The availability of functional imaging methodologies occurred relatively recently. Although the earliest techniques (e.g., positron emission tomography [PET]) were not suitable for child and adolescent populations because they rely on radioactive isotopes, more recent methodologies, such as functional magnetic resonance imaging (fMRI), can be used with younger participants without major risk. Of course, the field remains in its infancy, and when writing this review, we have had to rely on a variety of studies with adult populations and infer results that we anticipate with child and adolescent samples. However, it is a time of considerable and rapid advance, and thus, the gaps in our knowledge base will likely soon be filled.

The overarching goal of this review is to help understand the neurocognitive impairment(s) that increase the risk for antisocial behavior and thus may incur the Diagnostic and Statistical Manual for Mental Disorders (DSM; American Psychiatric Association, 1994) psychiatric diagnoses of oppositional defiant disorder (ODD) and conduct disorder (CD). The persistent feature of ODD is a pattern of persistently negativistic, hostile, defiant, provocative, and disruptive behavior, which is clearly outside the normal range of behavior for a child of the same age in the same sociocultural context. There is frequently low frustration tolerance. The persistent feature of CD is behavior in which the basic rights of others or major age-appropriate societal norms or rules are violated, as manifested by the presence of, for example, aggression, property destruction or theft. These diagnoses are partly age dependent. ODD is usually diagnosed before the age of 9–10 years. CD is usually diagnosed after the age of 9 years. Neither can be diagnosed after the age of 18 years when a diagnosis of antisocial personality disorder (APD) may be warranted.

Although not necessarily the consensus view, many authors consider individuals meeting the criteria of CD and ODD to be heterogeneous with respect to the pathology causing their disorder (Blair, Mitchell, & Blair, Reference Blair, Mitchell and Blair2005; Frick & White, Reference Frick and White2008; Steiner & Remsing, Reference Steiner and Remsing2007). This heterogeneity in causal pathology is arguably reflected in the extent of comorbidity of CD and ODD with other disorders (in some samples up to 80%; Steiner & Remsing, Reference Steiner and Remsing2007). Many diagnoses are associated with CD including posttraumatic stress disorder (PTSD), bipolar disorder, attention-deficit/hyperactivity disorder (ADHD), and psychopathic tendencies. Given that many of these disorders are associated with differing pathophysiologies, some of which appear to be mutually exclusive at the biological level (e.g., PTSD is associated with increased amygdala responsiveness, whereas psychopathic tendencies are associated with reduced amygdala responsiveness), it is plausible that there are different developmental routes to CD/ODD reflecting differing neurobiological bases. Thus, in this paper, we will examine the neuroimaging literature with regard to (a) the potential neurobiological bases of the developmental routes to CD/ODD, (b) why the aforementioned disorders are associated with an increased risk for CD, and (c) how the different forms of CD may be distinguished at the level of the individual.

In providing a selective review of the imaging literature related to antisocial behavior and the development of CD/ODD, we make a central assumption that there is a fundamental distinction between reactive (mean age = 8.4 years; Steiner, Saxena, & Chang, Reference Steiner, Saxena and Chang2003) and instrumental (instrumental or premeditated; cf. Steiner et al., Reference Steiner, Saxena and Chang2003) aggression (cf. Barratt, Stanford, Dowdy, Liebman, & Kent, Reference Barratt, Stanford, Dowdy, Liebman and Kent1999; Berkowitz, Reference Berkowitz1993; Crick & Dodge, Reference Crick and Dodge1996; Linnoila et al., Reference Linnoila, Virkkunen, Scheinin, Nuutila, Rimon and Goodwin1983). Reactive aggression is triggered by a frustrating or threatening event, and involves unplanned, enraged attacks on the object perceived to be the source of the threat/frustration. This aggression type is often accompanied by anger and can be considered “hot.” It is important that it is initiated without regard for any potential goal. This is in direct contrast with instrumental aggression, which is purposeful and goal directed (e.g., to obtain the victim's possessions). Furthermore, instrumental aggression need not be accompanied by an emotional state, such as anger, and can be considered “cold” (Steiner et al., Reference Steiner, Saxena and Chang2003).

Drawing a distinction between reactive and instrumental aggression is critical for three main reasons. First, considerable amounts of data strongly support the existence of two relatively separable populations of aggressive individuals: individuals who present with mostly reactive aggression and individuals who present with high levels of mostly proactive and some reactive aggression (Barratt et al., Reference Barratt, Stanford, Dowdy, Liebman and Kent1999; Connor, Reference Connor2002; Crick & Dodge, Reference Crick and Dodge1996; Linnoila et al., Reference Linnoila, Virkkunen, Scheinin, Nuutila, Rimon and Goodwin1983).

Second, reactive and instrumental aggression rely on partially separable neural systems. As such, a principled division can be made among individuals who display antisocial behavior according to the form of the antisocial behavior displayed. We will argue below that the pathophysiologies associated with individuals who present with mostly reactive aggression are notably distinct from the pathophysiology associated with individuals who present with high levels of instrumental and reactive aggression.

Third, and similar to the second reason, there are notable differences in the pathophysiologies implicated in those psychiatric conditions associated with an increased risk for only reactive aggression (e.g., PTSD, childhood bipolar disorder, and intermittent explosive disorder [IED]) as opposed to the pathophysiology associated with psychopathy, where there is an increased risk for instrumental and reactive aggression (Cornell et al., Reference Cornell, Warren, Hawk, Stafford, Oram and Pine1996; Frick, Stickle, Dandreaux, Farrell, & Kimonis, Reference Frick, Stickle, Dandreaux, Farrell and Kimonis2005).

