Evolved Architecture of Emotion and Cognition

An important insight is that the vertebrate brain has evolved largely through elaborating new structures on top of old ones, a process that the 19th century evolutionary biologists described as one of terminal additions. Very simple ancestors organized behavior through the specialized end of the neural tube (spinal cord) termed the hindbrain (rhombencephalon). From this basis the midbrain (mesencephalon) evolved with more complex integrative neural networks. Yet, these did not replace the hindbrain, but rather modulated its function, largely through inhibitory control. Almost certainly, the earlier structures have not remained in their original form, but have mutated in ways that are complementary to the more recently evolved elaborations at the differentiating end of the neural tube.

From the midbrain evolved two structures of the interbrain (diencephalon) that have been pivotal as organizing influences in the brain systems of more complex vertebrates. The first is the hypothalamus, a structure above the midbrain controlling homeostasis. Homeostasis is the visceral regulation of internal states (temperature, thirst, hunger, pain) required for the life of the organism. The second diencephalic structure is the thalamus, a set of nuclei on both left and right sides of the brain at the top of the brainstem. Within each thalamus are nuclei for both sensory and motor interfaces of the organism with the outside world.

On top of the interbrain is the final major division of the vertebrate brain, the endbrain, or telencephalon. The forms taken by the endbrain determine the intelligence capacity of the higher vertebrates, including reptiles, birds, and mammals. The problem is understanding the function of the basic circuits of the telencephalon, and the diverse variations they have taken in each species.

In this effort to understand a highly evolved organ system, the general rule of 19th century evolutionary biologists was to look to its more primitive roots. These roots are found both in phylogenesis, that is, evolutionary roots, and in ontogenesis, that is, embryonic roots. In the stages of the development of the human embryo, for example, we can trace the elaboration of the terminal additions in the major phyletic transformations of the brain.

For the specific problem of mapping out the telencephalon of vertebrates, C. Judson Herrick took the classical approach, and studied how the telencephalon elaborates the function of the diencephalon (interbrain), including both thalamic and hypothalamic nuclei (Herrick, 1948). This approach remains instructive for neural systems theorists today.

The thalamus is the gateway for sensory and motor traffic into and out of the telencephalon. In higher mammals including humans, the cerebral cortex of the telencephalon has evolved massive networks for elaborating sensory and motor processes, yet they continue to funnel communications with the outside world through the thalamus. In this pivotal role, the thalamus exerts essential regulatory controls on the inhibition, activation, and thus selection of information within the sensory and motor streams of experience and behavior. Corticothalamic processes are therefore essential to conscious attention and the executive control of sensorimotor function.

Whereas the sensory and motor processes are integral to the body's somatic function, interfacing with the outside world, there are equally important neural systems for regulating the visceral function, interfacing with the homeostatic mechanisms of internal bodily processes. These are organized at the diencephalic level in the hypothalamus, often described as the body's master control center. Following Herrick's guide (1948), if we look to the projections from the hypothalamus to the telencephalon, we find these are restricted to the limbic cortex, at the inner core (i.e., limbus or border surrounding the brainstem) of each cerebral hemisphere. With this relation to visceral bodily processes, the limbic cortex has evolved a general evaluative function, determining the biological or internal significance of the information traffic with the world.

A general model of emotion and cognition can be derived from these elementary bases of neural organization. At the core of each cerebral hemisphere (the left and right halves of the endbrain) are the internal regulatory systems. These provide global evaluative, that is, emotional, controls. On the lateral surface are the sensory and motor networks. These provide the articulation of the specific patterns of perception or action that must be coordinated tightly with the realities of the immediate environment. Each of these primitive functions appears to have evolved with unique representational capacities, that is, unique ways of forming ideas or concepts. At the limbic core there emerges a general representation of the significance of an event, often in relation to a bodily need. To organize a discrete action, or a particular perception, there must be a process of specification. This specification is achieved largely through inhibitory control of the elements of the percept or action, such as through the inhibition and selection operations performed by thalamic circuits. In the motor domain, a discrete action is therefore sculpted from a general motive urge and its various potentialities. In the sensory domain, a distinct percept is organized from the interface of motive roots with thalamic attentional direction and the specific data in sensory neocortex.

Thus, two bodily constraints (visceral at the inner core and somatic at the outer shell) form the boundaries of the mind. Emotion is organized close to the core. Not only sensorimotor specifics but explicit cognition is organized close to the somatic shell. Both of these boundaries are concrete, embodied in their unique biological constraints. What seems to happen is that abstract cognitive abilities, such as empathy, arise through mediating between these opposing constraints (Tucker, 2001).

