LHPA, emotional dysregulation, and psychopathology

Because cortisol secretion is the end product of the activation of the LHPA system, the study of cortisol in a variety of contexts provides important information about various aspects of a person's stress response system. First, extremely high or low levels of cortisol at rest (basal) or induced by stress can be viewed as reflecting a dysregulation of the LHPA system (Heim, Ehlert, & Hellhammer, 2000) and they have been associated with distinct emotional regulation problems (Checkley, 1996; Gunnar, 2001; Tobin, 2001; ver Ellen & Van Kammen, 1990). Second, individual differences in normative cortisol levels have been linked to specific affect-related behaviors. This suggests that both normative and dysregulated LHPA function may have direct links with behavior and personality traits that are directly related to emotional regulation (Guerra, Nucci, & Huesmann, 1994; Gunnar, 1994).

For example, LHPA dysregulation, both in terms of resting (basal) hormonal activity and abnormalities of the activation or inhibition of hormonal levels after a stressor (reactive), has been linked to a variety of affective psychiatric disorders. Major depressive disorder and posttraumatic stress disorder are among the psychopathological disorders associated with LHPA abnormalities (Checkley, 1996; Gunnar, 2001; Tobin, 2001; ver Ellen & Van Kammen, 1990). A common interpretation of these findings is that a dysfunctional LHPA system may result in extreme physiological activation during possible stressful situations or elevated activation during day to day activities. When combined with dysfunctional emotional regulation processes (such as distorted cognitive appraisals of situations), this elevated stress response may create problems with the emotional reaction to both daily events and actual stressful situations, ultimately producing atypical levels of negative emotionality.

The literature linking dysregulation of the LHPA axis to affective disorders is extensive and continues to grow in a rapid manner. This literature underscores the importance of LHPA axis functioning on the regulation of emotional experience and expression. However, here we are interested in examining the evidence linking normal variations of LHPA functioning to individual differences in patterns of emotional regulation, social withdrawal, and aggression. Thus, we present evidence suggesting that normal LHPA circadian activity, as well as activation and regulation of the system during stress, are important components of the phenomenology of social withdrawal and aggression via affecting a number of emotional regulation processes.

LHPA and social withdrawal

The development of social withdrawal, or the tendency to avoid and withdraw from social interactions, has been linked to a number of neurophysiological processes, including the functioning of the LHPA system. Schmidt, Fox, Rubin, and Sternberg (1997) found that 4-year-old children with high morning salivary cortisol engaged in more wary behavior during peer play and were reported as shy by their mothers. In another study of LHPA functioning throughout the day, Dettling, Gunnar, and Donzella (1999) found that shyness was a predictor of high basal cortisol levels during the day among 6- to 8-year-old children. Furthermore, conduct disordered children with high basal levels of cortisol also reported high levels of anxiety (McBurnett, Lahey, Frick, Risch, Loeber, Hart, Christ, & Hanson, 1991). The association between cortisol and anxious behaviors appears to be present during infancy, suggesting a relatively stable developmental trajectory. For example, studies by Kagan and colleagues (Kagan et al., 1988; Kagan, Reznick, & Snidman, 1988) showed that infants who displayed high levels of negative affect to sensory stimulation were later identified as withdrawn in a variety of contexts when they were 4 years old. These children also presented higher basal cortisol levels than more highly sociable children. Consistent with these results, high levels of cortisol were also found among withdrawn anxious infants after the acute stressful challenge of a mother–infant separation (Tennes, Downey, & Vernadakis, 1997). The implications of these findings are considered in a later section, but it is possible that these children's overactive LHPA functioning (high basal cortisol levels) serves as a behavioral motivator by creating subtle but consistent levels of social stress. Children may attempt to regulate this stress by avoiding social situations.

The studies described above suggest that high cortisol levels may be a biological marker that differentiates shy from uninhibited children throughout many stages of development from infancy to adolescence and beyond. However, other studies have provided the opposite results, suggesting that specific biosocial contexts may significantly alter the relationship between cortisol and stress, at least during the school-age period. For example, Gunnar and colleagues presented a series of studies showing that behavioral inhibition may actually blunt LHPA activation during specific social situations, which results in a relative elevation of cortisol secretion among extroverted children (Gunnar, 1994; Gunnar, Tout, de Haan, & Pierce, 1997). They repeatedly found that extroverted children display higher levels of cortisol during the early stages of a school year when compared to shy and inhibited children (Gunnar, 1994; Gunnar, Tout, de Haan, & Pierce, 1997). Davis, Donzella, Krueger, and Gunnar (1999) also examined the cortisol levels of 7- to 12-year-old children during the first week of a new school year. They found highly assertive and social children had higher levels of cortisol than shy children, but this trend was reversed after the formation of peer groups. Similarly, de Haan and colleagues (1998) found that cortisol measures taken during a similar school period were correlated with assertive behavior in children. Likewise, high levels of cortisol were found among extroverted children after engaging in a competitive task (Donzella, Gunnar, & Krueger, 2000).