It is critical that a failure to distinguish between populations at risk for reactive versus instrumental and reactive aggression can confound interpretation of experimental results because individuals with very different pathophysiologies are likely to be combined in the same group.

The Neurobiological Basis of Reactive Aggression

The neural systems mediating reactive aggression have been extensively investigated in animal studies. All mammalian species appear to exhibit reactive aggression (Blanchard, Blanchard, Takahashi, & Kelley, Reference Blanchard, Blanchard, Takahashi and Kelley1977; Panksepp, Reference Panksepp1998), and it is part of the mammalian gradated response to threat. In work with rodents and using an approaching threat model, the Blanchards demonstrated that low levels of danger from distant threats induce freezing. Higher levels of danger from closer threats induce attempts to escape the immediate environment. Higher levels of danger still, when the threat is very close and escape is impossible, initiate reactive aggression (Blanchard et al., Reference Blanchard, Blanchard, Takahashi and Kelley1977).

Animal work indicates that this progressive response to threat is mediated by a basic threat system that runs from medial amygdaloidal areas downward, largely via the stria terminalis to the medial hypothalamus and from there to the dorsal half of the periaqueductal gray (Gregg & Siegel, Reference Gregg and Siegel2001; Panksepp, Reference Panksepp1998). This system is organized in a hierarchical manner such that aggression evoked by stimulation of the amygdala is dependent on the functional integrity of the medial hypothalamus and periaqueductal gray, but aggression evoked by stimulation of the periaqueductal gray is not dependent on the functional integrity of the amygdala (Gregg & Siegel, Reference Gregg and Siegel2001; Panksepp, Reference Panksepp1998). This amygdala–hypothalamus–periaqueductal gray neural system is thought to mediate reactive aggression in humans also (Blair, Reference Blair2004a). Furthermore, this system is regulated by higher order structures such as the orbital, medial, and inferior frontal cortices (Blair, 2004a).

Reactive Aggression and CD

The animal data briefly described above suggest two possible pathologies that would increase the risk for reactive aggression. First, the responsiveness of the basic threat system might be exogenously or endogenously elevated; second, the functioning of regulatory regions of frontal cortex might be disrupted. These forms of pathology are, of course, not mutually exclusive. Both could give risk to an individual who is more likely to express reactive aggression.

As noted, CD and ODD are comorbid with a variety of mood and anxiety disorders including PTSD and bipolar disorder (Steiner & Remsing, Reference Steiner and Remsing2007; Zoccolillo, Reference Zoccolillo1992). Moreover, both PTSD (Silva, Derecho, Leong, Weinstock, & Ferrori, 2001, Zoccolillo, Reference Zoccolillo1992) and childhood bipolar disorder (Danielyan, Pathak, Kowatch, Arszman, & Johns, Reference Danielyan, Pathak, Kowatch, Arszman and Johns2007; Faedda, Baldeseorini, Glavinsky, & Austin, Reference Faedda, Baldessarini, Glovinsky and Austin2004; Leibenluft, Blair, Charney, & Pine, Reference Leibenluft, Blair, Charney and Pine2003) are associated with an increased risk for reactive aggression. The following section briefly reviews the neuroimaging literature with respect to these disorders and examines whether either or both are associated with increased basic threat circuitry responsiveness and/or decreased frontal regulation of emotional responsiveness.

Increased Basic Threat Responsiveness, PTSD, and Reactive Aggression

PTSD is an anxiety disorder that can occur following exposure to trauma. CD/ODD shows significant comorbidity with PTSD (Dixon, Howie, Starling, & Franzep, Reference Dixon, Howie, Starling and Franzcp2005; Ruchkin, Schwab-Stone, Koposov, Vermeiren, & Steiner, Reference Ruchkin, Schwab-Stone, Koposov, Vermeiren and Steiner2002), and PTSD is associated with an increased risk for reactive aggression (Silva et al., Reference Silva, Derecho, Leong, Weinstock and Ferrari2001, Zoccolillo, Reference Zoccolillo1992). Moreover, exposure to trauma, including violence in the home and neighborhood, has long been linked with an increased probability of reactive aggression (Frazzetto et al., Reference Frazzetto, Di, Carola, Proietti, Sokolowska and Siracusano2007; Neller, Denney, Pietz, & Thomlinson, Reference Neller, Denney, Pietz and Thomlinson2006).

PTSD is also a particularly interesting disorder from a developmental and translational neuroscience perspective. It develops as a consequence of exposure to stressors, although the mechanisms by which stress triggers the development of PTSD in humans remain to be fully elucidated.

