10
Affect, Cognition, and Symptom Perception

Of what value is a chapter on emotion, cognition, and symptom reporting for professionals involved in research and treatment of cancer pain? Two obvious links exist: first, pain is typically interpreted as a symptom, i.e., as an indicator of an underlying disease or injury and, second, as is true for all symptoms, pain is a subjective experience. The process by which a somatic sensation such as pain is converted and responded to as a symptom, and the cognitive and emotional factors affecting this process, therefore, should be of interest to both the investigator and clinician who use such reports to study and treat pain.

We have also been told that most of the scientific presentations at the cancer center focus on the neurological and pharmacological mechanisms underlying pain experience and the implication of these mechanisms for treatment. These presentations deal with the relationship among pain reports, the level of noxious stimulation, the amount of activity in various neural pathways, and the receptor sites that affect conduction in these pathways. When viewed from that perspective, emotional and cognitive factors affecting pain reporting appear as confounds that disrupt the systematic analysis and control of the pain process and efforts are made to identify and eliminate them. Examples can be seen in signal detection models designed to separate the affective ``bias'' from stimulus discrimination and research on individual differences that assign the bias in pain reporting to factors such as ``illness behavior'' (59,66) and hypochondriasis (3).

The physiological analysis of pain mechanisms, the use of signal detection analysis to obtain sensory thresholds, and the study of individual differences in illness behavior have been valuable in alerting us to the variety of factors that affect pain reports and have suggested new possibilities for pain management and the treatment of the pain patient. This chapter, however, takes a somewhat different approach to the symptom reporting and pain problem. Our main thesis is that somatic signals such as pain are embedded in cognitive and emotional processes. Thus, to understand the processes underlying this complex experience, we must identify the components of the sensory, cognitive, and affective mechanisms that contribute to symptom and pain experience and understand the processes involved in their integration. Thus, the proposed model treats sensory, cognitive, and emotional components as integral parts of an information-processing system that transmutes sensory experience into symptoms.

Our first task, therefore, is to provide an overview of a model that describes how a somatic sensation, such as pain, is processed to become a symptom; that is, how it is transformed from a sensory event to an indicator or representation of an underlying threat of disease or injury. Our work has attempted to identify the processes involved in the formation of representations and the way in which representations guide the procedures selected to manage these disease threats and to cope with the emotional distress associated with them. The terms most often adopted to identify this area of research have been (a) the study of illness cognition in self-regulation (77) and (b) the contribution of common-sense models of illness to self-regulative processes (52).

Our second task will be to address how emotional factors may affect the self-regu lative system and symptom reporting. We will raise a number of questions about the way emotion influences symptom reporting and warn you beforehand that there is a modest amount of data to support some of the speculative ideas that we will present.

A SELF-REGULATION MODEL FOR ILLNESS COGNITION

We assume that people are active problem solvers and that a self-regulation systems model provides a framework for describing and analyzing their behavior in elaborating and responding to somatic stimuli. Figure 1 provides a schematic view of the model. When you examine this figure, you should think of it as a picture of the individual's psychological system, that is, of the processes and steps involved in converting a sensory signal into a symptom report. It is worth remembering that the psychological mind represented by this figure is likely to have little or no knowledge of the neurological and physiological processes underlying its experience; i.e., it may know nothing about myelinated A and unmyelinated C fibers, may never have heard of the thalamus, the spinothalamic pathway, or periaqueductal gray, and may be completely ignorant of n and l receptor sites. Moreover, the system does not have direct access to these underlying processes; i.e., unlike the objects in the outer world, the neural receptors, conducting pathways, and n and l receptor sites are not represented in consciousness. Yet this mental apparatus will generate a representation of health threats from both somatic sensations and social stimuli, will marshal procedures to cope with these ``symptoms,'' will evaluate the outcomes of these procedural efforts, and will respond emotionally to these threats and coping outcomes. These reactions occur as a somatic stimulus is integrated with a complex system of declarative and procedural memory structures that process or elaborate ``raw'' sensory events into representations or indicators symptomatic of underlying disease.