How can understanding the complexity of the LHPA axis help us clarify the findings linking high levels of cortisol to both inhibited and uninhibited children? First, most studies linking high levels of cortisol with shy and withdrawn behaviors have used basal resting levels of cortisol (Gunnar, 1994; Gunnar et al., 1997; Kalin, Larson, Shelton, & Davidson, 1998; McBurnett et al., 1991; Schmidt et al., 1997). Second, conversely, studies linking high levels of cortisol to outgoing and assertive behavior have measured cortisol after or during prolonged stressful situations (Davis et al., 1999; de Haan et al., 1998; Donzella et al., 2000; Gunnar, 1994; Gunnar et al., 1997). It is possible that inhibited children may have an overactive basal LHPA system (low basal inhibitory tone), which may be linked to increased levels of anxiety, withdrawn, and avoidant behaviors. When placed in social situations that may result in increased arousal (such as during the initial stages of group formation), these children may regulate negative arousal by avoiding social interactions. This may contribute to reduced activation of their LHPA system. In addition, it is possible that any activation of the LPHA may be difficult to observe because these children already present high basal levels of cortisol. Thus, the activation of the LPHA by a stressor should be measured in the context of each child's prestress cortisol levels.

In contrast, uninhibited children may have a readily active tonic inhibition of their LHPA system that results in low basal cortisol levels. This may lead to a high stress threshold that allows the child to engage in more stress-inducing behaviors, which in turn maximally elevates his/her cortisol levels (because the child is already starting from a low cortisol level under a resting condition). Thus, it is possible that the individual differences in cortisol activation found in various contexts among children with different levels of social withdrawal are partly related to MR functioning, which leads to different patterns of basal cortisol secretion and behavioral correlates.

Several animal models provide more insight into the role of GRs and MRs in the manifestation of inhibited and uninhibited behaviors (Kabbaj, Devine, Savage, & Akil, 2000). When rats are exposed to novel environments, they can be categorized according to their explorative behavior. High responders engage in high levels of exploratory behavior in a novel environment, whereas low responders engage in limited exploration (Kabbaj et al., 2000). These behaviors resemble the sensation seeking styles that have been identified in humans (Gerra et al., 1997, 1999; Gerra, Zaimovic, Timpano, Zambelli, Delsignore, & Brambilla, 2000). MRs and GRs are thought to be involved in the generation of exploratory behavior among rats because high responders and low responders differ in the density of GR expression in the hippocampus and because artificial manipulation of GR and MR activity results in a reversal of these behavioral characteristics. In a very influential study of the physiological makeup of high and low responder rats, Kabbaj and colleagues (2000) found that in a basal nonstressed state, compared to low responder rats, high responders have increased levels of CRH in the PVN but lower CRH levels in the central nucleus of the amygdala. High responder animals also have lower basal GR expression in the hippocampus compared to low responder animals. In contrast, both high and low responder animals have equal levels of MR expression in the hippocampus. It is interesting that high and low responder animals were found to have equal basal corticosterone levels. However, they differed in the magnitude of the corticoid response to novel stress. Although we may intuitively think that high levels of exploration would be associated with a low corticosterone stress response in high responder rats, these animals have a high corticosterone response to novelty whereas low responder animals have a blunted response. When further scrutinizing the biological markers described above, this is not entirely unexpected. The profile described in high responder animals is consistent with increased responsiveness to stress (high CRH in PVN), coupled with decreased anxiety behavior (low CRH levels in the amygdala, a structure associated with fear and anxiety, where CRH is not directly related to LHPA activity). In addition, low GR hippocampal expression also suggests a diminution in the inhibitory capacity observed as a sustained corticosterone release after novelty, which has been reported by some investigators in high responder animals (Kabbaj, Piazza, Le Moal, & Maccari, 1996; Piazza, Deroche, Deminiere, Maccari, Le Moal, & Simon, 1993).

GRs appear to be key in modulating the corticosterone response induced by novelty stress. Direct antagonism of hippocampal GR (blocking GR action) changes the behavior of a low responder rat that has high GR expression in the hippocampus into behaviors consistent with those of a high responder rat. This is evidenced by changes in their exploratory behavior and their corticosterone response in a novel environment (Kabbaj et al., 2000; Maccari, Piazza, Deminiere, & Angelucci, 1991; Strohle, Poettig, Barden, Holsboer, & Montkowski, 1998). This suggests that high responder rats engage in exploratory behaviors, despite experiencing high levels of stress during exploration. It is possible that high responders engage in exploratory behaviors precisely because of such physiological reactions, which drives risk-taking behavior. Favoring this possibility is the vast evidence suggesting that corticoids are reinforcing in rats, which lead them to engage in self-administration of psychostimulants or to substitute stress for psychostimulant administration (Piazza et al., 1993). Finally, given the high level of distribution of GRs in the brain, including the limbic system, the activation of GRs may have an impact on multiple neuronal pathways, which may result in avoidant or escape behaviors. Thus, blocking of GR activation or a reduced number of GRs may prevent an organism from activating avoidant or escape behaviors, resulting in more exploratory and risk-taking activity.