Studies in animals provide some insight into the link between stress and/or threat and the development of PTSD; considerable animal data clearly demonstrate that stress and/or threat exposure can cause lasting alterations in behavior, brain chemistry, and neural circuitry. These changes occur after both acute and chronic stressors, and are particularly pernicious developmentally following exposure in juvenile animals (Graham et al., Reference Graham, Heim, Goodman, Miller and Nemeroff1999; Nemeroff, Reference Nemeroff2004). After even a single-exposure to a potentially threatening traumatic event, such as predator scent, animals showing the most extreme responses (e.g., behavioral changes, alterations in the hypothalamus–pituitary–adrenal axis or autonomic nervous system) to the threat appear to retain these changes long term (Cohen, Matar, Richter-Levin, & Zohear, Reference Cohen, Matar, Richter-Levin and Zohar2006). Similar changes in neurochemistry and the hypothalamus–pituitary–adrenal axis have also been found in nonhuman primates after maternal rejection (Maestripieri, Higley, et al., Reference Maestripieri, Higley, Lindell, Newman, McCormack and Sanchez2006; Maestripieri, McCormack, et al., Reference Maestripieri, McCormack, Lindell, Higley and Sanchez2006), stressful foraging environments (Coplan et al., Reference Coplan, Andrews, Rosenblum, Owens, Friedman and Gorman1996), and when exposed to an acute stressor (Kraemer, Ebert, Schmidt, & McKinney, Reference Kraemer, Ebert, Schmidt and McKinney1989). Furthermore, regions of threat–response systems, such as the amygdala, appear to be increased in responsiveness as a result of early exposure to stressors (Blaise, Koranda, Chow, Haines, & Dorward, Reference Blaise, Koranda, Chow, Haines and Dorward2008; Kraszpulski, Dickerson, & Salm, Reference Kraszpulski, Dickerson and Salm2006; Sabatini et al., Reference Sabatini, Ebert, Lewis, Levitt, Cameron and Mirnics2007; Salm et al., Reference Salm, Pavelko, Krouse, Webster, Kraszpulski and Birkle2004).

As noted above, PTSD is a developmental consequence of exposure to threat. Consistent with the animal data demonstrating that stress and/or threat exposure increases the responsiveness of the basic threat system, neuroimaging work with adult patients with PTSD have also indicated heightened basic threat responses to threat related material (unfortunately, there have been no neuroimaging studies of children or adolescents with PTSD to our knowledge). Specifically, there have been relative relatively consistent findings of increased amygdala responsiveness in patients with PTSD to personalized traumatic narratives (Shin et al., Reference Shin, Orr, Carson, Rauch, Macklin and Lasko2004; Rauch et al., Reference Rauch, van der Kolk, Fisler, Alpert, Orr and Savage1996), trauma-related cues (Liberzon et al., Reference Liberzon, Taylor, Amdur, Jung, Chamberlain and Minoshima1999, Protopopescu et al., Reference Protopopescu, Pan, Tuescher, Cloitre, Goldstein and Engelien2005), and fearful facial expressions (Armony, Corbo, Clement, & Brunet, Reference Armony, Corbo, Clement and Brunet2005; Bryant et al., Reference Bryant, Kemp, Felmingham, Liddell, Olivieri and Peduto2008; Rauch et al., Reference Rauch, Whalen, Shin, McInerney, Macklin and Lasko2000; Shin et al., Reference Shin, Wright, Cannistraro, Wedig, McMullin and Martis2005; Williams et al., Reference Williams, Kemp, Felmingham, Barton, Olivieri and Peduto2006; although see Lanius et al., Reference Lanius, Williamson, Densmore, Boksman, Gupta and Neufeld2001; Shin et al., Reference Shin, McNally, Kosslyn, Thompson, Rauch and Alpert1999). Moreover, a recent voxelwise meta-analysis of 14 of these PET or fMRI studies, all of which focused specifically on processing of negative emotional stimuli (i.e., processing of emotional expressions, trauma-related stimuli or scripts, or emotional words), found greater activation in the amygdala as well as in the insula and midcingulate cortex in PTSD patients than in matched comparison subjects (Etkin & Wager, Reference Etkin and Wager2007).

In short, animal work demonstrates that exposure to threat increases the responsiveness of the basic threat circuitry. Work with patients with PTSD, a condition consequent on exposure to threat, also reveals increased responsiveness of the basic threat circuitry (particularly, the amygdala). The suggestion here is that there is an increased risk for reactive aggression in patients with PTSD because of this heightened responsiveness of the basic threat circuitry; a threat stimulus that might elicit freezing in a healthy individual is more likely to elicit reactive aggression in a patient with PTSD because the basic threat system is primed to respond by the earlier trauma.

In this context, it is worth noting that in addition to data indicating trauma associated with PTSD increases the responsiveness of the basic threat systems, PTSD has also been associated with decreased responding within regions of the middle prefrontal cortex (PFC) following presentation of traumatic narratives (Lanius et al., Reference Lanius, Williamson, Densmore, Boksman, Gupta and Neufeld2001; Shin et al., Reference Shin, Orr, Carson, Rauch, Macklin and Lasko2004), trauma-related cues (Bremner et al., Reference Bremner, Narayan, Staib, Southwick, McGlashan and Charney1999), fearful facial expressions (Shin et al., Reference Shin, Wright, Cannistraro, Wedig, McMullin and Martis2005; Williams et al., Reference Williams, Kemp, Felmingham, Barton, Olivieri and Peduto2006), and during the performance of emotional Stroop interference tasks (Bremner et al., Reference Bremner, Vermetten, Vythilingam, Afzal, Schmahl and Elzinga2004; Shin et al., Reference Shin, Whalen, Pitman, Bush, Macklin and Lasko2001; but see Bryant et al., Reference Bryant, Felmingham, Kemp, Barton, Peduto and Rennie2005; Shin et al., Reference Shin, Kosslyn, McNally, Alpert, Thompson and Rauch1997). Moreover, the recent meta-analysis revealed an apparent hypoactivation in PTSD patients in response to negative emotional stimuli in multiple regions including the ventromedial PFC (vmPFC), rostral anterior cingulate cortex, parahippocampal gyrus, dorsal amygdala and anterior hippocampus, orbitofrontal cortex, dorsomedial PFC, dorsal anterior cingulate cortex, and midcingulate cortex (Etkin & Wager, Reference Etkin and Wager2007). Given that these frontal regions may regulate the responsiveness of the basic threat circuitry (Blair, 2004a), it is possible that the increased risk for reactive aggression in patients with PTSD reflects a reduction of this frontal regulation.

