatively lateralized to the right lateral temporal lobe, involving Brodmann’s areas 21 (superior temporal gyrus) and 22 (middle temporal gyrus), but also involved the left lateral temporal lobe (Brodmann’s areas 21, 22). The neural response to neutral voice prosody, compared to emotional prosody, was left -lateralized to the inferior frontal gyrus and premotor cortex (Brodmann’s area 6). In an fMRI study of healthy subjects, Sander et al. 75 found that angry voice prosody activated the right amygdala and bilateral Brodmann’s area 22 (superior temporal sulci) regardless of whether subjects had been directed to attend to or ignore the stimulus. Listening to angry voice prosody that was “to be attended to” additionally recruited the orbitofrontal cortex. These regions correspond to regional abnormalities in impulsive aggressors, supporting the notion that prosodic disturbances leading to misperceptions of emotional subtexts may contribute to impulsive aggression.

Dimensions of Language Processing Involving Abstraction and Increasing Degrees of Perceptual Difficulty

PET studies have demonstrated that increasing degrees of (linguistic) perceptual difficulty and abstract dimensions of language processing involve right hemisphere homologues of left hemispheric language processing regions, including right Brodmann’s area 47. 76 , 77 In a study of 19 healthy subjects during semantic and syllabic decision-making based on degraded speech stimuli (degraded to harsh whisper) compared to clearly articulated speech stimuli, 76 there was greater activation of left ventrolateral (Brodmann’s area 47) and dorsolateral prefrontal cortex during semantic decision-making based on clearly articulated speech stimuli. In contrast, greater activation in the right ventrolateral (Brodmann’s areas 9, 47) and dorsolateral prefrontal cortex occurred during semantic and syllabic decision-making based on degraded speech stimuli. Sharp et al. 76 concluded that these findings supported the hypothesis that as perceptual difficulty of tasks (including language tasks) increases, right hemispheric regions are recruited. In a different study, Bottini et al. 77 found that during a task involving sentence processing compared with lexical decision-making, right hemispheric activation, including right Brodmann’s area 47, accompanied the expected left hemispheric activation, which included left Broca’s area (Brodmann’s areas 45, 47). These investigators also found that metaphorical language processing, in contrast to literal language processing, activated many right hemispheric regions. Another study of sentence processing that required syntactic and phonologic decision making regarding the presence or absence of errors found that regions corresponding to left Brodmann’s area 47, as well as right Brodmann’s areas 44 and 45, were activated. 55

The previously described bilateral processing of prosody and the recruitment of right homologues of left language regions for abstraction and increased perceptual difficulty underscore the significance of New et al.’s 9 , 10 findings of bilateral abnormalities in language regions, as well as findings of other investigators of bilateral prefrontal lateral cortical abnormalities in impulsive aggressors. These findings support the contention that diminished capacity to process and interpret prosodic and complex or abstract linguistic stimuli may contribute to impulsive aggression. These findings also relate to the factor analysis of the Barratt Impulsivity Scale, which supports decreased capacity for cognitive complexity as a dimension of impulsivity, 23 and suggest the hypothesis that impaired recruitment of right hemisphere homologues of left language regions may contribute to impulsive aggression.

Language and Modulation of Emotions

Prefrontal verbal cognitive processing plays an important role in the modulation of emotional processes in humans. Functional MRI findings by Hariri et al. 34 , 78 and Ochsner et al. 79 support the modulatory role of cognitive verbal appraisal upon fear responses associated with amygdala activation. Kim and Hammon 40 demonstrated, using fMRI, that Brodmann’s areas 44, 45, and/or 47 (Broca’s area) and Brodmann’s area 6 (premotor cortex) were recruited, in conjunction with regions of the ACC and OFC, for tasks involving willful regulation (increasing and decreasing) of both negative and positive emotions.

These findings support that the interconnectedness of the anterior cingulate with both limbic mediators of emotion (e.g., amygdala) and lateral prefrontal language regions makes it ideally situated to exert top-down regulatory influence on the amygdala in response to verbal cognitive evaluation. This is consistent with the findings of Hariri et al. 34 that ventral/medial prefrontal cortical regions of the anterior cingulate gyrus and the orbitofrontal cortex, important to modulation of emotional states, rely on interaction with the ventrolateral (e.g., Broca’s area/inferior frontal gyrus) and dorsolateral prefrontal (e.g., the premotor cortex) regions for more complex and adaptive reward-value computations and behavioral responses necessary for self-control and the modulation of emotional experience. Hariri et al. 34 point out that linguistic cognitive appraisal of emotional experience appears to engage executive prefrontal lateral cortical networks by which humans can uniquely regulate emotional experience, speech, and behavior. This idea fits well with LeDoux’s conceptual model 28 that depicts the lateral prefrontal cortex as exerting executive modulatory control over medial prefrontal cortical regions known to inhibit limbic mediators of negative emotions such as fear and rage.