Experimental data (45) suggests there are two, partially independent arms to the self-regulation system (see Figure 1 ). The upper, cognitive arm provides the meanings that define the nature of the health threat, procedures for controlling it, and criteria for evaluating the success of coping efforts. The lower, emotional arm represents the affective response to the sensation and its representation, the procedures for regulating these emotional reactions, and the criteria for evaluating the success of coping with these feeling responses. If the stimulus is a somatic sensation, the intensity of the emotional reaction will vary with its intensity and its meaning. For example, very severe pain can elicit high levels of distress and fear and lead directly to help seeking. The emotional reactions and coping procedures adopted for a mild level of pain, however, will differ primarily as a function of the meaning attributed to the pain; pain interpreted as a muscle sprain will likely arouse less fear than pain attributed to a cancer. The two arms are in constant interaction with one another.

Although this model of the conversion of sensations to symptoms was generated to fit the data generated in social psychological studies of fear communications (45), it also has been used to design and interpret studies examining the reduction of stress during medical treatment (35,36). The homology between the psychological model and physiological models of the pain system (12), i.e., parallel and interacting arms for informational and emotional processing, encouraged this application (27,51).

In most instances, a representation of a health problem is generated by a set of somatic sensations rather than by a single sensation. Symptom sets such as (i) headaches, face flushing, and chest pain, (ii) hard or tender growths in various parts of the body accompanied by fatigue and general malaise, and (iii) coughs, running noses, and fatigue could lead to self-diagnoses of heart attack, cancer, and a cold, respectively (6). Given their implied meaning, it would not be surprising to find that each set evokes different procedures for self-evaluation and self-treatment. The heart attack set might call for cessation of activities and consultation with a family member or a physician while the cold set might call for home remedies such as vitamin C, fluids, and rest. We can expect the feedback generated by these procedures to validate or invalidate their usefulness and to change the representation depending upon the fit between the feedback and the biological factors underlying the symptom.

Illness Representations: Attributes and Categories

Five attributes have been identified in studies of disease representations, and combinations of the five define different common-sense illnesses. The five are (i) the identity of the problem, i.e., its name and symptom indicators (62); (ii) the antecedent causes (e.g., exposure to infection, hereditary weakness, stress) (4); (iii) the possible consequences (physical, economic, and social) (20,50); (iv) the time line, time to occurrence, duration until cure or death (18,62); and (v) the controllability, i.e., whether there is a procedure available that can remove or stabilize the problem (42). These features guide the selection of coping procedures and shape their execution. For example, symptoms represented as a coronary threat stimulate rapid help seeking and the cessation of activity (58), those of a possible cancer may lead to help seeking or to a ``wait and see'' strategy to determine whether they are getting worse (11), and the symptoms of a common cold may stimulate self-care procedures such as getting extra sleep, taking aspirin or vitamin C, and drinking fluids.

Coping Procedures and Outcome Appraisals

The appraisal following a coping procedure is typically focused on the evaluation of the adequacy of the procedure (e.g., did the aspirin relieve the headache?). However, it can also lead to reevaluations of the self-diagnosis (e.g., as aspirin did not relieve the headache, the headache could be serious and not simply due to stress or to a cold), and to reevaluations of the competency of oneself and the support system used for managing threat (e.g., I can't do anything about this nor can my doctor). As this process extends over time, the representation of an illness threat will be updated and become more elaborate, its representation progressing from that of a mild condition of questionable meaning (Is it a cold or just fatigue?) to one that is more specific and possibly more threatening (11). The level of threat coupled with the perceived ease of control can stimulate a variety of emotional responses such as anger, fear, and depression if there are great threat and poor possibility of control, or relief and joy if the threat is mild and/or controllable.

Process Assumptions

Three assumptions need to be delineated with respect to the operation of the model, as they are important for symptom reporting. First, illness cognitions are products of active, constructive systems; this point is readily apparent when representations are observed over time. Second, at least three types of memory systems, abstract, perceptual, and procedural, appear to be involved in the transformation process, i.e., in converting somatic sensations into meaningful representations. Finally, specific procedures, i.e., ways of asking and answering questions, are involved both in the transformation of symptoms to representations and in the ongoing effort to minimize the impact of an illness-symptom threat and its emotional concomitants (81).