A System for Pain Evaluation

The pain of frustrated love has, of course, been a palpable feature of experience for many people. What is important for neuropsychological theory is to recognize that social behavior is motivated by viscerally significant motives, such as pain and longing, that are as integral to mammalian life as hunger or thirst. With this recognition, we can examine the neural substrates of the drives and motivations that form the foundation of affiliation. As we will see, the use of pain to describe psychological distress is not merely a semantic convenience, but rather an accurate reflection of a psychological experience that is, at its core, viscerally controlled.

Pain is not a unitary phenomenon. Psychometrically, pain ratings are multidimensional, being made up of sensory and emotional components. The distinction between localization and evaluation is a classic one (Rivers & Head, 1908), and has found its way into modern concepts of somatosensation. The system involved in conveying nociceptive information to the brain is known as the anterolateral system. This system contains several ascending tracts. One begins in the spinal cord where fibers from the dorsal horn crosses over into the contralateral space and travels up into the intralaminar and ventral posterior lateral (VPL) nuclei of the thalamus (Albe–Fessard, Berkley, Kruger, Ralston, & Willis, 1985). The spinal projections to the VPL of the anterolateral tract terminates in a topographic manner. The receptive fields of nociceptive neurons in this route are restricted. In contrast, the projections from the spinal cord to the centrolateral and parafascicular nuclei of the thalamus via the anterolateral tract are not topographically organized. Neurons in these thalamic nuclei have diffuse receptive fields, often covering an entire body, and their response latencies to the nociceptive stimuli are long, and once activated, their responses are prolonged. These characteristics make it clear that cells within this pathway are poor at spatial and temporal localization. Albe–Fessard et al. argued that the pathway to the VPL is involved in the localization of noxious stimuli and the pathway to the intralaminar nuclei is involved with the affective evaluation of the stimuli.

This localization/evaluation division of the pain pathway has been extended to the cortex (Vogt, Sikes, & Vogt, 1993). The VPL projects to the somatosensory cortex, and Vogt et al. reviewed evidence showing that cells within the somatosensory cortex do respond to nociceptive stimuli by coding information about the intensity, duration, and location of the noxious stimuli. The centrolateral and parafascicular nuclei have been shown to send projections to the anterior cingulate cortex (ACC) as well as the orbitofrontal cortex (Jones & Leavitt, 1974; Vogt & Pandya, 1987). Because of the projections from the centrolateral and parafascicular nuclei, cells in the ACC have been found to respond to nociceptive stimuli (Vogt et al., 1993), but their receptive fields are large and their responses outlast the duration of the stimulus. Thus, the ACC, like the intralaminar nuclei, appears not to be involved in localization and identification of the noxious stimulus, but rather it is involved in the subjective, evaluative aspects of noxious stimuli as we will see.

Pain, Learning, and Attention

The contributions of the anterior limbic system (ACC and anterior insula) to pain evaluation have been studied in isolation from the apparent contributions of these networks to the self-regulation of effort, attention, and discrepancy detection. Yet there are important reasons to believe that the evaluative mechanisms engaged in many complex forms of self-regulation are extensions of mechanisms that evolved for evaluating and responding to pain.

As might be expected, learning in response to aversive stimuli appears to be mediated by the neural systems that respond to pain (Papini, 2002). However, these same systems have been found to play more complex roles in mediating learning in situations such as frustrative nonreward (Amsel & Stanton, 1980). Traditional behaviorist models of associative learning cannot account for many well-known phenomena of learning, including, for example, the improved strength of learning due to variable reinforcement. Animals that are rewarded on an irregular schedule are more resistant to extinction than those on a consistent reward schedule. Although this cannot be explained by passive associative conditioning principles, Amsel and Stanton (1980) have shown how such phenomena can be explained by mechanisms of frustration learning. The frustration engendered by nonreward typically leads to an aversive response to the situation, and to a loss of the hedonic tone that is critical to reward expectancy and continued behavioral effort (Amsel & Stanton, 1980). These responses appear to reflect self-regulatory responses, similar to anxiety and frustration in humans, that evolved from the limbic mechanisms of pain evaluation (Papini, 2002).

It is important that, under the conditions of a variable reward, the frustration becomes “counterconditioned” by the occasional reward, such that the animal develops a tolerance for the frustration in that learning context. When reward is withheld, the animal trained on a variable reward schedule can then tolerate the frustration and has much greater capacity for continued behavioral effort than the one trained with constant rewards. These elementary motive and self-regulatory processes may be important in interpreting the cognitive neuroscience evidence on anterior cingulate and insula contributions to the emotional evaluation of events, monitoring of conflict, and engagement of effortful self-control (Luu & Tucker, 2003).