Frontal Regulatory System Dysfunction and an Increased Risk for Reactive Aggression

Although the amygdala is thought to be relatively complete in its structural and functional maturity by puberty (Machado & Bachevalier, Reference Machado and Bachevalier2003, Yakovlev & Lecours, Reference Yakovlev, Lecours and Minowski1967), PFC continues to undergo structural and developmental changes during adolescence (Blumberg et al., Reference Blumberg, Kaufman, Martin, Charney, Krystal and Peterson2004). Work with primates shows significant regression of synaptic varicosities in PFC during adolescence (Bourgeois, Goldman-Rakic, & Rakic, Reference Bourgeois, Goldman-Rakic and Rakic1994; Lewis, Reference Lewis1997), whereas structural fMRI studies with humans show preadolescent increases, and postadolescent decreases, in cortical gray matter (Giedd, Reference Giedd2004; Giedd. Blumenthal, & Jeffries, Reference Giedd, Blumenthal and Jeffries1999). It has been suggested that these changes in PFC structure are associated with greater frontal modulation of emotional responding in the postadolescent period (Blumberg et al., Reference Blumberg, Kaufman, Martin, Charney, Krystal and Peterson2004). Following on from this, it is plausible that there are conditions where the appropriate development of regulation over the basic threat system is dysfunctional.

Early neurobiological models of antisocial behavior stressed a role for frontal dysfunction (Pennington & Bennetto, Reference Pennington and Bennetto1993; Raine, Reference Raine2002). These models, however, were relatively nonspecific with respect to the form of antisocial behavior (they typically did not distinguish reactive and instrumental aggression). Moreover, they frequently failed to specify which regions of frontal cortex are critical for antisociality, although implicitly they favored dorsolateral PFC, a region that does not appear to be integrally involved in the regulation of basic threat systems, and when damaged, is not associated with an increased risk for reactive aggression (Grafman et al., Reference Grafman, Schwab, Warden, Pridgen, Brown and Salazar1996). Nevertheless, these models were informative, and fostered an early neuroimaging literature on aggressive individuals (Critchley et al., Reference Critchley, Simmons, Daly, Russell, Van and Robertson2000; Goyer et al., Reference Goyer, Andreason, Semple, Clayton, King and Compton-Toth1994; Raine et al., Reference Raine, Buchsbaum, Stanley, Lottenberg, Abel and Stoddard1994; Raine, Buchsbaum, & Lacasse, Reference Raine, Buchsbaum and Lacasse1997; Raine, Phil, Stoddard, Bihrle, & Buchsbaum, Reference Raine, Phil, Stoddard, Bihrle and Buchsbaum1998; Schneider et al., Reference Schneider, Habel, Kessler, Posse, Grodd and Muller-Gartner2000; Soderstrom et al., Reference Soderstrom, Hultin, Tullberg, Wikkelso, Ekholm and Forsman2002; Soderstrom, Tullberg, Wikkeslso, Ekholm, & Forsman, Reference Soderstrom, Tullberg, Wikkelso, Ekholm and Forsman2000; Volkow & Tancredi, Reference Volkow and Tancredi1987; Wong et al., Reference Wong, Fenwick, Fenton, Lumsden, Maisey and Stevens1997; reviewed by Blair, Reference Blair, Charney and Nestler2004b). Most of these studies examined mainly reactive populations, with the exception of Raine (1998), who identified 9 reactive and 15 instrumental murderers, and found that relative to comparison individuals, the reactive murderers showed lower prefrontal functioning, whereas the instrumental group did not. Although many of these studies did not differentiate which specific prefrontal regions were dysfunctional, an investigation into the relationship between damage to PFC and aggression suggested that lesions to orbitofrontal and medial (as opposed to dorsolateral) PFC were associated with violence (Grafman et al., Reference Grafman, Schwab, Warden, Pridgen, Brown and Salazar1996). Similarly, Goyer et al. (Reference Goyer, Andreason, Semple, Clayton, King and Compton-Toth1994) examined cerebral blood flow using PET under resting conditions in 17 patients with personality disorder (antisocial, borderline, dependent, and narcissistic) and 43 comparison individuals and found that lower normalized cerebral blood flow in lateral orbitofrontal cortex (Broca area [BA] 47) correlated with a history of (mainly reactive) aggression. These studies provided early evidence that dysfunction of higher order cortical structures such as the orbital and medial PFC may contribute to aggression in particular populations, potentially via inappropriate regulation over the basic threat system.

There are several psychiatric conditions that appear to be related to dysfunction in frontal systems involved in the regulation of emotion. Two of these, IED and childhood bipolar disorder, have relatively recently begun to be examined using imaging technologies. Both of these disorders are developmental conditions in that they are thought to first manifest during childhood or adolescence. Furthermore, IED and childhood bipolar disorder are both associated with an increased risk for reactive aggression.