The Constructive Process

The representation of a current or potential health threat, the coping procedures in response to it, and the evaluations of outcomes, are products of the ongoing constructive activity involved in the individual's attempt to understand and regulate a specific domain or segment of psychological reality. The somatic sensation and the ongoing problem-solving behavior focused upon it define the individual's problem space. Although Figure 1 describes these components in linear order, they are in constant interaction; representation, emotional reactions, procedures, and appraisals affect and are affected by one another. The specifics of this picture will change from moment to moment as procedures are performed for the elaboration and regulation of a somatic sensation.

Memory Systems and the Constructive Process

The representations, coping procedures, and outcome appraisals generated by a somatic sensation are products of the interaction between the sensation and the individual's knowledge base, i.e., the declarative and procedural memory systems. Our data and the data in the pain literature suggest that the knowledge base used to transform a somatic sensation into a symptom includes both semantic memory, i.e., memories of labels such as heart disease, cancer, and colds, and concrete or perceptual memories of specific illness experiences, e.g., memory of painful sensations in specific parts of the body during specific illness episodes. Thus, both abstract semantic and concrete perceptual memories are involved in the processing or transformation of a sensation into a symptom, the abstract guiding how we think, feel, and talk about a symptom and develop long-term strategies or plans for its management, the concrete or perceptual memory providing an immediate link of a sensation to emotional reactions and automatic procedures for its control.

The hypothesis that perceptual memory participates in the constructive process is supported by the dramatic reports of phantom pain. These reports indicate that a combination of somatic sensations, e.g., the sensation of a limb and the pain within it, is stored as a perceptual memory (i.e., a perceptual schema) and remains active in memory after the removal of sensory input (61). These memories are unlikely to persist in the nonamputee, as new somatosensory inputs will interact with (retrieve and change) these memory traces. Additional evidence for perceptual memory pro[chcesses can be seen in studies of dream imagery (72) and studies of visual perception (1). In all of these cases, perceptual experience is shaped by central memory stores in the absence of unambiguous external sensory inputs.

In summary, somatic inputs and the performance of procedures for their management and evaluation create new and alter existent central memory processes at the very same time that they are shaped and enriched by these central memory pro[chcesses. A somatic sensation, therefore, is not neutral; it has an identity (what it indicates), a time line or expected temporal existence, potential for control, and a set of possible consequences and antecedent determinants. An individual's response to a somatic sensation is a product of this entire system. Moreover, the perceptual and abstract aspects of a representation may not always be in accord with one another and this discrepancy can produce behavioral outcomes that are both interesting and a source of danger to the individual enacting them. The discrepancies between concrete and abstract aspects of illness representations can be seen in a wide range of studies of illness cognition.

EMPIRICAL DATA ON THE CONSTRUCTIVE PROCESS

Concrete and Abstract Features of Illness Representations

Although each of us is familiar with the unfolding and development of illness episodes, the stimuli, memories, and actions determining our experiences of illness are often inaccessible to consciousness. Even when they are accessible, they may be poorly recalled because they are fleeting and given relatively little attention and/or cognitive effort. As a consequence, the individual's knowledge of the constructive process will be vague, poorly differentiated, and rife with inconsistencies. Systematic studies of this domain, therefore, should produce more than an occasional surprise by showing that monitoring and coping with symptoms can lead to behaviors that are both risk inducing and risk reducing, depending upon the illness and surrounding circumstances.

Symptom Monitoring by Hypertensives

As an example, our studies of symptom monitoring by hypertensives suggest that abstract and concrete memory structures can make contradictory contributions to the representation of this disease that put patients at risk. The risk arises because the perceptually driven features of common-sense representation of hypertension are inconsistent with patients' abstract knowledge of the disease and its underlying biomedical mechanisms. A study by Meyer et al. (62) illustrates these points. Fifty patients in continuing treatment for hypertension were asked if they agreed that ``people cannot tell when their blood pressure is elevated.'' Eighty percent of these patients thought that ``people cannot tell''; they accepted their doctors' statements that hypertension is asymptomatic, at least at a conceptual level. However, 92% of these patients said that they could tell when their own blood pressure was elevated. They used symptoms such as headaches and face flushing as indicators. Interestingly, a sizable proportion asked the interviewer not to convey this information to their doctors.