Encephalization of Pain Evaluation

In mammals, and perhaps in dinosaurs and birds, there are primitive motivational systems, such as those regulating the response to pain, that appear to have been extended through natural selection to support and motivate the capacity for social attachment. Modern analysis of neural mechanisms of attachment behavior in juvenile mammals, such as by Panksepp, suggests that there is a pain of social abandonment that motivates attachment, and that this is mediated by similar mechanisms that are engaged in physical pain (Panksepp, 2003a, 2003b).

Encephalization is a concept that is complementary to the construct of terminal additions. It describes how functions of more primitive neural structures were elaborated with more recently evolved terminal additions. In particular, the elaboration emphasizes the shift away from rigid and reflexive actions to flexible and temporally delayed behaviors. Thus, encephalization of the pain system can be regarded as an evolutionary trend that parallels the development of more complex social and cognitive abilities. This means that reactions to pain are no longer reflexive, immediately providing the motivation to move the organism out of harm's way. In its most basic form, the pain system can be regarded as another homeostatic system; a system that regulates adjustments to restore a balance. As homeostatic mechanisms evolved and became encephalized to form allostatic systems, allowing for anticipations of homeostatic imbalance and making adjustments in advance (McEwen, 2000) so has the pain system. In its encephalized form, the pain system forms the foundation for learning, allowing for future actions to not only avoid physical harm but also to prevent loss of reward. That is, it allows for the regulation of actions in context (Luu & Tucker, 2003). Attachment can be regarded as an extension of this motive function.

An interesting observation was made by Foltz and White (1962) on cingulotomy for the treatment of pain. They described that certain patients with chronic pain showed “augmented” emotional responses. These included displays of anxiety, preoccupation with the pain to the point of causing them depression, and emotional lability. These patients showed immediate changes in demeanor when the ACC was lesioned.

Perhaps consistent with these findings, in modern neuroimaging studies the ACC has been repeatedly shown to be involved in the subjective evaluation of the unpleasantness (i.e., distress) of physical pain (e.g., Derbyshire et al., 1994; Rainville, Duncan, Price, Carrier, & Bushnell, 1997). We have argued that this basic function forms the motivational basis for ACC monitoring functions. When expectancies (i.e., how actions relate to the world) are violated, such as when errors are committed or expected rewards are not delivered the evaluative pain system is engaged to motivate learning of new behaviors (Luu & Tucker, 2003; Papini, 2003). Indeed, the ACC, along with the amygdala, comprises part of a system that regulates rapid learning of avoidance behaviors (Gabriel, Burhans, Talk, & Scalf, 2002). This is the first consequence of encephalization of the pain system; a subjective evaluation of pain that is removed from the evaluation of the physical properties of pain and produces a process of frustration that can motivate new learning.

Vertical Integration of Emotional Vocalizations

Another manifestation of the encephalization of nociceptive mechanisms is directly related to attachment. This is the vocalization associated with experienced pain. Vocalization and pain overlap closely in their cerebral representation. Sites that are able to elicit vocalization, such as the cranial nuclei, periaqueductal gray matter (PAG), mediodorsal amygdala (MD), and ACC (Jürgens & Ploog, 1970), are also involved in nociception and rich in opiate receptors (Jones et al., 1991; Wamsley, Zarbin, Young, & Kuhar, 1982). However, it must be more than just the state of physical pain that is related to vocalization; it must be the state of psychological pain as well. Closely related to social pain and abandonment are distress vocalizations. Young birds or mammals separated from the mother will often emit distress vocalizations. It is the role of the ACC in pain evaluation and vocalization that exemplifies how the encephalization of the pain system relates to attachment, sympathy, and the development of empathy.

The study of emotional vocalizations has been important in understanding the integrated roles of multiple levels of the evolved neuraxis in emotional experience and behavior. Unlike human speech, which can be separated into verbal and nonverbal (affective) components, primate vocalization is inevitably linked to affect. This fact is important in that it emphasizes the motivational component of speech and behavior. Studies of primate vocalization by Jürgens and his colleagues (Jürgens, 1979; Jürgens & Ploog, 1970; Jürgens & Pratt, 1979) have shown that it is possible to find sites throughout the extent of the neuraxis that are involved in vocalization. In these studies it has been shown that vocalizations can be elicited from cranial nuclei in the brainstem, such as the solitary nucleus, trigeminal motor nucleus, and the reticular formation. Vocalization can also be evoked from the PAG, mediodorsal nucleus of the thalamus, the amygdala, and ACC.