These animal models may provide insight into the underlying neural mechanisms that modulate inhibited and uninhibited human behavior. Although we must be careful in translating the significance of novelty as a stressor in rodents to its equivalent in human children, we could extrapolate based on the high–low responder animal model that perhaps extroverted children should not have an underactive LHPA response to stress when exposed to unfamiliar surroundings or novel social interactions. As with the high responder rats, these children may experience a physiological stress reaction when engaging in assertive social behaviors. Their high levels of assertive social behaviors cannot be explained by an “underactive” LHPA response to stressful situations. Instead, their behavior could be explained by a sensitive LHPA system, perhaps high levels of CRH in the PVN that are just waiting to be released upon demand, decreased levels of CRH in fear-related centers (amygdala), and reduced GR expression in the hippocampus that limits the occurrence of avoidant behaviors. As in the high responder animals, this combination would result in increased or normal responsiveness to stress but a decrease in anxiety-like behaviors when exposed to or prior to novel social environments. Likewise, it is plausible that inhibited children have high levels of CRH in the amygdala and high GRs in the hippocampus, both of which could explain their nonexploratory freezing behavior that is displayed during stressful situations and their low levels of cortisol that are secreted during these situations because their actual exposure to the novel situation is self-regulated through avoidant behaviors.

The application of animal models to human data appears adequate because the interpretation of these results is consistent with the work by Gunnar and colleagues, who have shown that children with different temperamental characteristics have different LHPA activation when placed in specific environmental contexts. During situations that may lead to stressful situations, such as the start of a school year, inhibited children display very low levels of cortisol whereas uninhibited children display high levels of cortisol (Davis et al., 1999; Donzella et al., 2000; Gunnar, 1994; Gunnar et al., 1997). They postulate that inhibited children engage in nonsocial behaviors in order to avoid possible stressful situations, which results in decreased levels of cortisol. Conversely, uninhibited children actively seek social interactions despite their normal LHPA response to stress, which in novel contexts may explain their high LHPA activation during these encounters. Considering the multiple functions that cortisol subserves, it is possible that elevations of cortisol prior to these encounters lead inhibited children to engage in socially avoidant behaviors. Instead, uninhibited children seek these social interactions that activate their LHPA axis. Their active LHPA axis mobilizes fuel sources (glucose) for heightened activity during these encounters. In addition, facilitation of cognitive function and association, consolidation, and memory processes are linked to elevated (but optimal) levels of corticoids in plasma during learning tasks (de Kloet, 1991). It is possible that during social encounters the uninhibited children are consolidating and processing information about social situations that would allow them to be much more adept in future social interactions. Indeed, this may very well explain the lower cortisol levels detected in uninhibited children later in the school year once peer groups are established.

LHPA and aggression

The literature linking LHPA functioning and aggression is also seemingly contradictory. Although a large number of studies have found an inverse association between cortisol and aggression, other studies have failed to find this association or have found the opposite results. Thus, both blunted and overactive LHPA functioning has been found among aggressive children of different ages. McBurnett, Lahey, Rathouz, and Loeber (2000) studied the relationship between basal cortisol levels and aggression among elementary school boys (ages 7–12 years) participating in a longitudinal study. They found that low basal cortisol levels in the morning were associated with high levels of aggressive conduct disorder, peer ratings of aggression, and symptoms of oppositional defiant disorder. Other studies of young children have indicated that low levels of morning cortisol were associated with hostility toward teachers and aggression toward peers (Tennes & Kreye, 1985; Tennes, Kreye, Avitable, & Wells, 1986). Low morning cortisol levels were also found among school-aged children reporting clinically significant levels of externalizing behavior problems (Cicchetti & Rogosch, 2001b). In addition, studies of adolescents and young adults have provided similar results. For example, Vanyukov, Moss, Plail, Blackson, Mezzich, and Tarter (1993) found that low morning basal cortisol levels were associated with increased conduct disorder symptoms among 150 adolescents.