These data led us to suggest a symmetrical relationship between labels and symptoms. Awareness of a somatic sensation encourages a search for a label and being labeled as hypertensive leads to a search for symptoms. The search for symptoms can be relatively swift and instantaneous or it can take place over a fairly extended period of time. That both may occur was suggested by the reports of another sample of hypertensives who were interviewed both at their first treatment session and 6 months later. At baseline, 71% of these 65 newly treated subjects believed they could monitor their blood pressure; the percentage holding this belief increased to 92% after 6 months of treatment (62). In comparison to both the continuing treatment and the newly treated hypertensives, only 46% of nonhypertensive patients from other clinics in the same hospital believed they could tell when their blood pressure was elevated, and those who did reported the same types of symptoms as did the hypertensives. The comparison between the baseline value for newly treated hypertensives and nonhypertensive controls, i.e., 71% versus 46%, suggests that symptom identification is fairly rapid, and the change from 71% to 92% during hypertension treatment suggests it can take place over an extended period of time. The data, however, do not provide any precise assessment as to how quickly this occurred.

Patients' beliefs in the symptomatic nature of hypertension were not benign. In the continuing treatment group, patients using symptoms as indicators of blood pressure were less likely to use their medications as prescribed and were in poorer blood pressure control if they believed their medications did not have a positive effect on their symptoms (62). Similarly, a greater proportion of those newly treated patients who believed at the outset that they could use symptoms to monitor their blood pressure had dropped out of treatment within 6 months compared to those who did not hold this belief.

Symptom Monitoring and Cancer

The response of patients receiving chemotherapy treatments for cancer provided another set of examples of discrepancies between concept and symptom perception, though increased distress in treatment rather than quitting treatment (which was very rarely observed) was the behavioral outcome. Women in adjuvant treatment for breast cancer, perceiving themselves as cured as their cancers had been surgically removed, appeared to be more distressed by the noxious side effects of the chemotherapy treatment than were women in treatment for metastatic disease. The adjuvant treatment group had to reconcile being ``surgically cured'' with the ``sickness'' induced by chemotherapy, whereas women with metastatic disease could perceive and feel the noxious treatment effects as a defense against existent tumors. A similar effect arose with patients afflicted by malignant lymphoma: those whose tumors regressed abruptly at the outset of chemotherapy found continuing this noxious treatment more distressing and less ``sensible'' than did those whose tumors regressed more gradually (63).

Symptom Monitoring in a Clinical Trial

We encountered an interesting example of symptom monitoring by ex-cancer patients participating in a tamoxifen toxicity trial (56). All of the 140 women participating in this double-blind, randomized trial had been successfully treated for a breast lesion. The trial objectives were to determine if tamoxifen, an estrogen antagonist, accelerated bone loss and had any deleterious effect on blood lipids. Each participant was interviewed at the outset of the trial and at 3-month intervals during the first year with respect to their guess as to what drug they were receiving, their symptom experience, breast self-examination, and other behaviors. Although the trial was double-blind, the placebo was inert and the subjects receiving tamoxifen experienced symptoms such as hot flashes and bone pain. Not surprisingly, participants were accurate in guessing whether they were receiving tamoxifen or a placebo (53). This awareness also affected their behavior; subjects in the tamoxifen group performed breast self-examination on a regular (nondeclining) basis during the first 6 months of the study, whereas those receiving placebo showed a significant decline in this important self-appraisal action. This effect occurred despite the extensive training in breast self-examination that was given all participants at the outset of the trial.

The precise reason for these differences in breast self-examination over time is unclear. One possible interpretation is that subjects on tamoxifen believed they were receiving protection against cancer recurrence and felt less threatened that self-examination would lead to the detection of a recurrence of cancer. Those on placebo had no such assurance. Another possibility is that tamoxifen's systemic symptoms served as a reminder for breast self-examination. Finally, it is conceivable that subjects who believed they were on placebo felt neglected or poorly treated, and for this reason rejected staff recommendations to engage in breast self-examination.¹ Regardless of which of these interpretations is correct, the change in behavior of the placebo subjects could reduce the possibility for early detection of disease.