Impaired Frontal Regulation, IED, and Reactive Aggression

IED is categorized as a DSM-IV Axis I (American Psychiatric Association, 1994) impulse disorder that manifests as impulsive aggression that is disproportionate to the provocation. It emerges clinically during adolescence (Coccaro, Posternak, & Zimmerman, Reference Coccaro, Posternak and Zimmerman2005; Coccaro, Schmidt, Samuels, & Nestadt, Reference Coccaro, Schmidt, Samuels and Nestadt2004). Given that by definition IED is characterized by reactive aggression, it is logical to hypothesize that dysfunction in amygdala–prefrontal cortical regions may underlie IED's etiology, and there is some evidence supporting this idea. For example, interictal aggression occurring in people with temporal lobe seizure disorders has been related to IED, and quantitative MRI studies indicate decreases in frontal lobe grey matter (Woermann et al., Reference Woermann, van Elst, Koepp, Free, Thompson and Trimble2000) and amygdala atrophy (van Elst, Woermann, Lemieux, Thompson, & Trimble, Reference van Elst, Woermann, Lemieux, Thompson and Trimble2000) in patients diagnosed with temporal lobe epilepsy plus IED.

Imaging studies in people suffering from IED are limited. Behavioral data, however, suggest that the integrity of their amygdala–orbital/medial PFC network is compromised (Best, Williams, & Coccaro, Reference Best, Williams and Coccaro2002), and this appears to be supported by a recent fMRI study. In this study, individuals with IED had a greater blood oxygen level dependent (BOLD) response in the amygdala in response to angry faces compared with healthy control subjects (Coccaro, McCloskey, Fitzgerald, & Phan, Reference Coccaro, McCloskey, Fitzgerald and Phan2007). Furthermore, in the IED group, there was a breakdown of the amygdala–orbitofrontal cortex (BA 10 extending to BA 11 m) negative functional connectivity observed in healthy volunteers in response to angry faces. Notably, the BOLD responses to angry faces were correlated with scores on the lifetime history of aggression–aggression subscale (r = .546, p < .02).

Impaired Frontal Regulation, Childhood Bipolar Disorder, and Reactive Aggression

Notably, IED has been linked to bipolar disorder (McElroy, Reference McElroy1999; McElroy, Soutullo, Beckman, Taylor, & Keck, Reference McElroy, Soutullo, Beckman, Taylor and Keck1998), and the two disorders symptomatically overlap, although they are currently viewed as categorically different. As with IED, there is mounting evidence that regions involved in threat reactivity, especially the amygdala, may be altered in bipolar disorder. Importantly, amygdala dysfunction appears to be an early feature of the disorder as evidenced by structural, behavioral, and imaging data. Multiple structural imaging studies reported reduced amygdala volume in bipolar adolescents (average age of participants across studies is in the very late teens), a results that has been fairly consistent among studies (Blumberg et al., Reference Blumberg, Kaufman, Martin, Whiteman, Zhang and Gore2003, Reference Blumberg, Fredericks, Wang, Kalmar, Spencer and Papademetris2005; Chen et al., Reference Chen, Sassi, Axelson, Hatch, Sanches and Nicoletti2004; DelBello, Zimmerman, Mills, Getz, & Strakowski, Reference DelBello, Zimmerman, Mills, Getz and Strakowski2004). Structural studies also point to higher order dysfunction in childhood bipolar disorder, which may contribute to reactive aggression. Differences in PFC morphology, including ventrolateral PFC and anterior cingulate, have been reported in adults with bipolar disorder (Lochhead, Parsey, Oquendo, & Mann, Reference Lochhead, Parsey, Oquendo and Mann2004; Lyoo et al., Reference Lyoo, Kim, Stoll, Demopulos, Parow and Dager2004; McIntosh et al., Reference McIntosh, Job, Moorhead, Harrison, Forrester and Lawrie2004), but such findings have been less consistent in adolescents (mean age = 19.9 years; Adler et al., Reference Adler, DelBello, Jarvis, Levine, Adams and Strakowski2007). This may suggest that ventral PFC abnormalities occur later on in development, potentially resulting in an impairment of cognitive control over threat processing. If this is the case, it would indicate that the developmental trajectory of cognitive control processes over emotional responding differs between individuals with and without bipolar disorder.

Although there are few functional imaging studies investigating emotional processing in bipolar adolescents, there is some evidence that dysfunction in subcortical and prefrontal regions of the brain may occur early in development in this disorder. For example, in response to happy and angry faces, adolescents with bipolar disorder (mean age = 14.9 years) had greater activation than the healthy comparison group in the rostral anterior cingulate, right amygdala, caudate, and thalamocortical loop (Pavuluri, O'Connor, Harral, & Sweeney, Reference Pavuluri, O'Connor, Harral and Sweeney2007). These deficits in subcortical activation during emotional processing appear to extend into adulthood in bipolar disorder (Altshuler et al., Reference Altshuler, Bookheimer, Proenza, Townsend, Sabb and Firestine2005). Similar to the structural data for childhood bipolar disorder, bipolar adolescents have also been shown to have altered activity in cortical regions known to be involved in emotional regulation. For example, activation in orbitofrontal and dorsolateral PFC in response to angry faces was lower in bipolar adolescents than that occurring in the healthy comparison group (Pavuluri et al., Reference Pavuluri, O'Connor, Harral and Sweeney2007).