However, a small but notable number of studies have failed to replicate these findings (Gerra et al., 1997; Kruesi, Schmidt, Donnelly, Hibbs, & Hamburger, 1989; McBurnett et al., 1991; Schulz, Halperin, Newcorn, Sharma, & Gabriel, 1997). Kruesi et al. (1989) compared adolescents with aggression and conduct problems to nonaggressive adolescents. They reported no significant differences in urinary cortisol levels among these groups. It is important to note, however, that Kruesi and colleagues measured 24-hr urinary cortisol output, which integrates intraday variations that could eliminate differences among the groups that may be present in discrete times of the day. A similar study of aggressive and nonaggressive adolescents also failed to find a relationship between morning plasma basal cortisol levels and aggression (Schulz et al., 1997). Furthermore, some studies indicate that high levels of cortisol are associated with high levels of aggression in some contexts. Gerra et al. (1997) experimentally induced aggression in young adults (18–19 years old) and found that participants with high basal cortisol levels responded more aggressively than participants with low basal cortisol. McBurnett et al. (1991) also found that 9-year-old children who display both conduct problems and anxiety also have higher levels of cortisol than normal children. Nevertheless, we should note that cortisol collection in this study was conducted during a clinic visit, which varied in time during the day. Thus, it is not clear whether such findings represent variations in basal or stress-induced cortisol levels.

The same principles that help explain the relationship between LHPA functioning and social withdrawal may explain the relationship between LHPA and aggression. LHPA functioning does not have a direct causal relationship with aggression. Instead, it affects emotional regulation by determining chronic and event-related stress levels and by eliciting regulatory behaviors. When these LHPA activation patterns interact with other physiological systems and environmental contexts, the likelihood of occurrence of specific aggressive responses is maximized. For example, children with low basal cortisol levels may have a higher anxiety threshold (Kruesi et al., 1989), which would prevent them from experiencing anxiety and guilt when engaging in inappropriate behavior and/or in anticipation of discipline encounters. Their basal, MR mediated system may be overactive (high basal tone). This overactivation of MRs would result in reduced basal cortisol levels around their circadian clock. However, their LHPA response to stress may be intact. When a child with high basal tone of the LHPA system (low basal cortisol levels) is placed in a context where aggression is either expected or necessary, he/she may lack the inhibitory capacity to engage in escape or avoidant behaviors and may proceed to engage in aggressive behaviors. Cortisol levels would then become maximally elevated as the result of the aggressive encounter (greater delta and greater mobilization of glucose for energy use).

Therefore, the data on cortisol and aggression may be highly sensitive to the contextual sampling procedure and the coping mechanisms evoked. At least during early childhood, if cortisol levels are measured in a familiar, nonstressful setting, low levels of cortisol may be more likely to predict high levels of aggression. However, if taken in a context where aggressive interactions are prevalent, high levels of cortisol may better indicate high levels of aggression. Because the LHPA axis is a dynamic system, these relationships may change over time as the LHPA axis is exposed to these patterns of secretion. Thus, in the adolescent and young adult the neuroendocrine system may have a new “setpoint” at which basal cortisol levels are now elevated as the result of “downregulation” or decreased expression of corticoid receptors in key brain regions as a direct consequence of high cortisol response for many years. Although higher basal cortisol levels may prevail in this age group in comparisons to nonaggressive controls, the peak response to stressful challenge between these groups may not differ significantly. Rather, the change between basal to peak levels would be different (i.e., a small change in aggressive adolescents and a large change in nonaggressive adolescents; Dallman, Akana, & Scriber, 1991).

Another possible explanation of the conflicting findings concerning the relationship between cortisol and aggression may be found in non-human primate models of aggression. Similar to Dodge's concept of proactive and reactive human aggression (Dodge & Coie, 1987; Dodge, Lochman, Harnish, Bates, & Petit, 1997), Kalin (1999b) and colleagues (1998) suggest that there are two specific types of primate aggression: offensive and defensive. Defensive aggression is motivated by excessive fear, whereas offensive aggression has its roots in difficulty with impulse control and underarousal. Monkeys display these two types of aggression and the neuroendocrinological correlates for each aggression type are unique (Kalin, 1999a). Monkeys displaying defensive aggression have increased asymmetry of their PFC, an area of the brain that filters and associates multiple neural pathways to allow the formal execution of a task or behavior (executive function). Monkeys displaying defensive aggression also have high morning basal levels of cortisol, exhibit a stable state of hypervigilance, and display aggressive behaviors in response to minor threats (Kalin, 1999a, 1999b). Conversely, monkeys displaying offensive aggression have low serotonin levels in their cerebrospinal fluid, high plasma testosterone levels, and low basal cortisol. Monkeys displaying offensive aggression are uninhibited and active and display aggressive behaviors in an impulsive and unpredictable manner.

Low brain serotogenic activity and high plasma testosterone levels have been consistently associated with decreased impulse control and increased levels of aggression in humans (Cleare & Bond, 1997, 2000; Gerra et al., 1997; Kalin, 1999b; Van Goozen, Frijda, & Van de Poll, 1994; Van Goozen, Matthys, Cohen–Kettenis, Thijssen, & van Engeland, 1998). The low levels of cortisol found among monkeys displaying offensive aggression may also create a state of underarousal, which could result in predatory aggression when combined with low brain serotonin levels and high circulating testosterone. However, monkeys with low levels of basal cortisol also displayed high levels of defensive aggression in particular contexts. This suggests that, although LHPA overactivation (high basal cortisol) is associated with increased levels of defensive aggression only, LHPA underactivation (low basal cortisol) is associated with increased levels of both defensive and offensive aggressive responses.