Although the entire extent of the neuroaxis is involved in emotional vocalization, experimental evidence indicates that sites at each level of the neural hierarchy play different roles. Jürgens and Ploog conducted a landmark series of studies of the brain systems controlling emotional vocalization in mammals (Jürgens & Ploog, 1970; Ploog, 1992). Jürgens and Ploog were able to characterize the increasing complexity in the organization of vocalization at each higher level of the brain. Whereas discrete actions, such as respiratory and phonetic aspects of vocalization, resulted from electrical stimulation of the brainstem, more complex species-specific patterns of vocalization were elicited from stimulation of the midbrain. At the end-brain level, differing roles were found for the limbic cortex (core) and neocortex (shell). The limbic cortex was found to be essential to the appearance of vocalization in an appropriate emotional context, as if it provides a spontaneous emotional response that then subordinates the lower brain centers of patterning and discrete actions. In studies with monkeys, the neocortical networks (i.e., ACC), at the highest level of the telencephalon, were found to be required for what Jürgens and Ploog called “voluntary call initiation.” In these calls, the monkey used vocalization in what appeared to be a deliberate, instrumental communication with another (which is not involuntarily yoked to motivations).

The organization of the cingulate vocalization region is particularly important to the organization of the voluntary aspects of vocalization (Vogt & Barbas, 1986). The cingulate vocalization region is heterogenous, involving rostral area 24, 32, and 25. These regions are interconnected and receive projections from the amygdala. The regions have connections with dorsal aspects of area 24. Consistent with the evolutionary trend in which areas 25, 32, and rostral 24 give rise to dorsal area 24 (Pandya & Yeterian, 1990), dorsal area 24 has been shown to exert inhibitory control over the cingulate vocalization region, providing contextual appropriateness to calls (Vogt & Barbas, 1986).

The evolution of neural mechanisms that integrate pain/distress motive with vocalization lays the foundation for understanding the vertical integration of attachment and regulation of parent/infant bonds that can then be extended to more general social bonds. Early in evolution, pain vocalization is just a reflex to experienced pain, as seen by the coupling of pain and vocalization in the PAG. Most social animals have the capacity for emotional response to the emotional displays of other of their species, described as emotional contagion. The mechanisms of emotional contagion seem to share neural substrates common to the pain system. Emotional contagion may provide the foundation for learning of the association of another's pain-related vocalization to the concept of pain or danger. With increasing encephalization of the nociceptive function, brain mechanisms for support of maternal and parenting behaviors came online, allowing for the recognition of pain vocalizations to be understood as expressions that signal another's pain.

Evolved Mechanisms of Attachment

As recognized by Paul MacLean's classical studies, mammals gave rise to new classes of behavior, including parenting, attachment, and play, that support dependent juveniles, and the extended embryonic plasticity of their brains (MacLean, 1987). For evolutionary theory, of course, we must try to understand the more primordial changes that anticipated these mammalian capacities. An example might be the dinosaurs' ability to care for their young, an ability that might still be seen in rudimentary form in today's crocodilians. Although perhaps not as fully elaborated and flexible as the attachment and parenting of mammals, the attachment mechanisms of birds continue to provide us with important instruction on the progressive nature of developmental experience. The young bird imprints on the significant caregiver of its infancy. Regardless of whether this is its mother or a strange alien of another species (like an ethologist), the bird then maintains an attachment with that object throughout life.

Modern analysis of neural mechanisms of attachment behavior in juvenile mammals suggests that there is a pain of social abandonment that motivates attachment and this is mediated by similar mechanisms that are engaged in physical pain (Panksepp, 2003a, 2003b). Young birds or mammals separated from the mother emit distress vocalizations. Small doses of opiates, that is, doses that are too small to be sedating, stop the distress vocalizations, and stop the animal's searching for its mother. The implication is that contact with the mother must be associated with an endogenous (natural) opiate release. Abandonment then results in a kind of withdrawal syndrome, one that may be directly analogous to heroin or morphine withdrawal. Thus, when an opiate is administered to the young animal, the drug is fully effective in relieving the pain of abandonment, suggesting that opiates and their deprivation are the natural mechanisms of social bonding. Panksepp points out that human social bonds may share important features with opiate addiction.

In many cases, heroin users turn to drug use in response to social or sexual abuse, or to interpersonal abandonment. Whatever leads to opiate abuse in an individual case, the effects of addiction are remarkably consistent. The abuser loses interests in social bonds, and substitutes heroin for both sexual and family relations. For example, Haertzen and Hooks (1969) conducted a study in which prisoners were given heroin or morphine for 60 days and then brought into withdrawal. During the acute stage, the subjects described themselves as having physically pleasant feelings, being excited and less discouraged, and being less worried and more patient and accepting. However, in the chronic state the subjects reported less motivation for a variety of social and sexual activities. Social activities were considered by subjects to be a waste of time. They preferred individual pursuits over social events and expressed boredom with other people.