The Neurobiological Basis of Instrumental Aggression

Instrumental aggression is a form of instrumental behavior, that is, a behavior undertaken to achieve a particular goal. In the case of instrumental aggression, an individual uses violence in the expectation that this will allow the individual's goal to be met. When faced with a goal (“obtain $50”), the individual has a range of behavioral options (e.g., “find a job,” “mug someone”). Choosing between these choices will be a function first of availability (Can a job/potential victim be found?) and then of action costs (social punishment and the distress of the victims). Thus, an individual might choose an antisocial behavior because they are under social or situational pressure to do so, and in turn, the antisocial acts are simply the most viable options available. In this case, the instrumental aggression/antisocial behavior, whereas perhaps socially undesirable, need not be maladaptive; from the individual's perspective, it may be the only way to achieve their goals. An individual may also choose antisocial behavior because their calculation of the costs related to the actions, empathic responding, or use of this information is deficient. Instrumental aggression/antisocial behavior can thus be considered maladaptive if the decision making that leads to the choice of the action is dysfunctional.

A crucial process for socialization is the ability to form stimulus–reinforcement associations. To learn that hitting another individual is bad, a representation of this action must be associated with an aversive unconditioned stimulus. Considerable literature implicates the role of the amygdala in the formation and processing of stimulus–reinforcement associations (Baxter & Murray, Reference Baxter and Murray2002; Everitt, Cardinal, Parkinson, & Robbins, Reference Everitt, Cardinal, Parkinson and Robbins2003). The aversive unconditioned stimulus that is particularly important for moral socialization is the distress of other individuals such as the expressions of fear and sadness (Blair, Reference Blair2003). The expressions of fear and sadness serve as social reinforcers; as such, they allow caregivers to teach the societal valence of objects and actions to the developing individual (Blair, Reference Blair2003). In healthy developing children, actions or objects associated with the sadness and/or fear of others are avoided from an early age (within the second year of life). A considerable body of work attests to the power of facial expressions to transmit valence information in both humans and other primates (Blair, Reference Blair2003; Klinnert, Emde, Butterfield, & Campos, Reference Klinnert, Emde, Butterfield and Campos1987; Mineka & Cook, Reference Mineka and Cook1993). The amygdala is crucially involved in the response to fearful and sad expressions (Adolphs, Reference Adolphs2002; Blair, Reference Blair2003). Moreover, recent fMRI work has shown that the amygdala's response to fearful expressions is modulated by whether there is an object to make the association of the distress of the expression with (Hooker, Germaine, Knight, & d'Esposito, 2006). Specifically, Hooker et al. (Reference Hooker, Germine, Knight and D'Esposito2006) demonstrated that healthy adults show significantly greater amygdala activity to individuals displaying fear when looking at a novel object rather than when looking at an empty space. In short, the amygdala shows significantly greater activity when given an opportunity to engage in stimulus–reinforcement learning (stimulus = novel object, reinforcement = fearful expression) than when there is no novel object about which to learn.

Moral socialization by parents and peers helps the developing child appropriately represent the costs of antisocial behavior. An important temperamental variable influencing the ease with which a child can be socialized is fearfulness (Kochanska, Reference Kochanska1997; Steiner & Remsing, Reference Steiner and Remsing2007). We believe that this temperamental variable can be considered to index the functional integrity of the amygdala. The punishment that best achieves moral socialization is the victim's distress; empathy induction, focusing the transgressor's attention on the victim, particularly fosters moral socialization (Eisenberg, Reference Eisenberg, Davidson and Harrington2002; Hoffman, Reference Hoffman1994).

The above suggests that pathology interfering with the ease of socialization would increase the risk for instrumental aggression. It is argued that this is what has occurred in the developmental condition of psychopathy (Blair, Reference Blair2007). Certainly, the relationship between effective parenting techniques (e.g., low levels of corporal punishment, consistency, good monitoring, and high parental involvement) and decreased antisocial behavior that is seen in typically developing children (with low callous-unemotional [CU] traits), does not hold for children who present with the emotional dysfunction associated with psychopathy (mean age = 8.4 years; Oxford, Cavell, & Hughes, Reference Oxford, Cavell and Hughes2003; Wootton, Frick, Shelton, & Silverthorn, Reference Wootton, Frick, Shelton and Silverthorn1997).

