Abstract

There are two broad themes in psychosomatic medicine research that relate emotions to physical disease outcomes. Theme 1 holds that self-reported negative affect has deleterious effects and self-reported positive affect has salubrious effects on health. Theme 2 holds that interference with the experience or expression of negative affect has adverse health consequences. From the perspective of self-report these two traditions appear contradictory. A key thesis of this paper is that the foundational distinction in cognitive neuroscience between explicit (conscious) and implicit (unconscious) processes, corresponding to Themes 1 and 2, respectively, provides a unifying framework that makes empirical research on unconscious emotional processes more tractable. A psychological model called "levels of emotional awareness" is presented first that places implicit and explicit emotional processes on a cognitive-developmental continuum. This model holds that the ability to become consciously aware of one's own feelings is a cognitive skill that goes through a developmental process similar to that which Piaget described for other cognitive functions. Empirical findings using the Levels of Emotional Awareness Scale are presented. A parallel hierarchical model of the neural substrates of emotional awareness is presented next supported by recent neuroimaging and lesion work. The evidence presented in this review suggests that the neural substrates of implicit and explicit emotional processes are distinct, that the latter have a modulatory effect on the former, and that at the neural level Theme 1 and Theme 2 phenomena share critical similarities. The implications of this psychobiological model for research in psychosomatic medicine are discussed. Key words: neuroscience, health, implicit processes, emotional awareness, emotion regulation. ACC ϭ anterior cingulate cortex; BA ϭ Brodmann Area; dACC ϭ dorsal anterior cingulate cortex; HF ϭ high frequency; HRV ϭ heart rate variability; IAPS ϭ International Affective Picture System; IBS ϭ irritable bowel syndrome; LEAS ϭ Levels of Emotional Awareness Scale; PET ϭ positron emission tomography; rCBF ϭ regional cerebral blood flow. INTRODUCTION T he physiology of emotion is arguably the cornerstone of psychosomatic medicine. The American Psychosomatic Society was launched by the publication of Emotions and Bodily Changes by H. Flanders Dunbar (1). Conferences were held on this same topic in 1936 and 1937 and the journal Psychosomatic Medicine was inaugurated in 1939 with Dr. Dunbar as Editor. The journal was to be "devoted not to the isolated problems of the diseased mind or the diseased body, but to the interrelationships between emotional life and bodily processes" (2). To this day, a majority of papers in this journal involve emotion and emotion regulation, broadly construed. Yet, although emotion and its physiology have been foundational for the field, the approaches taken have not been unified. There have been two distinct themes in how it has been approached over the past 70 years. The first theme, which is the dominant perspective within the field today, holds that aversive emotional states are associated with adverse health outcomes. There is now a long list of negative emotional states and traits for which this has been found to be true, including depression (3), hostility (4), worry (5), anxiety (6), hopelessness (7), and perceived stress (8), among others. According to this perspective, self-reports are assumed to provide a satisfactory way of identifying these states and traits. We now have unequivocal evidence, for example, that self-reported depression is associated with decreased survival in patients with coronary artery disease (9). Depression is also associated with increased mortality in the context of diabetes (10) and certain forms of cancer The second theme, which predominated in the early years of the Society, is that interference with the experience or expression of negative emotion is deleterious for health (2). This perspective can be traced to the work of Breuer and Freud in 1895 on conversion disorder (14). Their core psychoanalytic hypothesis was that conversion disorder or hysteria arose because affect that was activated during the traumatic event could not be expressed at the time of the trauma. Symptoms were a result of this failure to express the emotions associated with the event, as the unexpressed emotions prevented the memory of the trauma from dissipating. The "strangulated" affect was expressed indirectly, however, in a somatic symptom which symbolically represented that which had been repressed; e.g., paralysis of an arm was related to the unexpressed wish to strike out. This foundational concept of repressed emotion led to a variety of applications, one of which was to the mind-body problem and what came to be known as the modern field of psychosomatic medicine (15). Franz Alexander at the Chicago Psychoanalytic Institute advanced psychosomatic theory by articulating the concept of visceral neurosis, which was based on the observation, modifying Freud's idea, that the symptom itself did not have symbolic meaning for the patient (16). Rather, a psychological conflict led to repression of affect, which continued to be expressed physiologically, resulting in dysregulated physiology and disease. Grinker and Spiegel (17), also from Chicago, observed in World War II veterans that psychosomatic conditions, such as peptic ulcer, arose when the physiological expression of emotion persisted in the absence of the conscious experience of emotional distress. This classic psychoanalytic approach to psychosomatic