Positive Effects of Symptom Monitoring

The effects of monitoring and interpreting symptoms need not always be negative. For example, while hypertensives who monitor their symptoms are more likely to miss their medication, we found the hypertensives were more likely to quit smoking. There are some diseases for which monitoring symptoms may prove valuable for self-treatment if patients are trained to make proper use of their somatic experiences. Diabetes is one such disorder. Although the monitoring and interpretation of somatic symptoms by untrained diabetics may be risk inducing if these patients interpret the dysphoric effects of hypoglycemia as sickness and the euphoria of mild hyperglycemia as wellness, diabetics can be trained to identify symptoms associated with either blood sugar elevation or depression and to treat (self-regulate) these symptoms appropriately (29).

Regardless of the direction of the effects of symptom monitoring on the self-regulation process, i.e., whether it enhances or interferes with health-promoting action, the practitioner who ignores these effects does so at his own peril and that of his patients. The risk exists because the self-regulation system is guided by cues, variations that are imperfectly correlated to, and on occasion inversely related to, changes in the underlying disease process. One result is patient dissatisfaction with health care providers, which can lead to extensive shopping for third, fourth, and fifth opinions and result in multiple prescriptions for medications whose combinations may pose serious dangers. While the hectic pace of contemporary practice restricts the time available for exploration of patients' representations and coping procedures, a grasp of the basic variables involved in the self-regulative system should establish a clear framework for communication and allow the practitioner to focus directly on these key factors.

Procedures for Constructing Representations

One aspect of the self-regulation model in need of elaboration is the identification of the various procedures people use in constructing their illness representations. These procedures help people decide whether a somatic symptom merits a symptom label and will be reported as such. Few, if any, procedures will be visible when the symptoms initiating an illness episode are clear and threatening, e.g., bleeding, severe pain, and unexpected fainting. In most cases, the representation of threat is immediate but not necessarily identified. When a presenting somatic stimulus is vague and/or mild, however, a more elaborate and complex set of evaluative procedures will be used over an extended period of time before the somatic sensations are updated to become full-blown symptoms and indicators of an illness threat.

Investigators have found it useful to divide the elaboration process into an initial, or appraisal, phase represented by the time period from the first notice of the sensation until the decision that one is ill; a second, or illness, phase extending from the decision that one is ill until calling for health care; and a third, or utilization, phase of the period from calling for care until seeing a practitioner (10,44,74,80). The duration of the first two phases is primarily under the control of the patient, whereas that of the third is controlled by the provider.

Our assumption that somatic sensations are elaborated by both abstract and perceptual memories suggests that a variety of procedures will be used in constructing representations and that these procedures will vary across the phases. We can also expect procedures to vary in the degree to which they are performed automatically or deliberately, depending on whether or not they are guided by perceptual memories activated by a somatic sensation or products of abstract (conceptual) thought. It will also be more difficult to learn about automatic procedures, as they are less likely to be accessed by consciousness and be reported on than procedures that are guided by abstract conceptual structures that take place over extended periods of time.

Automatic, Perceptually Driven Procedures

The results from the Meyer et al. study (62) of hypertensives suggested that a pressure toward symmetry to match labels and symptoms exists in the perceptual cognitive system is likely automatic and determined by the underlying schema of hypertension (7). That the schema may be culture wide is suggested by the data that both hypertensives and nonhypertensives report the same symptoms, e.g., heart beating, headaches, and face warming, when judging their blood pressure to be elevated.