Thus, the evidence from opiate abuse provides convergent support for Panksepp's (2003a, 2003b) hypothesis that the body's natural opiates are essential mechanisms in the normal process of social comfort and bonding. This hypothesis may also help to explain the otherwise anomalous behavior of self-injury. Some emotionally disturbed people, particularly women who have suffered sexual or emotional abuse, may intentionally injure themselves, such as by lacerating an arm or leg repeatedly with a knife. Most theories of motivation begin with the assumption that people seek pleasure and avoid pain, and thus cannot explain why someone would intentionally hurt themselves. Yet, if the pain of severe interpersonal abuse is mediated by opiate mechanisms, then it may be that compensatory opiate release, which is caused by physical pain, is experienced as a relief, a relief to a chronic state of psychological pain. Consistent with this reasoning, Roth, Ostroff, and Hoffman (1996) observed that after administration of naloxone, an opiate antagonist, patients with self-injurious behaviors reduced or ceased their self-injurious behaviors.

Up to this point, we have reviewed what we believe evidence of the necessary neural system for the development of empathic behavior. In the next sections we outline the psychological components of empathic behaviors, neural correlates of empathic behaviors, and the role of socialization in the development of empathic behaviors.

Psychological Components of Empathic Behavior

Although attachment may be an essential element to a child's development of empathic behavior, it must be accompanied by cognitive skills that allow the child to understand the perspective of the other person (Taylor, 1999). These skills have been studied in recent years in relation to the concept of “theory of mind,” the notion that the child can form a mental proposition that another person has a mind, and thus a unique subjective perspective on events (Moses, Baldwin, Rosicky, & Tidball, 2001). As a cognitive skill, perspective taking can be studied with experimental methods, allowing an analysis of its emergence in child development (Hoffman, 1984). The capacity for perspective taking has been seen as necessary for empathy, understanding another's feelings, as opposed to sympathy, having one's own feelings in response to observing another's emotional displays or emotionally significant situation.

Although the cognitive process of perspective taking is accepted as a necessary component of the psychology of empathy, modern emotion theorists have recognized that a sympathetic response may also be essential. It is possible, for example, to have and use a clear understanding of another's subjective perspective to cause harm to that person, such as in physical or psychological torture. Caring for the person's welfare is also required for prosocial empathic behavior. The interesting psychological question is how sympathy contributes to this caring. It may be that the addition of sympathy, in the primitive sense of emotional contagion, to perspective taking is sufficient to constitute empathy. However, actually caring for the other may also be required, and the basis for caring proves as important in a biological analysis as it is in a psychological one.

Cognitive Neuroscience Models of Empathic Behavior

When a monkey observes a person in the laboratory picking up an object, this observation is accompanied by activity in motor neocortex in the same (inferior prefrontal) motor neurons that the monkey uses to pick up an object (Rizzolatti & Fadiga, 1998). Intracranial recordings in humans show that there are ACC cells that are both responsive to painful stimulus applied to the subject as well as observation of the painful stimulus applied to others (Hutchison, Davis, Lozano, Tasker, & Dostrovsky, 1999). These cells are not just stimulus responsive; rather, they seem to respond to the anticipation of pain, consistent with the encephalized nature of the ACC pain response.

At more abstract levels of perspective taking, several studies have implicated the involvement of the ACC. Error monitoring has been an active domain of research within the past decade, and repeated studies have shown the involvement of the ACC in this critical cognitive function (Dehaene, Posner, & Tucker, 1994; Holroyd, Dien, & Coles, 1998; Luu, Collins, & Tucker, 2000). A recent study by van Schie, Mars, Coles, and Bekkering (2004) show that the brain response to observing errors committed by another individual is similar to self-committed errors. In other tasks that require subjects to attribute intentions to the behaviors of others, or that require subjects to infer the affective and motivational states of others, activation is observed in the same ACC regions (primarily the rostral ACC) as those tasks that require subjects to monitor their own actions (Frith & Frith, 2001).

The neural substrate of perspective taking involves multiple brain structures, of course. The reason for the emphasis of ACC in our review is to demonstrate the consistency between the pain system and its role in the support of empathic abilities. Nevertheless, because of the involvement of multiple cerebral structures in perspective taking and considering the associative connections that could be developed among the different structures, Keysers and Perrett (2004) proposed that learning may lead to the capacity for representing the actions of others simply through a process of Hebbian association.

Studies in humans have suggested that multiple components of empathy developed from a psychological analysis may be related to activity in specific neural networks. Drawing from both psychoanalytic and developmental psychological analyses of empathy, Decety and Jackson (2004) have proposed that empathy requires three primary components: (a) a sharing of another's affect, (b) the maintenance of a separate self-representation, and (c) flexible mechanisms of emotional self-regulation to allow engagement of the differential perspectives of self and other. In reviewing neuroimaging findings related to empathic processing, Decety and Jackson cite evidence not only of brain activations that appear to mirror the experience of the other, but midline frontal activity (posterior cingulate, anterior cingulate, and frontal pole) that appears to reflect self-regulatory mechanisms integral to the shifting of self-other perspectives. These corticolimbic networks overlap in an important way with the attentional control networks of the brain (Posner & Rothbart, 1991).