disorders reached its zenith in the 1940s when Alexander formulated what came to be called "Specificity Theory" (16). Using the premier "science of the mind" at that time-psychoanalysis, he sought to emulate the successes of infectious disease by identifying the specific psychological contexts in which specific physical disorders occurred (2). Alexander held that each of seven disorders, including essential hypertension, bronchial asthma, ulcerative colitis, peptic duodenal ulcer, neurodermatitis, rheumatoid arthritis, and thyrotoxicosis, was associated with a specific unconscious conflict (16). For example, those prone to hypertension had a conflict about expressing aggressive tendencies for fear of jeopardizing important relationships. Hypertensive individuals seemed to be meek and compliant and were thought to not be in touch with or consciously aware of their aggressive tendencies. Episodes of hypertension were thought to be elicited in contexts in which aggressive tendencies were activated but could not be consciously processed, resulting in the bodily expression of the emotion (i.e., hypertension). This was a visceral neurosis in the sense that the blood pressure elevation itself did not have symbolic meaning. It was recognized that such a conflict alone was not sufficient to bring on the disorder and that constitutional (e.g., genetic) and other factors contributed. Although the nature of the unconscious conflict was unique to each disease, all core conflicts were associated with repression of emotion. Alexander asserted that the essence of the analytic process in the treatment of psychosomatic patients was bringing into the patient's consciousness emotions and motivations of which he or she was unaware. Thus, although each disorder was postulated to be associated with a specific conflict, there was also a common denominator involving emotions that were not consciously processed that applied to all of the disorders. Alexander's Specificity Theory has not stood the test of time. Although some empirical support for the theory was obtained (18), empirical tests using standardized measures and unbiased sampling methods typically failed to provide supportive evidence (19). These results were due, in part, to the fact that conditions such as hypertension had diverse etiologies and because the psychological profile of patients with a given disorder varied as a function of the stage of the disorder (15). However, Theme 2 is still an accepted approach to understanding how psychological factors contribute to physical disease, as illustrated by constructs that are still current such as the repressive coping style (20), alexithymia (21), denial (22), suppression (23), Type C personality (24), and the Type D personality (25). All of these constructs have in common that negative emotions, particularly in the context of stressful life circumstances, are either not consciously experienced or are somehow stymied in their outward expression. Each is associated with findings demonstrating a linkage to adverse health outcomes. One of the challenges associated with this line of research is demonstrating the presence of an emotional state or trait that is not reportable by the subject. An alternative variant of this approach is to assess specific emotion regulatory functions or skills, the impairment of which may be manifested as interference with emotional experience or expression. Examples include attention to emotion (26), emotional approach coping (27), acceptance of emotion The continued interest in the Theme 2 approach is illustrated by the four successful international conferences on "The (non)Expression of Emotion in Health and Disease," held in Tilburg, Netherlands, in 1996Netherlands, in , 1999Netherlands, in , 2003Netherlands, in , and 2007. In Germany, there continue to be separate departments of Psychotherapy and Psychosomatics in which the psychoanalytic approach to understanding psychological factors in physical disease is the predominant model. Although these approaches remain viable because of their clinical and cost-effectiveness (33), they are clearly in the minority in the US today. Given the extent of research associated with Themes 1 and 2, it is surprising to consider that there has been relatively little cross-talk between them. Reconciliation of the two perspectives is challenging because, at the level of self-reported experience, they seem to be contradictory. Theme 1 holds that self-report is a reliable and useful indicator of emotional state or trait. Theme 2 holds that what is most salient in the domain of emotions and health is that which cannot be experienced or expressed. Yet, given their demonstrated value over decades, each must contain at least some element of truth. How might these two approaches be reconciled? A main thesis of this paper is that neuroscience provides a way to integrate Themes 1 and 2 to create a unified framework for psychosomatic medicine research on emotion. A neuroscientific approach reconciles these two perspectives and provides an explanation for how these different psychological constructs can lead to peripheral physiologic changes that are potentially pathogenic. Before turning to neuroscience, however, we consider how the relationship between Themes 1 and 2 can be understood from a psychological perspective. of preoedipal factors in psychopathology, the conceptualization of the relationship between affect and disease shifted from a focus on conflict to a focus on deficits in affective development (15). In addition, as a counterpoint to Specificity Theory, there was growing interest in common personality characteristics that contributed