Impaired Systems for Socialization and Psychopathy

Psychopathy is not currently included in DSM-IV (American Psychiatric Association, 1994) and should not be considered synonymous with CD or APD; fewer than 25% of those meeting diagnosis for CD or APD would meet cutoffs for psychopathy (Hare, Glass, & Newman, Reference Hare, Glass and Newman2006). Psychopathy has behavioral components (e.g., criminal activity and poor behavioral control), but it also has important affective-interpersonal components (notably, a lack of empathy and guilt; Frick et al., Reference Frick, O'Brien, Wootton and McBurnett1994; Harpur & Hare, Reference Harpur and Hare1994). It is a developmental disorder (Lynam et al., Reference Lynam, Caspi, Moffitt, Loeber and Stouthamer-Loeber2007) where recent work has confirmed the stability of CU traits in particular and the disorder more generally from childhood into adulthood (Lynam, Caspi, Moffitt, Loeber, & Stouthamer-Loeber, Reference Lynam, Caspi, Moffitt, Loeber and Stouthamer-Loeber2007, Munoz & Frick, Reference Munoz and Frick2007, Obradović, Pardini, Long, & Loeber, Reference Obradović, Pardini, Long and Loeber2007). Moreover, CU traits indexed in adolescents have been shown to predict adult measures of psychopathy in early adulthood (Burke, Loeber, & Lahey, Reference Burke, Loeber and Lahey2007, Lynam et al., Reference Lynam, Caspi, Moffitt, Loeber and Stouthamer-Loeber2007). In addition, the functional impairments seen in adults with psychopathy (e.g., in responding to emotional expressions, aversive conditioning, passive avoidance learning, reversal learning, extinction) are also seen, with some caveats (see below) in adolescents with psychopathic tendencies (Blair, in press).

There appears to be a strong genetic contribution to the emotion dysfunction that underlies the CU traits. For example, in a recent large study with around 3,500 twin pairs, CU traits were shown to be strongly heritable (67% heritability) at both 7 and 9 years (Blonigen, Hicks, Krueger, Patrick, & Iacono, Reference Blonigen, Hicks, Krueger, Patrick and Iacono2005; Viding, Blair, Moffitt, & Plomin, Reference Viding, Blair, Moffitt and Plomin2005). However, as yet, the details of this contribution, the molecular genetics, are almost completely unknown. It can be assumed, given the known pathophysiology (see below), that they influence the functional integrity of the amygdala and vmPFC. Moreover, several genes have been identified that do influence the functional integrity of these structures (serotonin transporter 5-HTTLPR, and monoamine oxidase A MAOA; Hariri et al., Reference Hariri, Drabant, Munoz, Kolachana, Mattay and Egan2005; Meyer-Lindenberg et al., Reference Meyer-Lindenberg, Buckholtz, Kolachana, R Harri, Pezawas and Blasi2006). However, we do not know whether these genes are implicated in psychopathy.

Social factors influence the manifestation of psychopathy. For example, socioeconomic status (SES) is associated with the emergence of the full syndrome; it is significantly less likely to appear in individuals of higher social status (Silverthorn & Frick, Reference Silverthorn and Frick1999). Reduced SES does not predispose toward reduced emotional and amygdala responsiveness; indeed, as a factor it is unrelated to the emotional component of psychopathy (Hare & Neumann, Reference Hare, Neumann and Patrick2006). However, reduced SES does supply a motive: a lack of finances. Reduced SES increases the risk for antisocial behavior in individuals, including those with psychopathy.

The core deficits seen in psychopathy all relate to the functional role of the amygdala in stimulus–reinforcement learning and in responding to fearful and sad expressions. The impairment of individuals with psychopathy in stimulus–reinforcement association is evidenced by their impairment in aversive conditioning, the augmentation of the startle reflex following the presentation of visual threat primes and passive avoidance learning (Flor, Birbaumer, Hermann, Ziegler, & Patrick, Reference Flor, Birbaumer, Hermann, Ziegler and Patrick2002; Levenston, Patrick, Bradley, & Lang, Reference Levenston, Patrick, Bradley and Lang2000; Newman & Kosson, Reference Newman and Kosson1986). Moreover, recent fMRI work has demonstrated that individuals with psychopathy show reduced amygdala responding during aversive conditioning relative to comparison individuals (Birbaumer et al., Reference Birbaumer, Veit, Lotze, Erb, Hermann and Grodd2005). Several studies across groups have now demonstrated impairment in the recognition of fearful and, to a lesser extent, sad expressions in individuals with psychopathy (Blair et al., Reference Blair, Colledge, Murray and Mitchell2001; Dadds et al., Reference Dadds, Perry, Hawes, Merz, Riddell and Haines2006; Dolan & Fullam, Reference Dolan and Fullam2006; but see Kosson, Suchy, Mayer, & Libby, Reference Kosson, Suchy, Mayer and Libby2002). In addition, fMRI work has indicated that individuals with psychopathic traits show reduced responding to fearful expressions (Gordon, Baird, & End, Reference Gordon, Baird and End2004). Other work has shown that individuals with psychopathy show reduced amygdala responding to emotional words in the context of emotional memory paradigms (Kiehl et al., Reference Kiehl, Smith, Hare, Mendrek, Forster and Brink2001) and less amygdala differentiation in responding when making cooperation relative to defection choices in a prisoner's dilemma paradigm (Rilling et al., Reference Rilling, Glenn, Jairam, Pagnoni, Goldsmith and Elfenbein2007).