Developing Empathy

The capacity to integrate conscious attention with emotional experience may be fundamental to empathic experience. Although we can now analyze the neural networks and circuits that are required, these neural mechanisms do not appear to become functional without the appropriate developmental experience. Early experiences seem to be particularly critical, forming a template for later cognition and behavior. This analysis of early formative mechanisms may be important to understand both normal and psychopathological development.

Just as birds form an enduring bond with their caregivers through experience at a critical stage of imprinting, infant primates enter life prepared to bond with the mother. With Rhesus monkeys, systematic research by Harlow (1986) has shown the deficits in both social adaptation and physical health when the infant's bond with the mother is blocked. Importantly, an effective relationship requires both the infant and mother to respond appropriately. Maternal deprivation causes a female infant to grow up with deficits not only in peer interactions but in her parenting skills as well. In recent research, Suomi has extended this line of work in important ways. For example, Suomi and associates have studied young monkeys with genetic temperaments predisposing them to anxiety or impulsivity (Barr et al., 2004; Suomi, 2003). It is important that these monkeys may not grow up to suffer from behavior disorders if their mothers or other caregivers are able to provide them with adequate love and social support.

These studies of Rhesus monkeys have been extremely important in providing theoretical models for understanding the interaction between physiology and socialization in humans. In human disorders of empathy, a similar interaction between genetic temperament and social support can be observed. In autism, for example, the infant appears to be born with some deficit of the mechanisms required for successful attachment and social interactions. Recent experiences suggest that highly involved and competent parenting can overcome the social deficits of the autistic child to a considerable degree. Nonetheless, in less fortunate cases the interactive effect of the child's deficits and the response of caregivers to this deficit often produces a negative spiral, in which the child's disinterest in social contact leads to the parent's disinterest and neglect.

Whether from genetic or social causes, autistic children show a profound inability to understand the perspectives and intentions of other people. This inability is striking when compared with other cognitive skills, which may be relatively intact. Although it may be influenced by, or even caused by, the autistic child's lack of emotional or sympathetic response to another's emotions, the deficit in understanding another's perspective is clearly a cognitive impairment. Lacking what has been called a “theory of mind,” the autistic person simply cannot reason in ways that explain the other's perspective.

Careful studies of the normal communication patterns shared by infants and mothers have supported the realization that social perspective taking is a fundamental component of intelligence. Trevarthen and associates have observed the typical interactions between mothers and infants, and has provided a systematic analysis of patterns of reciprocity, turn taking, and other causal relations of the dialog (Trevarthen & Aitken, 1995). These patterns suggest that both infant and mother are able to understand not just the signals of the communication, but the intentions of the other that must be held in mind to participate in such well-orchestrated behavior.

A consistent finding has come from studies of how toddlers learn the meanings of words (Baldwin, 1993; Moses et al., 2001). They do not learn if there is a simple paring of the spoken word with the presence of an object, as traditional behaviorist theory would imply. Rather, the child looks to see what the mother is attending to, or what she intends to do, as they learn the words that go with the objects or actions. Thus, the meaning of words is not an arbitrary association, but is embedded in the child's understanding of the mother's intentions.

These studies have shown that there is a critical mental capacity required for effective social communication, one of understanding the intention and perspective of the other, and of communicating one's own state of mind in a reciprocal fashion. In his study of infant–mother interactions, Stern has described the shared mental representations that support such interactions as “intersubjectivity” (Stern, 1985). The important point for an understanding of empathy is that the skills of intersubjective representation are cognitive skills, and they are essential for effective human intelligence.

However, these skills are not enough to create effective empathic capacity. This is illustrated by another human developmental disorder, psychopathy. The psychopath may be quite skilled in understanding another person's perspective, and in fact, may be remarkable in the ability to manipulate others through application of these skills. Yet, there is a surprising lack of concern for the welfare of other people, such that the psychopath easily causes pain or injury to others with no apparent remorse.

Given its societal importance, psychopathy has received remarkably limited scientific attention. As a result, we have only limited evidence on the interplay of genetic temperament and socialization in the crystallization of this highly dangerous personality disorder. However, family studies have suggested some degree of heritability of psychopathy and related conditions, including histrionic personality disorder in women and drug abuse in men (Guze, Woodruff, & Clayton, 1971). Furthermore, anecdotal clinical observations have suggested that in many severe cases of psychopathy there is evidence of childhood neglect or abuse.