to psychosomatic disorders. Thus, in a landmark paper in Psychosomatic Medicine in 1948 titled "The Infantile Personality," Jurgen Ruesch (34) attributed the psychoneuroses to "pathological development" and psychosomatic conditions to "arrested development." He posited a host of potential causes for the personality structure associated with psychosomatic conditions, such as a lack of consistent parenting or trauma that overwhelms the child's sense of mastery. Ruesch proposed that these experiences caused deficits in social learning and an "infantile" form of "self-expression" that manifested as somatic symptoms. Similarly and independently, in the 1950s and early 1960s, Marty and de M'Uzan of the French Psychosomatic School observed a specific cognitive style ("pensée opératoire") in individuals prone to psychosomatic disorders, characterized by a lack of fantasy and a preoccupation with the concrete details of external events (35). Psychological Model At about the same time, Nemiah and Sifneos Although it was hypothesized that alexithymia arose from an arrest in development, the developmental line along which arrest had occurred had not been described. The theory of "levels of emotional awareness" sought to describe the developmental line in question (38). It was inspired by observations of emotion processing deficits in patients in a medical setting such as the following case. A 41-year-old unmarried woman was hospitalized for work-up of abdominal pain. Abdominal ultrasound and endoscopy were negative. A psychiatric consultation was requested. The patient reported that persistent pain had been present for 3 months. She also stated that 3 months previously her mother died. This was actually her adoptive mother who took her in at the age of 8 years. Her biological parents were alcohol dependent and physically abusive. She was placed in a series of foster homes until she was adopted. Her adoptive mother was a very kind and nurturing person. She suffered from diabetes and peripheral vascular disease. During the last 5 years of the mother's life, the patient cared for her on a daily basis. One week before her death, after a medical setback, the patient contemplated the possibility of her mother's death and was overwhelmed with a feeling of grief. A week later, mother died of a gastrointestinal vascular obstruction. After her mother's death, the patient experienced no feelings of grief or sadness. In addition to pain, she suffered from anhedonia, had neurovegetative symptoms of depression, and met criteria for major depressive disorder. A case such as this, which is commonly seen, raises a number of intriguing questions. What happened to the grief that she had experienced 1 week before mother's death? Why was the absence of experienced grief concomitantly associated with pain? Was her apparent difficulty in experiencing grief related to her early childhood history of abuse? Lane and Schwartz Emotional awareness is considered to be a separate line of cognitive development that may proceed independently from other cognitive domains (39). The concept that development can proceed at different rates in different domains of knowledge is known as horizontal decalage (40). In principle, it is possible that a developmental arrest can occur in one domain whereas development in other domains of intelligence continues unabated. The model posits five "levels of emotional awareness" that share the structural characteristics of Piaget's stages of cognitive development (41). The five levels of emotional awareness in ascending order are awareness of physical sensations, action tendencies, single emotions, blends of emotions, and blends of blends of emotional experience (the capacity to appreciate complexity in the experiences of self and other). The five levels therefore describe the cognitive organization of emotional experience. The levels are hierarchically related in that functioning at each level adds to and modifies the function of previous levels but does not eliminate them. For example, blends of emotion (Level 4 experiences), compared with action tendencies (Level 2 experiences), should be associated with more differentiated representations of somatic sensations (Level 1). The feelings associated with a given emotional response can be thought of as a construction consisting of each of the levels of awareness up to and including the highest level attained. The trait level of function is the level at which a given individual typically functions Modern conceptions of cognitive development have refined Piaget's views but are still consistent with the model proposed here. Karmiloff-Smith To illustrate with the clinical example above, the anticipatory experience of grief 1 week before the mother's death was a Level 3 experience. After the mother's death, the patient experienced abdominal pain, which was in her case a Level 1 (somatic) experience of grief. This also illustrates the distinction between explicit (grief) and implicit (abdominal pain) emotional processing. It deviates from psychodynamic thinking, however, by not attributing a motivational basis or mechanism (such as repression) that would explain why the level of processing changed, although such a motivational basis is possible. Levels of Emotional Awareness Scale (LEAS): Psychometric Findings The LEAS is a written performance measure that asks a person to describe his or her anticipated feelings and those of another person in each of 20 vignettes described in two to four sentences (29). Scoring is based on specific structural criteria aimed at determining the degree of differentiation in the use of emotion words (the degree of