The amygdala appears to feed reinforcement information associated with stimuli forward to medial orbitofrontal cortex, which then represents this outcome information. The function of this representation of reinforcement information within orbitofrontal cortex is to allow the normalization of competing outcomes so that the value of differing rewards such as apples and oranges can be compared (Montague & Berns, Reference Montague and Berns2002; Padoa-Schioppa & Assad, Reference Padoa-Schioppa and Assad2006; Schoenbaum & Roesch, Reference Schoenbaum and Roesch2005). This reinforcement expectancy information can then be used in decisions as to whether to approach the object or action or not. Given the suggestion made above, that the amygdala's role in stimulus–reinforcement learning is disrupted in psychopathy, it might be expected that individuals with psychopathy will show anomalous medial orbitofrontal cortex activity in the context of tasks that activate the amygdala. This is exactly what is seen. In Kiehl's study of emotional memory, the individuals with psychopathy not only showed reduced amygdala responses to the emotional words but also reduced rostral anterior cingulate cortex/medial orbitofrontal cortex activation (Kiehl et al., Reference Kiehl, Smith, Hare, Mendrek, Forster and Brink2001). There was also reduced medial orbitofrontal cortex activity in the individuals with psychopathy during aversive conditioning (Birbaumer et al., Reference Birbaumer, Veit, Lotze, Erb, Hermann and Grodd2005) as well as less medial orbitofrontal cortex differentiation in responding when making cooperation relative to defection choices in the prisoner's dilemma paradigm (Rilling et al., Reference Rilling, Glenn, Jairam, Pagnoni, Goldsmith and Elfenbein2007). Moreover, work has demonstrated reduced functional connectivity between the amygdala and vmPFC in adolescents with psychopathic traits when responding to emotional expressions.

Some recent fMRI work has specifically involved adolescent samples. Sterzer, Stadler, Krebs, Kleinschmidt, and Paustka (Reference Sterzer, Stadler, Krebs, Kleinschmidt and Poustka2005) reported reduced responding in dorsal anterior cingulate cortex in children with CD (mean age = 12.85 years) to negative images. In addition, following correction for anxiety and depressive symptoms, they also showed less differential responsiveness within the amygdala to negative relative to neutral images in the children with CD (Sterzer et al., Reference Sterzer, Stadler, Poustka and Kleinschmidt2007). However, these data must be treated with some caution given the very significant groups differences in IQ (p < .001). It was interesting that, in a follow-up study with IQ-matched groups and an optimized voxel-based morphometry protocol, these authors observed significant reductions in gray matter volume in bilateral anterior insular cortex and the left amygdala in the children with CD relative to healthy comparison individuals (Sterzer et al., Reference Sterzer, Stadler, Poustka and Kleinschmidt2007). Moreover, bilateral anterior insular gray matter volume was significantly correlated with self-reported empathy scores in the children with CD.

Three studies have examined children with psychopathic tendencies. Two of these compared responding to fearful and neutral expressions in adolescents with psychopathic tendencies (mean age = 14.5 years) relative to comparison children (E. Viding, personal communication; Marsh et al., Reference Marsh, Finger, Mitchell, Reid, Sims and Kosson2008). In both of these studies, the children with psychopathic tendencies showed a reduced amygdala response to the fearful expressions relative to the comparison populations. Moreover, the Marsh et al. study also reported reduced functional connectivity between the amygdala and vmPFC in children with psychopathic tendencies relative to comparison children. In addition, in this study, the children with psychopathic tendencies showed reduced amygdala responding to fearful expressions not only relative to healthy developing children but also children with ADHD. This is important because CD/psychopathic tendencies are frequently comorbid with ADHD (e.g., Colledge & Blair, Reference Colledge and Blair2001; Taylor, Schachar, Thorley, & Wieselberg, Reference Taylor, Schachar, Thorley and Wieselberg1986); roughly 50% of the children with psychopathic tendencies in the Marsh et al. study also met criteria for ADHD.

The third study examined reversal learning in children with psychopathic tendencies (mean age = 13.8 years; Finger et al., Reference Finger, Marsh, Mitchell, Reid, Sims and Budhani2008). In this study, children with psychopathic tendencies and comparison children performed a probabilistic reversal learning paradigm similar to that described by Budhani, Marsh, Pine, and Blair (Reference Budhani, Marsh, Pine and Blair2007). For this paradigm, the vmPFC showed a reduced BOLD response to punished incorrect reversal phases responses relative to correct responses in healthy adults (Budhani et al., Reference Budhani, Marsh, Pine and Blair2007). This was consistent with the suggestion that the vmPFC was responding to the prediction error, the failure to receive the anticipated reward. In the Finger et al. study, the healthy developing comparison children also showed significant reductions in BOLD responses in the vmPFC to punished incorrect reversal responses relative to correct responses. In contrast, the children with psychopathic tendencies did not.

The developmental trajectory of these deficits in individuals with psychopathic tendencies remains largely unknown. There is almost no work on the neurocognitive impairments associated with psychopathy in populations less than 10 years of age. The first fMRI studies of adolescent individuals with psychopathic tendencies have only emerged this year (2008). The genetic contribution might suggest an early emergence of the basic pathology, particularly the amygdala dysfunction, given that it is a relatively early maturing structure (Machado & Bachevalier, Reference Machado and Bachevalier2003; Yakovlev & Lecours, Reference Yakovlev, Lecours and Minowski1967). Certainly, the deficits associated with amygdala dysfunction (e.g., in aversive conditioning and passive avoidance learning) appear to be present in adolescent samples (Blair, Reference Blair2007). However, this has not been empirically confirmed. One difference that does appear between adolescent and adult samples with the disorder is the extent of the pathology, in particular, the deficit in reversal learning, linked to the vmPFC. This deficit is far more marked in adult individuals with psychopathic traits relative to adolescents (Budhani & Blair, Reference Budhani and Blair2005; Budhani, Richell, & Blair, Reference Budhani, Richell and Blair2006). This may indicate a later impact of the genetic contribution or of the amygdala dysfunction itself on the development of the vmPFC. However, again, the critical empirical studies have yet to be conducted.