Drawing from the model of interacting genetic and social factors from the monkey studies, we can hypothesize that psychopaths share a degree of temperamental vulnerability, but their developmental outcomes depend heavily on the quality of their parenting and socialization. Whatever the developmental course of the disorder, the empathic deficits of the psychopath are caused by fundamentally different psychological mechanisms than those of the autist. With good interpersonal perspective taking and inferencing, the psychopath remains deficient in sympathy for fate of other people, showing what may be described as “emotional autism.”

Neural Mechanisms of Sympathic Resonance and Intersubjective Reasoning

By understanding both cognitive and emotional foundations of empathy, and their underlying neural mechanisms, we can formulate a biological model for how empathy evolved in its particular form in humans.

A key element is sensitivity to the emotional state of the other person. We can describe this as sympathic resonance: the subject's own emotional responses resonate to the emotional display of the other. In a primitive form, the subject's responses are simply emotional contagion, but as we will see the human cortex provides a representational capacity such that, when balanced against articulation within somatic networks, the visceral emotional response may itself become more complex and sophisticated. Whether primitive or more complex, the sympathic resonance in limbic core-brain networks is syncretic, meaning that there is holistic fusion of multiple experiential elements, including both perceptual and interoceptive (from visceral sensations; Derryberry & Tucker, 1991). Because many occasions calling for empathic concern involve pain experienced by the other person, we can infer that the child's sympathetic response engages his or her own pain evaluative networks.

People differ in the capacity for sympathic resonance, and they therefore differ in an important component of empathy. Autistic, obsessive–compulsive, and paranoid persons are deficient in their own affective responses and therefore in the capacity to use their sympathic resonance to understand the emotions of another. Just as a person's own visceral, affective responses are emergent from homeostatic mechanisms and largely spontaneous, so, too, is sympathic resonance a largely spontaneous reaction to the experience of another's emotional display or situation. We propose that emotional contagion is only developed into a more complex representation when the sympathic response is mediated by intersubjective reasoning.

Intersubjective reasoning refers to the cognitive process through which the subject is able to infer the perspective, intention, and subjective state of another person. From the studies of mother–infant interaction and of toddler social referencing, we can see the complex set of skills that a young child gains in inferring the mother's subjective state.

In considering the development of intersubjective skills, it may be important to recognize the error in the common philosophical assertion that we can never know the subjective state of another person. It is not just that children are able to gain considerable skill in intersubjective reasoning, because this skill does remain limited. The error is the naive philosophical assumption that we have full and pristine knowledge of our own subjective experience. There is extensive psychological evidence that when asked to report on their subjective experience, people have nothing like a veridical report of their mental processes. Rather, they must confabulate an answer that seems plausible, relying on inferences and reasoning not unlike that used by the child in understanding another's subjective perspective. Intersubjective interpretation is highly imperfect, but so is the understanding of our own subjective experience, and this inferencing limitation is essential to understanding human empathy (Decety & Jackson, 2004).

Just as understanding the unique representational skills of the limbic networks helps to explain the visceral mechanisms of sympathic resonance, an analysis of the articulation and differentiation of information within the somatic neocortical networks helps to explain the fundamental basis of intersubjective reasoning. Within the somatic sensorimotor networks, the representation of another's actions appears to allow a kind of analysis by synthesis, through which the person models what it would be like to carry out the other's actions. The simulation of actions in mirror neurons described above may be the somatic analogue to the visceral response of emotional contagion. There is a matching or synchronization of the internal response with the observed actions.

In both cases, the subject's internal mirroring of the other provides a basis for perspective taking, for interpreting what it must be like to be the other. The differing natures of the conceptual representations at visceral and somatic levels provide differing frameworks for interpretation. At the visceral level, sympathic resonance gives a fast and syncretic affective context for what might be the other's subjective state. At the somatic level, action mirroring provides an equally concrete representation of what it must be like to carry out those actions. However, these concepts have differing neural mechanisms, with differing representational properties. Whereas the visceral representation is spontaneous and dynamic, the somatic representation is more differentiated, articulated, and sequenced. Because of its more structured form, the somatic representation provides scaffolding for more conscious and deliberate reasoning about the actions of the other, leading to the more complex capacity for intersubjective reasoning.

The importance of more deliberate reasoning is implied by Trevarthen's observation of the limited utility of pure synchrony in the most primitive and elemental interactions of the infant with the mother. Rather than synchrony, communication requires coordination. This means alternating turns in signalling, emphasizing a response to a familiar gesture, or making a bid to create a new shared event. Through such active and deliberate engagement of their knowledge of their own and the other's mental processes, children gain important skills in intersubjective reasoning.