specificity in the terms used and the range of emotions described) and the differentiation of self from other. The scoring involves essentially no inference by raters. Because the scoring system evaluates the structure of experience and not its content, individuals cannot easily enhance their scores or create a socially desirable impression in their responses, as is the case with some self-report instruments. A glossary of words at each level was created to guide scoring. Each of the 20 vignettes receives a score of 0 to 5 corresponding to the cognitive-developmental theory of emotional awareness that underlies the LEAS (38). A score of 0 is assigned when nonaffective words are used, or when the word "feel" is used to describe a thought rather than a feeling. A score of 1 is assigned when words indicating physiological cues are used in the description of feelings (e.g., "I'd feel tired"). A score of 2 is assigned when words are used that convey undifferentiated emotion (e.g., "I'd feel bad"), or when the word "feel" is used to convey an action tendency (e.g., "I'd feel like punching the wall"). A score of 3 is assigned when one word conveying a typical, differentiated emotion is used (e.g., happy, sad, angry). A score of 4 is assigned when two (or more) Level 3 words are used in a way that conveys greater emotional differentiation than would either word alone. Respondents receive a separate score for the "self" response and for the "other" response ranging from 0 to 4. In addition, a total LEAS score is given to each vignette equal to the higher of the self and other scores. A score of 5 is assigned to the total when self and other each receive a score of 4 and are differentiated from one another; thus, a maximum total LEAS score of 100 is possible. The LEAS has consistently been shown to have high interrater reliability and internal consistency (46). The test-retest reliability at 2 weeks has been shown to be good. Norms for age, gender, and socioeconomic status have been established. A variety of studies (but not all) (47,48) support the construct validity of the LEAS. The LEAS correlates moderately positively with two cognitive-developmental measures-the Sentence Completion Test of Ego Development by Loevinger et al. (49, IMPLICIT AND EXPLICIT EMOTIONAL PROCESSES Clinically, it has been shown that patients with borderline personality disorder score lower on the LEAS than age-matched control subjects (56) and that individuals with the "disorganized attachment style" have lower LEAS scores than those with the "organized attachment style" (57). Patients with irritable bowel syndrome (IBS) do not on average have lower LEAS scores than healthy controls but, among those with IBS, lower scores on the LEAS are associated with greater pain (58). Patients on a psychosomatic inpatient ward with somatoform disorders had lower LEAS scores than patients with disorders involving psychological distress, such as depression. This same study showed that somatoform patients showed significant increases in LEAS scores after 3 months of multimodal inpatient treatment that integrated body-based techniques with intensive group and individual psychotherapy (59). These findings support the theory that impairments in emotional awareness can occur developmentally, that lower emotional awareness is associated with a greater tendency to experience emotional distress as bodily symptoms, and that emotional awareness can improve with therapeutic interventions that facilitate the transition from implicit to explicit processing. The LEAS has also yielded useful findings in a variety of other clinical settings. Patients with essential hypertension had lower LEAS scores than those with hypertension secondary to other medical conditions, such as renal disease (60). Patients with eating disorders (anorexia and bulimia) were observed to have lower LEAS scores than matched controls (61), consistent with Hilde Bruch's classic observation that eating disorders are associated with an impairment in interoceptive awareness of one's own emotions (62). Patients with posttraumatic stress disorder (PTSD) have lower LEAS scores than matched controls, and LEAS scores were inversely correlated with the severity of PTSD symptoms, particularly symptoms involving dissociation (63). Patients with morbid obesity were observed to have lower LEAS scores than controls, and it was also observed that, among the obese patients, the higher the LEAS scores the greater their social anxiety (64). The latter finding indicates that greater emotional awareness is associated with a greater awareness of the negative emotional responses that morbid obesity elicits from others. A related finding is that individuals with generalized anxiety disorder have greater LEAS scores than matched controls (65), indicating that emotional awareness can be a double-edged sword. In contrast, patients with depression were found to have decreased awareness of the emotions of others (66,67), consistent with the pathological introspective focus that can occur with depression. Together, these findings indicate that the LEAS can detect variations in emotional awareness that have meaningful clinical correlations. A key issue in the assessment of emotion-processing deficits is the need to distinguish between Theme 1 and Theme 2 phenomena. The LEAS typically does not correlate with selfreported negative affect, such as anxiety or depression, in the absence of anxiety or depressive disorders (29), unlike the 20-item Toronto Alexithymia Scale, which typically does (47). In three separate studies, including studies of essential hypertension (60), eating disorders (61), and