Therefore, it appears that the relationship between LHPA and aggression is not only highly sensitive to the context in which the cortisol samples are taken but also may be closely related to the subtype of aggressive behavior displayed by the individual. We can again extrapolate these findings to understand the literature on humans. It is possible that the fairly consistent findings associating low basal levels of cortisol to aggression among children and adolescents may be a by-product of the methodology used on these studies. The method use to measure aggressive behavior may oversample events of “offensive aggression.” Given the non-human primate findings, even if the methodology used equal samples of offensive and defensive aggression, it would not be surprising to find an association between low levels of basal cortisol and aggression because such an LHPA profile is associated with high levels of both offensive and defensive aggression. Instead, high levels of cortisol and aggression would be expected among children and adolescents engaging in mostly defensive aggressive acts and during aggressive interactions or highly aggressive contexts.

Amygdala

Several structures within the LHPA system have direct and indirect connections to the amygdala (Afifi & Bergman, 1998). For example, the amygdala sends signals to the hippocampus through direct, temporal lobe white matter pathways. The amygdala also sends signals to the hypothalamus and PFC through the ventral amygdalofugal pathways. Input signals are received from the PFC, all sensory systems, and the thalamus. Because of the strong links between the amygdala and limbic structures, it is not surprising that the amygdala has been associated with emotional regulation, and consequently, with both aggression and social withdrawal, mostly through its role in modulating fear-induced behaviors. The study of the amygdala and its relation to emotion, memory, social behaviors such as reproduction and aggression, and the modulation of the autonomic and neuroendocrine systems follows three different lines of research. These include case studies of patients with anterior temporal lobe damage, lesion studies using rodents and non-human primates, and more recent imaging studies using clinical and nonclinical human populations. Based on these findings, the amygdala has been viewed as a key component in the circuitry of emotional regulation and experience (see Davidson, Jackson, & Kalin, 2000, for a review). We now present a brief review of this literature, paying special attention to research linking the amygdala to the activation and regulation of the LHPA axis.

The amygdala appears to play a key role in the processing of emotional sensory information. For example, clinical neuropsychological studies examining the cognitive effects of bilateral damage to the amygdala indicate impaired processing of facial emotions. Specifically, bilateral lesions to the amygdala can result in an inability to recognize negative emotions and difficulty in learning procedures that involve emotional components (Adolphs, Tranel, Hamann, Young, Calder, Phelps, Anderson, Lee, & Damasio, 1999; Boucsein, Weniger, Mursch, Steinhoff, & Irle, 2001; Calder, Young, Rowland, Perrett, Hodges, & Etcoff, 1996; Hamann & Adolphs, 1999).

In the largest study of humans with amygdala damage, Adolphs and colleagues (1999) found that subjects with bilateral amygdala damage were significantly impaired in recognizing negative facial emotions, particularly fear. However, the subjects were not impaired in recognizing neutral or positive emotions such as happiness. Adolphs et al. suggested that the amygdala is a component of a complex neural system that recognizes facial cues signaling the presence of danger. They argued that this is consistent with animal studies indicating that damage to the amygdala results in an inability to recognize potentially dangerous stimuli. However, they also noted that individual performance on the facial recognition task varied from complete impairment to normal performance. It is not clearly understood why some patients with complete bilateral amygdala lesions did not have problems recognizing all facial expressions including fear. However, it is possible that, given the length of time between the time of the injury and the time of assessment, some patients had already developed compensatory strategies or new neural pathways to help them identify negative facial emotions. Hamann and Adolphs (1999) examined the ability of two subjects with complete bilateral amygdala damage to recognize similarities among similar emotions. These patients were able to correctly identify the similarities among all emotions. This suggests that the inability to recognize negative emotional expressions among people with bilateral damage to the amygdala may be related to an inability to provide meaning to specific emotions rather than an inability to recognize the features of an emotional expression. Furthermore, there is also evidence that the amygdala's role in the recognition of negative emotions is not limited to facial expressions, because bilateral damage to the amygdala also appears to affect recognition of auditory stimuli that represent danger or anger (Scott, Young, Calder, Hellawell, Agglenton, & Johnsons, 1997). More recent functional magnetic resonance imaging (fMRI) studies provide similar results indicating that, when clinical and nonclinical subjects are exposed to stimuli that are associated with fear or anxiety, there is significant activation of the amygdala (Morris, Frith, Perrett, & Rowland, 1996; Phillips et al., 1997; Rauch, van der Kolk, Fisler, & Alpert, 1996). In particular, the exposure to faces showing expressions of fear has been found to consistently activate the amygdala (Morris et al., 1996; Phillips et al., 1997).