Correlational data, however, can provide only presumptive evidence for symmetry and its impact on symptom reporting. A laboratory experiment by Baumann et al. (4) provided additional experimental evidence (see also ref. 18). Eighty undergraduates who were told they were participating in a health attitude study had their blood pressures taken at the beginning of the laboratory session and were randomly assigned to either normal or high feedback conditions. In the ``normal reading'' condition, the experimenter casually mentioned that the subjects' pressures were normal, ``118 over 72.'' In the ``high reading'' condition, the experimenter, looking and sounding a bit concerned, mentioned that their pressures were ``142 over 88.'' Before leaving the room, the experimenter asked all subjects to complete a questionnaire that included a symptom checklist. The data showed that subjects in the high feedback condition reported more symptoms than did subjects in the normal feedback condition, and the symptoms they reported were the same as those reported by hypertensives, i.e., headache, heart beating, and tension.² The data support the hypothesis that labeling someone as hypertensive can lead to higher levels of symptom reporting. It is unclear whether this result would appear for feedback about other illness labels or for populations other than college students. That the results are not specific to this particular sample is suggested by data showing that adult hypertensives report similar symptoms and that the symptoms are not associated with actual blood pressure readings (5,65).

Reports by individual subjects point to a variety of automatic search procedures that are guided by an underlying illness schema. A sore, lump, or pain in the throat or a limb can elicit self-exploration and palpation of the tender area in search of correlated symptoms or in search of symmetrical symptoms. Each of these procedures is useful for answering a variety of questions. If a sore throat is accompanied by redness in the throat and stuffiness in the nasal passages, the syndrome would suggest the common cold; if the soreness increases to a gentle touch, it may suggest an infection; and if the lump exists in a corresponding body area, it may suggest a normal body structure rather than an unusual growth.

Because many of these procedures are executed ``unconsciously,'' i.e., without deliberate thought, it will be difficult to study them through verbal report. Observational techniques, filming people during feedback episodes when they are alone or with others, could provide interesting data respecting the conditions under which they are performed.

Extended/Deliberate Procedures

Many procedures for the evaluation of somatic sensations are volitional or deliberate, and extended over time. Perhaps the most familiar of these is the decision to ``wait and see,'' an option that is likely to be chosen during the appraisal phase when the somatic sensation initiating the appraisal process is vague and mild and not clearly definable as a symptom or disease indicator (11). Somatic changes that are slow to develop and mild are also more likely to elicit an attribution to aging and lead to delays in care seeking rather than be interpreted as signs of illness (67).

Data also suggest that people engage in environmental searches to answer questions about somatic indicators. For example, mild and vague somatic sensations are likely to be seen as indicators of stress rather than illness if they are contingent with the appearance of new environmental stressors (4). When a somatic sensation is contingent with a new stressor, it is unlikely to lead to seeking medical care. On the other hand, when a symptom appears in association with a chronic stressor, it can motivate care seeking because the problem may be seen as serious (11). Further research is likely to identify a variety of environmental search procedures, such as a review of daily activities to account for limb and leg pain, dietary review to account for gastric distress, and so on.

Social Comparison and Minimization

In a series of elegant experimental studies, Croyle and Jemmott (20) showed that individuals compare themselves to others when judging whether they are or are not at risk for a serious health threat. The subjects, college students, complete what is presumed to be a test for the presence of a newly discovered enzyme indicating abnormal pancreatic activity (the test is supposed to assess the presence of the enzyme in saliva by a color change in a paper that is responsive to sucrose). Subjects are run in groups; some of them receive a positive test (conveniently arranged by sweetening the mouth rinse in which the subject inserts the test strip), others a negative test. Some who receive a positive test are the only ones in their group to test positive (enzyme present), whereas others are one of several receiving a positive test; receiving a positive test is randomly arranged (19). The key data are the subjects' judgments of the seriousness of the disorder. As predicted, subjects who are uniquely afflicted (the only ones in their group to test positive) judge the deficiency to be quite serious, whereas subjects who share the disorder with others judge it to be relatively minor (20). The data are very clear in showing that the social context, i.e., perceiving the threat as shared, reduces its judged severity.