somatoform disorders (59), it has been shown that associations between lower emotional awareness and the clinical condition were not altered by partial correlations removing variance due to negative affect, whereas control for self-reported negative affect rendered associations with the TAS-20 nonsignificant. These findings suggest that the LEAS may have an advantage in those contexts in which distinguishing between Theme 1 and Theme 2 phenomena is important. Brain Model-Early History Scientifically credible models of how the human brain processes emotion have only been formulated in the past 100 years. The model to be presented here builds on important forerunners early in the 20 th century. In the early 1920s, Walter Cannon suggested that the physiology of emotion provided a key link between mental states and physical disease. He pointed out that subcortically generated emotion could be routed downstream to the hypothalamus associated with physiological expression or upstream to the neocortex for symbolic representation and expression (68). However, Cannon did not attempt to formulate a detailed model of how the brain mediates emotion (69). In 1937, James Papez published a landmark paper in which he proposed the first complete neural circuit mediating emotion (70). This proposal postulated a role for brain structures, whose function was not understood, in the mediation of a function, emotion, whose locus in the brain was not understood. Papez proposed that emotion could be induced either through stimulus perception or thought by engaging a reverberating circuit that included the mamillary bodies of the hypothalamus, the anterior thalamic nucleus, the anterior cingulate cortex (ACC), and the hippocampus. Although current models of how the brain mediates emotion are very different from the Papez model, his hypothesis that the dorsal anterior cingulate cortex (dACC) was the seat of emotional experience was surprisingly prescient. In the 1930s and 1940s, Kluver and Bucy MacLean had therefore proposed a neural basis for the interference or blockage of emotion processing inherent in Theme 2. He quoted the model by Ruesch (34) (described above) that itself was a forerunner of the alexithymia construct and the levels of emotional awareness model. Although MacLean's model was consistent with the concept of interference or blockage of emotion processing, it did not require a motivated basis for the lack of transfer, consistent with Ruesch's deficit model. This neural model lay dormant for many years, however, because the primary tools for studying the brain in humans at the time, such as the study of patients with brain lesions identifiable during surgery or postmortem studies, did not lend themselves to examining how brain function might contribute to the development of psychosomatic disorders in the typical context of individuals with intact brains. However, with the advent of modern techniques for functional and structural brain imaging, MacLean's model becomes highly relevant and testable. One might therefore ask, "How do we currently understand the role of cortical and subcortical structures in the mediation of emotional awareness?" Neural Substrates of Emotional Awareness To date, there have been two imaging studies in which the LEAS has been correlated with brain activity during emotion in healthy subjects. The first included 12 right-handed female volunteers who were free of medical, neurological, or psychiatric abnormalities. Happiness, sadness, disgust, and three neutral control conditions were induced by film and recall of personal experiences (12 conditions). Twelve 60-second positron emission tomography (PET) images of regional cerebral blood flow (rCBF) were obtained in each person, using intravenous bolus injections of 15 O-water (46). We examined the correlates of rCBF due to emotion by subtracting the neutral conditions from the emotion conditions. Subtracting the neutral conditions removes brain activity due to the experimental paradigm that is not due to emotion. To ensure sufficient statistical power, we combined the three emotion conditions together and did the same for the neutral conditions. We then correlated LEAS scores with the emotionminus-neutral rCBF difference to identify regions of the brain that correlated with emotional awareness during emotional arousal. Findings from this covariate analysis revealed one cluster for film-induced emotion with a maximum located in the right midcingulate cortex. For recall-induced emotion, the most statistically significant cluster was located in the right ACC. An analysis was then performed to identify areas of significant overlap between the two covariance analyses. Based on an a priori statistical threshold, a single cluster was observed in the right dACC (BA 24) (46). Traditionally, the dACC has been thought to have a primarily affective function (70,75). However, in addition to emotion, it is now recognized to play important roles in attention, pain, conflict monitoring, response selection, maternal behavior, vocalization, skeletomotor function, and autonomic control (76). How can these different functions be reconciled with the present findings involving emotional awareness? One answer is that these various functions of the dACC reflect its superordinate role in executive control of attention and motor responses (46). According to this view, emotion, pain, or other salient exteroceptive or interoceptive stimuli provide moment-to-moment guidance regarding the most suitable allocation of attentional resources for the purpose of optimizing motor responses in interaction with the environment. The findings from this study