Studies of non-human primates suggest that the amygdala can also modulate the behavioral response to fearful stimuli. Although early studies have reported that lesions to the amygdala result in drastic behavioral and personality changes usually known as the Kuver–Busy syndrome (reduced fear responses, hypersexuality, and hyperorality; Aggleton & Passingham, 1981), more recent studies indicate that lesions to the amygdala reduce behavioral responses to some unconditioned fear responses (such as the exposure to a snake), yet fear responses to some social human interactions remain intact (Kalin, Shelton, Davidson, & Kelley, 2001). For example, Kalin and colleagues (2001) were able to lesion the amygdala of several sexually mature rhesus monkeys while sparing adjacent areas as well as fibers passing through the amygdala. In adult male monkeys, these lesions reduced fear responses after exposure to a snake. However, lesioned monkeys did not reduce their freezing behavior during a human intruder paradigm and had levels of frontal lobe asymmetry similar to those of nonlesioned monkeys. Kalin suggested that the amygdala mediates behavioral responses to acute fear, but it does not affect behaviors that reflect more stable behavioral traits such as social anxiety. These findings have important implications about the role of the amygdala in influencing the behavioral responses to fear and its relationship to social behaviors such as withdrawal and aggression.

The point in development at which neuronal pathways are interrupted also appears to be important. Amaral (2002) is currently following a group of macaque monkeys that endured bilateral amygdalectomy shortly after birth. In contrast to monkeys that had amygdala lesions as adults, these monkeys do not show a blunted behavioral response to a social stressor. Instead, they appear extremely fearful when exposed to an unfamiliar animal. This indicates that there are discrete temporal windows in which the amygdala is developing its connections to other brain structures and cortical associative areas that are important for the development and modulation of fear-related behaviors. Thus, this implies that the time that an insult is created during development, be it anatomical, toxic, or “experiential,” may lead to very different behavioral and perhaps neuroendocrine manifestations later in life.

The amygdala appears to be related to social withdrawal and aggression in at least two specific ways. Because the amygdala plays a role in the activation of the LHPA, dysregulation of the amygdala may affect LHPA functioning. Although there are few direct amygdala connections to the PVN in the hypothalamus, a number of researchers have argued that social inhibition may be the result of an overstimulated amygdala (much like the “freezing” response to fear in monkeys). Freezing behavior or social inhibition may underlie the observed elevated basal levels of cortisol found among inhibited–shy children in particular contexts (Kagan, 1994; Kagan et al., 1988; Kagan, Reznick, & Snidman, 1988; Schmidt et al., 1997). Thus, the amygdala may indirectly affect LHPA functioning and thus influence both social withdrawal and aggression.

There are some conflicting reports about the role of the amygdala in affecting LHPA functioning, which may again be related to the extent of the lesions performed by the investigator and/or the age of the animal at the time of the intervention. Nevertheless, stimulation of the amygdala results in increased corticoid production (Feldman, Conforti, & Siegel, 1982), whereas damage to this structure (specifically, the central nucleus of the amygdala) results in blunted LHPA activity (ACTH release) in response to physical stress (Van de Kar, Piechowski, Rittenhouse, & Gray, 1991). Despite the blunted ACTH response, the adrenocortical response to stress (cortisol response) is unaffected by bilateral amygdala lesions in adult rhesus monkeys (Kalin et al., 2001), suggesting that the adrenal response is sensitized to low ACTH levels following amygdala damage. Furthermore, basal tonic cortisol levels are likewise unaffected (Kalin et al., 2001). In conclusion, these studies suggest that, although the amygdala may be related to LHPA functioning, it is neither necessary nor sufficient for the regulation of tonic LHPA function or the adrenocortical response to stress.

The amygdala has also been implicated in the facilitation of fear-induced learning and the expression of fear-related behavior. Particularly, animal and human studies suggest that the amygdala facilitates the learning of emotional information as well as stimulus–response (SR) contingencies related to fear (Goldstein, Rasmusson, Bunney, & Roth, 1996; Rosen, Hamerman, Sitcoske, Glowa, & Schulkin, 1996). For example, Goldstein and colleagues (1996) found that amygdala lesions in rats prevented the learning of simple SR contingencies and blocked previously learned responses. Damage to the amygdala may therefore prevent the animal's anticipation of the possible consequences of behavior.