Empathy Through Abstract Reasoning

The hypothesis of reentrant organization of concepts across visceral and somatic networks can be applied to the psychological experience of empathy. The core-shell hypothesis of corticolimbic interaction proposes that abstract concepts are the result of a structural process. This is a structural process that parallels the structuralist account of psychological development by Werner (1957). In this process, the syncretic quality of densely interconnected limbic networks (with their visceral, hypothalamic controls) is opposed to the differentiation of specific features of the sensorimotor interface with the environment in neocortical networks (with their somatic inhibitory thalamic controls). At the core, the syncretic concepts rapidly grasp the significance of visceral responses and demands. They are well suited to the task of sympathic resonance. At the somatic shell, the differentiated concepts allow specification of representations in relation to the actual referents in the world, both sensory and motor. This is a quality of information representation that is well suited to both sensory and motor mirroring.

Both of these bodily frames of experience are concrete boundaries of mental representation. However, through the extended consolidation between them, that is, through their dialectical opposition in structural forms, there seems to emerge a new, interlinked representational form. This form achieves an abstract relation between the concept and the referent. The differentiated referents at the shell are assembled again in a syncretic integration at the core, but now the differentiated referents remain.

For example, in an initial syncretic visceral response, a child may have sensed anxiety or urgency in a person's nonverbal cues, and responded with a diffuse sympathic resonance of her own, relying in part on her own capacity for experiencing anxiety and pain. She also watched the person picking up food, and mirrored this action with her own sensory motor perspective, through analysis by synthesis, on what it would be like to carry out exactly that set of differentiated actions. Through reentrant interaction between these representational forms, created at opposite boundaries of her corticolimbic networks, she forms a more abstract concept, now differentiated and integrated, in which she understands how the person's motivated visceral state must go with the sensorimotor actions. The person must be hungry.

Empathic Concern Through Nociceptive Capacity

Thus, empathy may require sophisticated operations of intelligence that arbitrate between visceral emotional contagion and somatic sensorimotor mirroring to create more abstract representations. These representations would elaborate more complex forms of sympathic resonance at the visceral core of experience, in which the child comes to integrate elements of the feeling of the other, but now supported by more deliberate processes of reasoning. At the same time, the abstract representations support intersubjective reasoning, in which more deliberate propositions about the other's intentions and perspective are organized within an explanatory context of felt meaning.

Even though it is a remarkable product of evolution, this highly organized intersubjective intelligence is not guaranteed to produce an effective concern for the welfare of the other person. Even powerful and abstract forms of intelligence may remain bound by egocentric concerns. To go beyond a theory of intelligent empathy to explain effective prosocial experience and behavior, it may be necessary to understand the residuals of childhood experiences of attachment and bonding that were integral to the exercise of empathic capacities. For the child whose elemental needs for social contact were met effectively, it is very likely that interpersonal communication has been grounded in concern for others. The exercise of empathy in both visceral and somatic domains is founded on the experience of love, both in giving and receiving. It is not necessary that the experienced relationships were entirely gratifying; having experienced frustration and loss may also give the child an important personal basis for interpreting the experiences of another.

We propose that research on limbic evaluative processes, evolved from primitive nociceptive mechanisms, may be integral to understanding both interpersonal attachment and the capacity for empathic concern that develops from it. The opiate mechanisms that mediate the child's love relationships, and the psychological pain that results from their frustration, become integral motive forces in directing the reentrant corticolimbic process of forming abstract concepts of empathic understanding. Limbic evaluative processes are not simply raw emotions, but may act as motives for extended concept formation and memory integration, including the reentrant processing that we have described as organizing the hierarchic structure of abstract concepts. The person with the capacity for tolerating the frustration of experiencing sympathetic pain may achieve the extended corticolimbic processing required for abstract empathic understanding.

The research on frustration and learning may suggest how the capacity for tolerating sympathetic pain develops to allow effortful empathic reasoning. From a stable attachment relationship, the child develops self-regulatory operations that are not only associated with opiate release during reunion and affection, but extended intervals of psychological pain when the love relationship is frustrated. Because the bond is strongly regulated by elaborated (encephalized) nociceptive mechanisms, the evidence and theory on frustration and learning may be explanatory for this developmental process. Children whose early relationships do not provide for strong bonding experiences may not have the self-regulatory experience to tolerate the frustration of sympathic pain in later relationships. In fact, the learning theory principles suggest that children whose needs are continually gratified may not be adequately frustrated to allow the counterconditioning of frustration within the successful bonding relationship. It may be, then, that it is the adaptive negotiation between parental and child needs, within a strong relationship, that allows the experience with frustration and reunion that is optimal for the capacity to experience psychological pain in a way that it is productive for deep, reflective empathic experience and extended coping efforts.