suggest that attentional resources are more readily engaged by one's own emotional responses among those individuals who are more emotionally aware. To the extent that people who are more emotionally aware attend more to internal and external emotion cues, the subsequent cognitive processing of this information can contribute to ongoing emotional development. The fact that the dACC can be engaged by so many other kinds of nonemotion stimuli may mean that less emotionally aware individuals tend to use their attentional resources more readily for purposes other than processing internal or external emotional signals. These findings were replicated and extended in a second PET study using the same radiotracer, 15 O-water (77). Pleasant, unpleasant, and neutral pictures from the International Affective Picture System (IAPS) (78) were presented in blocks so that we could obtain scans corresponding to high arousal pleasant, low arousal pleasant, high arousal unpleasant, low arousal unpleasant, and neutral conditions. This design enabled us to disentangle whether LEAS correlated with valence (pleasant, unpleasant), arousal (high, low arousal), or both. We also were able to consider sex differences as 22 men and 22 women were studied. We observed a significant correlation between dACC activity and LEAS in the high arousal (pleasant and unpleasant) conditions relative to the low arousal (pleasant and unpleasant) conditions. This finding was observed in the combined sample of men and women but was greater in women. The association between LEAS and dACC was not observed as a function of pleasant or unpleasant valence relative to neutral or in the low arousal conditions relative to neutral. IMPLICIT AND EXPLICIT EMOTIONAL PROCESSES These results illustrate that engagement of the dACC differs as a function of individual differences in emotional awareness in the context of conditions that induce high arousal emotions. Individuals who are more emotionally aware are better able to tolerate and consciously process intense emotions than those who are less aware (79 -81). Conversely, individuals functioning at a lower level are more likely to behave impulsively and be less aware of what they are feeling in the context of high arousal emotions (44). This may be understood as a greater ability among more highly aware individuals to be cognizant of their own emotional reactions in the context of high arousal and to anticipate and evaluate the consequences of their actions in advance of their behavioral expression. This greater ability may be mediated at least in part by the dACC, consistent with Paus' view that the dACC is fundamentally involved in translating intentions into actions in the context of emotional arousal These findings suggest an important role for the dACC in the conscious processing of emotional information. How does the brain process emotional information that is not conscious? This has been studied in the functional imaging context, using a technique called backward masking, in which the target stimulus, typically a picture of a facial expression such as fear or anger, is presented very briefly (typically Ͻ30 milliseconds) and prevented from being consciously processed by a "masking" stimulus, typically a neutral face, that follows immediately and is presented for a longer duration (typically 100 milliseconds). This approach has been used effectively to demonstrate amygdala activity when facial expressions of fear (84) or anger (85) are not consciously perceived. These imaging studies have not, however, attempted to simultaneously examine subjective emotional responses to these stimuli. The most rigorously conducted study to date that examined whole-brain activity during implicit and explicit processing of emotional faces in the same subjects was published in 2006 by Leanne Williams and colleagues (86). Consistent with previous work Another feature of this study was that amygdala activity was examined as a function of implicit and explicit processing during the first half and second half of the experimental trials, which averaged 1300 milliseconds in total. Results showed that amygdala activity was greater during the first half relative to the second half. The decrease in amygdala activity during the second half, however, was greater with explicit rather than implicit processing. The small decrease from the first to the second half during implicit processing is likely due to the inhibitory influence from the ventromedial prefrontal cortex. The greater decrease with explicit processing is likely due to the combined effect of ventral and dorsal medial prefrontal top-down inhibitory influences. Thus, conscious, relative to unconscious, processing is associated with greater inhibition of the amygdala, and late amygdala activity was numerically greater during unconscious than conscious processing (p Ͻ .098). These findings are reminiscent of those with the LEAS in which higher levels of conscious processing were associated with greater attenuation of Level 1 and Level 2 phenomena. Model of the Neural Substrates of Implicit and Explicit Emotional Processes The distinction between implicit (nonconscious) and explicit (conscious) processes is foundational in cognitive neuroscience because their neural substrates are dissociable (88). The distinction was first applied to memory. Explicit memory for facts and events requires participation of medial temporal lobe structures (such as the hippocampus) and diencephalon, whereas implicit memory requires structures such as the striatum (skills and habits), neocortex (priming), amygdala and cerebellum (classical conditioning), and reflex pathways (nonassociative learning). Implicit processes have also been demonstrated in a variety of other cognitive domains, including attention, perception, and problem solving (88). This body of research has led to a growing recognition that consciousness is the tip of the cognitive iceberg in the sense that the vast majority of cognitive processing occurs outside of conscious awareness (89). Antonio Damasio's distinction between primary emotion and feeling and their dissociable neural substrates (90) paved the way for the application of the implicit-explicit distinction to emotion. Primary emotion is the phylogenetically older behavioral and physiological expression of an emotional response. Primary emotion occurs automatically and without the necessity of conscious processing. Feeling, on the other hand, involves the conscious experience of that emotional state. According to Damasio, primary emotion and feeling are separable, both conceptually and neuroanatomically (90). Whereas primary emotion is necessary for successful adaptation to environmental challenges and the physiological adjustments needed to meet those challenges, a conscious feeling state enables previous emotional experiences to be consciously recalled or current experiences to be accessed and used for decision-making and navigation of the social world. In this paper, the term "affect" encompasses both primary emotion and feeling. One might question whether a distinction applied in the cognitive domain applies to emotion. My colleagues at the University of Arizona and I hold that there is nothing about emotion that is not cognitive if one equates cognition with information processing (91). Others including Kihlstrom and colleagues (45), who were early investigators of implicit cognition (92), argued that the implicit-explicit distinction applies to emotion as well as cognition. Parallels between the corresponding neuroanatomical and psychological models are schematically depicted in Although the neural substrates of each level of emotional awareness are not yet known, a rudimentary neuroanatomical model that distinguishes between implicit and explicit processes can be formulated (44). This model distinguishes between the neural substrates of implicit emotion, on the one hand, and three different aspects of the conscious experience of emotion: background feelings, focal attention to feelings, and reflective awareness Several fundamental principles characterize this model. 1. There are distinct neural substrates of implicit and explicit emotional processes. 2. The structures that are uniquely responsible for emotion generation operate implicitly, i.e., outside of conscious awareness. 3. The brain structures needed for conscious emotional experiences emerged more recently during phylogenesis compared with those required for implicit processes. 4. The structures that preferentially participate in the conscious experience of emotion are not unique to emotion but rather perform domain-general cognitive functions. 5. Emotions compete with other inputs for conscious processing. Implicit Aspects of Emotion Brain imaging studies in recent years demonstrated that amygdala activation can occur in response to emotional stimuli in the absence of conscious awareness of the stimulus. This finding is consistent with the notion that the amygdala performs its functions implicitly and plays a key role in establishing whether exteroceptive stimuli have an emotional meaning. Echoing the distinction by Papez between emotion induced by perception versus thought (70), LeDoux has written widely about the distinction between the thalamo-amygdala pathway for processing exteroceptive stimuli rapidly and crudely in the absence of conscious awareness (69), and the neocortical-amygdala pathway that provides more precise and differentiated identification of the stimulus requiring an additional 12 milliseconds of processing time. The time saved by having behavior directed by implicit processes could potentially mean the difference between life and death in lifethreatening situations. This is not inconsistent with the view that the capacity for explicit processes, such as reflection, deliberation, delay of impulses, planning and the like, are arguably more important to adaptive success in modern human cultures, in which success or failure rather than life or death is the more salient issue. There are several additional lines of evidence that further support the hypothesis that the amygdala executes its functions implicitly. Studies of the subjective experience of emotion in patients with unilateral and bilateral amygdala lesions show that emotional experience is not appreciably altered in these patients compared with controls (94). During ictal fear, cortical structures such as the ACC and prefrontal cortex are activated in addition to the amygdala, whereas when amygdala activation alone is observed during seizures detected by depth electrodes, the experience of fear does not occur (95). Other structures that participate in implicit emotion processing include the thalamus, hypothalamus and pituitary, basal ganglia, and brainstem nuclei, the latter including those involved in autonomic regulation and ascending neurotransmitter systems. An important new development is that "unconscious emotion" is now recognized as a legitimate topic of scientific inquiry in academic psychology (96). For example, studies have examined how emotional stimuli that are not consciously detected influence behavior and subjective experience. A dialogue between researchers in this area and researchers in psychosomatic medicine will likely be of mutual benefit as this work progresses.