It is possible that this property of the amygdala directly affects the development of aggression in some contexts. For example, some researchers have explained the evidence linking low cortisol levels to aggression to be a function of reduced stress response during discipline encounters (Kruesi et al., 1989). Kochanska (1993, 1994, 1995, 1997) has suggested that children with different temperamental predispositions may benefit from different types of parenting styles. It is possible that some children's failure to learn SR contingencies and the inability to consider the possible consequences of behavioral responses may be partly related to the level of amygdala function and may be impacted by specific parenting styles. Thus, the low cortisol levels found among some of these children may be related to the differential development of the functional amygdaloid–hippocampal circuitry that indirectly influences key regulatory elements of the LHPA axis. Recent advances in fMRI technology could help answer this question. It would be revealing to observe the pattern of amygdala activation during SR learning paradigms among normal children and those with externalizing behavior problems. Unfortunately, most of the research on amygdala functioning has been based on animals or humans with severely damaged amygdalas. Moreover, studies focusing on normal variations of amygdala functioning and the developmental trajectory of this structure are noticeably absent.

Prefrontal lobe asymmetry

The amygdala and other limbic structures have unilateral and reciprocal connections with the forebrain, particularly with the prefrontal cortical regions (Afifi & Bergman, 1998). It is therefore relevant that a number of studies of the neurophysiology of emotional expression have implicated several areas of the PFC in the processing and expression of both negative and positive emotions. Specifically, activation of the left prefrontal areas is associated with the experience and processing of positive emotions whereas activation of the right prefrontal areas is associated with negative emotions (Gainotti, 1972; Gladue, 1991; Morris, Robinson, Raphael, & Hopwood, 1996). Extensive reviews of this relationship have already been presented (Davidson et al., 2000). Thus, we only provide a brief summary of the literature and focus on discussing how understanding the neural projections between the PFC and several limbic structures can help us clarify the relationship between cerebral asymmetry and emotional regulation as it pertains to aggression.

The role of cerebral asymmetry in the regulation of emotional expression has been reported in the clinical neuropsychology literature since the early 1970s. These studies showed that patients with damage to the left side of their brain tended to report more depressive symptoms than did patients with right hemisphere injuries (Gainotti, 1972). It has been argued that the disruption of specific circuitry within the left cortex may preclude the expression of positive affect, which in turn may result in depressive symptomatology. In addition, damage to the left hemisphere results in a marked cerebral asymmetry favoring activation of the right hemisphere, which controls the regulation of negative affect (Davidson, 1993; Gainotti, 1972; Morris et al., 1996).

A number of studies using EEG recordings of cerebral activation have implicated the right frontal lobe in the activation of negative affect and related behavioral responses. Nathan Fox and colleagues have found evidence linking frontal asymmetry in the regulation of emotions, for example, 7- to 22-month-old infants showed marked right frontal activation during a mother–infant separation paradigm (Fox, 1994; Fox, Bell, & Jones, 1992; Fox, Calkins, & Bell, 1994; Fox, Rubin, Calkins, Marshall, Coplan, Porges, & Long, 1995; Fox, Schmidt, Calkins, & Rubin, 1996). Conversely, socially outgoing toddlers showed increased left frontal activation during social situations (Fox et al., 1992). This group of investigators also found that relative right frontal asymmetry is associated with later behavioral problems among both socially inhibited and uninhibited toddlers, suggesting that right frontal asymmetry is related to negative emotionality (Fox et al., 1996; Rubin, Coplan, Fox, & Calkins, 1995).

Other experimental studies with adults have shown similar results (Davidson, Ekman, Saron, Senulis, & Friesen, 1990; Wittling & Genzel, 1995). Davidson et al. (1990) found a marked activation of the right PFC and right temporal lobe after the presentation of an aversive film and a contralateral activation of the same regions after the presentation of a film designed to elicit positive affect. By presenting aversive films to subjects targeting specific hemispheres, Wittling and Genzel (1995) expanded on Davidson et al.'s findings. They were able to present an aversive or neutral film to either the left or the right visual field of a subject and found that a physiological reaction to the aversive film occurred only after it was presented to the left visual field. This suggested that the right hemisphere is the predominant regulatory site for negative emotions. There is also evidence linking basal patterns of cerebral asymmetry to specific temperaments and social behaviors. Particularly, relative right prefrontal asymmetry has been found among infants with more difficult temperaments and in socially inhibited children (Davidson & Fox, 1989; Schmidt et al., 1997; Schmidt, Fox, Schulkin, & Gold, 1999; Schmidt, Fox, Sternberg, Gold, Smith, & Schulkin, 1999).

In order to understand how PFC asymmetry relates to emotional regulation and aggression, we must explore the relationship between the PFC and other subcortical structures previously identified as major regulators of emotional experience and expression. In addition, given the role of the amygdala and LHPA axis in the regulation of emotional experience, we can expect the PFC to be functionally and anatomically related to these two systems. In fact, a number of investigators have recently proposed ways to explain the possible circuitry between the PFC and these subcortical structures (Davidson & Irwin, 1999; Feldman, Conforti, & Weidenfeld, 1995). Specifically, reciprocal neural projections between the PFC and the amygdala can be used to explain the role of the PFC in regulating amygdala function and the role of these two systems in the processing and integration of aversive SR contingencies. Furthermore, specific neural projections from the PFC to several subcortical structures related to the LHPA axis (e.g., hippocampus, subiculum) may explain how the PFC may influence the regulation of the LHPA system (Feldman et al., 1995).