Additional data supporting the modification of the current defensive rage model of impulsive aggression are the findings by Barratt et al. 2 , 3 that the reduction of impulsive aggression may not rely on reduction of “rage” or anger. Barratt and colleagues found that phenytoin treatment of predominantly impulsive aggressive inmates did not result in statistically significant reduction in anger, despite a 71% reduction in the frequency of physically assaultive acts. These findings leave open the possibility that motoric inhibition of impulsive aggressive acts and modulation of anger may be dissociable and that anger may be a necessary but insufficient cause of impulsive aggressive acts.

Executive Functioning Relies on Integrated Prefrontal Functioning Including Language Processing Regions

Aggregate findings reviewed in this article support an overarching hypothesis that the ACC and OFC regions of the medial prefrontal cortex rely on connections with lateral prefrontal language regions (Broca’s area and premotor cortex) for the modulation of impulsive aggression. According to this model, lower-order appraisal capacity of the ACC and OFC regions rely on the interaction between more lateral cortical prefrontal regions for more complex, flexible, and adaptive reward-punishment computations and behavioral responses. In support of this hypothesis, Villemarette-Pittman et al.’s 5 findings imply an interdependence between the language processing regions and other prefrontal regions in concluding that impulsive aggressors demonstrate impairments in language processing involving increasing degrees of executive functioning.

Summary, Research Directions, and Treatment Implications

While it is not yet clear if impulsive aggression is a distinct entity given that it is seen in many conditions, we have reviewed evidence that unique pharmacological and neurobiological features distinguish impulsive aggression from premeditated aggression. These findings include robust treatment responsiveness to the anticonvulsant phenytoin and the abnormalities of language processing regions. We have also reviewed convergent lines of evidence that the current model of impulsive aggression and its regulation, largely based on animal models of defensive rage, warrants updating to incorporate impairments in language processing by which humans are uniquely capable of modulating emotional states and aggressive impulses. According to this updated model, perceptual miscategorizations of stimuli, and their associated affects (e.g., fear and defensive rage), are activated, and prevail in the face of nonintegrative prefrontal functioning, resulting in faulty modulation of defensive rage pathways. Abnormalities of amygdalar functioning may contribute to these impairments due to the dysfunction of regions involved in both defensive rage and discrimination of the emotions of others. In contrast to the defensive rage model, this updated model suggests how regions involved in the overall orchestration of language, thinking, and behavior—as well as the modulation of emotional experience—are important in impulsive aggression. Hypoactivation of ACC and OFC regions of the medial prefrontal cortex may not only reflect defective capacity to modulate limbic mediators of defensive rage but also reflect decreased connectivity with more lateral cortical language processing regions. Interactions with other regions, such as the basal ganglia, while not considered here, are likely important.

While the nature of these language processing abnormalities and the connection between language processing and modulation of impulsive aggression remain to be fully elucidated, several hypotheses emerge from this model. Language processing regions may mediate executive abilities important to the regulation of aggressive impulses, such as deductive reasoning, cognitive restraint of aggression (that may be separable from motoric inhibition of aggression), cognitive modulation of emotion, and/or reflective functioning. From a mechanistic perspective, activation of language processing regions of the brain may be critical to dampen limbic mediators of impulsive aggression via the ACC and OFC under normal conditions. Abnormalities leading to deficient, nonintegrative prefrontal cortical processing may lead to misperceptions that trigger defensive rage as well as deficient modulation of its expression. Within this model, it is possible that verbal impairments may be a manifestation of abnormal dimensions of language processing (e.g., syntactic processing) that provide a defective interface between structures involved in the modulation of impulsive aggression. Alternatively, activation of defensive rage (in response to a perceived threat) may inactivate prefrontal cortical language processing regions important to its modulation. Finally, it remains possible that verbal impairments may be an epiphenomenon in which these impairments are the outward, measurable manifestations of a shared function of overlapping brain regions. Thus, impulsive aggression may be caused by multiple etiologies leading to a common final pathway.

Treatment implications of the findings reviewed in this article include numerous promising targets for phenytoin or possibly other use-dependent anticonvulsants 2 and therapeutic approaches that take into account language disturbances. The findings also suggest that others may be less promising, such as efforts focused solely on the reduction of “anger.” A better understanding of the dimensions of language processing that are involved in impulsive aggression and the nature of its interaction with executive disturbances may lead to the design of effective language-based therapies, such as cognitive behavior therapy, for impulsive aggression. Moreover, this model suggests that pharmacologic agents, such as phenytoin and fluoxetine, that have been demonstrated to improve both impulsive aggression and disturbances in the language processing regions of the brain 2 , 3 , 10 may act synergistically with language-based therapies.