The PFC has direct neural projections from a variety of limbic structures that are directly linked to the amygdala (Afifi & Bergman, 1998). There are input projections from the thalamus to the PFC, and these connections contain information arising from the temporal cortex and the amygdala. Direct reciprocal connections from the PFC to the amygdala have also been identified (Afifi & Bergman, 1998). Output projections to the amygdala are both excitatory and inhibitory in nature. However, damage to the PFC results in an overactivation of the amygdala, suggesting that the effect of the PFC on the amygdala is predominately inhibitory (Gewirtz, Falls, & Davis, 1997; Morgan, Romanski, & LeDoux, 1993). How does this translate to the functional domain? Lesions to the PFC in rats reduce the PFC inhibitory action on the amygdala, resulting in increased difficulty in the facilitation of extinction of aversive responses (Morgan et al., 1993). In humans, damage to the PFC impairs the ability to anticipate future negative consequences (Bechara, Tranel, Damasio, & Damasio, 1996). Therefore, it appears that the PFC plays an important role in regulating the acquisition of aversive contingencies, the acquisition of new ones, and the extinction of aversive conditioned responses.

The role of the PFC in the acquisition and extinction of aversive conditioned behaviors may help us understand the relationship between cerebral asymmetry and emotional regulation. Right frontal lobe asymmetry has been associated with withdrawal-related behaviors, such as those activated to escape or avoid aversive stimuli (see Davidson & Irwin, 1999, for a recent review). These behaviors resemble aversive SR responses, whose extinction may be modulated by the PFC. Thus, it is possible that right prefrontal asymmetry results in reduced inhibitory signals toward the amygdala, which would then maintain aversive-related responses. This hypothesis can be applied to explain the relatively high rates of avoidant behaviors displayed by socially inhibited children. New social interactions are intrinsically stressful, thus explaining the findings showing that socially uninhibited children experience high levels of stress (as measured by cortisol secretion) during the early weeks of the school year when social groups are being formed (Davis et al., 1999; de Haan et al., 1998). The stress perceived by these children eventually subsides after repeated exposures to the same social situations. This can be viewed as a form of extinction or habituation. That is, with repeated social interactions, the events (or expectations of these events) that may have previously elicited stress (such as rejection and ridicule by unknown peers) are no longer present or expected, resulting in the extinction of the stress response. Children with right cerebral asymmetry may have difficulty habituating and reducing such stress response, which may result in increased rates of avoidant behaviors.

Paradoxically, damage to the PFC is also associated with the inability to regulate behavior based on possible aversive consequences. This can result in a pattern of impulsive behavior with little anticipation or recognition of the consequences, a pattern common among children with externalizing behavior problems. Because of their inability to anticipate negative consequences from their behavior, these children may be resistant to traditional methods of socialization, which require a basic understanding of SR relations. Therefore, severe asymmetry of PFC activation may be related to deficits in the extinction of aversive conditioned responses, which may explain the consistent association found between the PFC and negative affective style and withdrawal-related behaviors. Conversely, normal PFC functioning may be necessary to regulate behavior based on anticipated consequences, the lack of which could explain some externalizing behavior problems.

As mentioned previously, the PFC also appears to be involved in the regulation of LHPA activation. In a series of studies with non-human primates, Kalin and colleagues (1998) found an association between extreme right frontal asymmetry and elevated basal levels of cortisol. Although these findings do not provide concrete evidence explaining how these two processes may be related, the data from Wittling and Genzel's study (1995) suggest that right hemisphere activation is involved in the activation of the LHPA axis in humans. In this study, presentation of an aversive film to the right hemisphere (left visual field) resulted in increased cortisol levels. Presentation of the same film to the left hemisphere resulted in reduced cortisol levels, which were comparable to those produced by the presentation of a neutral film to each hemisphere. It appears that the processing of the aversive component of the film and/or the physiological reaction to the aversive film is partly under the control of the right hemisphere. Although there are direct connections between the PFC and the hypothalamus through the medial forebrain bundle (Afifi & Bergman, 1998), it is not clear if the excitatory effect of the right hemisphere on the LHPA axis is directly through PFC activation.

In conclusion, right prefrontal asymmetry has been linked to the processing and expression of negative emotions. Right prefrontal activation has also been associated with LHPA axis activation during the regulation of negative emotion. Most importantly, however, Fox and colleagues (1996) findings indicate that, although right frontal asymmetry is more commonly seen in socially inhibited children, such a pattern of activation predicts future behavioral problems for both socially inhibited and uninhibited children. These findings underscore the role of right frontal activation in the regulation, processing, and expression of negative emotions.