A number of studies report effects in which the social context serves to increase pain tolerance (e.g., ref. 73) and to reduce distress and speed recovery following surgery, e.g., patients whose preoperative roommates were in the postoperative recovery phase following a bypass operation have a speedier recovery (40). The precise mechanisms involved in these latter social comparisons, however, may be different from that studied by Croyle and Jemmott. The minimization of seriousness following a positive test for pancreatic enzyme deficiency appears to involve motivated denial, i.e., subjects who are told the deficiency is untreatable and test positive engage in psychological self-protection by down-rating the severity of the disorder, while those told the deficiency is treatable do not (24). Different processes, however, may be at work in other social contexts. For example, the presence of others that leads athletes to tolerate more intense, noxious stimulation probably reflects a socially elicited need to ``look macho,'' and coronary bypass patients who do better when they room with a patient in postoperative recovery may do so because they are reassured that they too can survive and manage the distress of surgery. Group processes, however, can enhance as well as decrease levels of pain and symptom reporting. This is regularly seen in the treatment of chronic pain patients where social relationships can enhance and sustain symptom and pain reporting (28,82,86). In sum, a variety of procedural mechanisms may operate in the social context leading to the minimization or maximization of symptom reports, depending upon the meaning (representation) of the social context.

EMOTION AND THE SELF-REGULATION PROCESS

As should now be clear, the self-regulation model requires that the question ``How do emotions affect symptom reporting?'' be rephrased to ask ``How do emotions affect the self-regulation process and how do these processes influence symptom reporting?'' Recent investigations focusing on the first of these two questions have examined the relationship between the reporting of somatic symptoms and the reporting of negative emotion in samples of healthy individuals (16,17,41,84). These newer studies made an important advance relative to earlier studies of illness behavior (59,66), somatization (38), depression (37,70), and hypochondriasis (3) by proposing that the association of negative mood reports with symptom reports appears in the general population as well as in the smaller subsamples defined by concepts such as somatization, hypochondriasis, and neuroticism.³

The association between negative affect and symptoms suggests the need for a careful theoretical and empirical analysis of the relationship between symptom reports and measures of both disease and negative feelings. Our second question, ``How do emotions affect the self-regulation process and how do these effects influence symptom reporting?,'' points to a wide range of hypotheses respecting different paths for such influence, and exploration of these paths could produce data that will deepen our understanding of the processes involved in emotion-symptom interactions. This way of framing the emotion-symptom link suggests the following three sets of questions: (i) Are there direct links between emotion and symptom and, if so, are these links symmetrical (64)? In other words, how and when can emotions generate somatic sensations that can be represented as somatic symptoms, and how and when can illness generate somatic sensations that can be represented as emotion? (ii) How and when do emotions affect the interpretation and/or representation of existent somatic sensations and how and when do these changes affect coping procedures and symptom reports? That is, do emotions affect whether sensations are represented as symptoms of illness or as indicators of other causes, e.g., stress, and how do these effects influence symptom reporting and seeking health care? (iii) Do emotions affect appraisal of coping outcomes and how and when does the appraisal of outcome affect emotion, and how do these processes affect symptom reporting? Unfortunately, there is a shortage of sound empirical evidence to address these issues, especially the third, and therefore we will focus on the first two. Given the paucity of data, we will have to resort to speculation more often than we would like, but the shortage of evidence should not deter hypothesis formation if it can generate fruitful research.

Emotion, Illness, and the Generation of Somatic Sensations

At least three pathways exist for direct links between emotion and symptom reports. First, an emotion may generate somatic sensations that are misinterpreted as symptoms of illness, and an illness may create somatic sensations that generate an emotion. Second, by reactivating memories of specific illness episodes, an emotion may reactivate the actual experience of symptoms. Third, an emotion may affect resistance to a pathogen and the occurrence of illness, which in turn leads to symptoms (the so called psychosomatic hypothesis). Each of these possibilities raises questions respecting the mechanism by which an activated emotion affects somatic experience, and the first, which is symmetrical, how an activated illness can generate an emotion.

Misinterpretation of Emotional Sensations and the Reactivation of Symptom Memories

Although we can distinguish between the first and second pathways, i.e., emotionally generated sensations interpreted as signs of disease and emotional reactivation of episodic illness memory, it may be difficult to separate the two in practice. Examples of the first, the interpretation of emotionally produced sensations as signs of disease, are seen in cases in which the activation of the sympathetic, adrenomedullary system, during active episodes of intense anger and anxiety, can generate palpable somatic sensations, e.g., heart racing, face warming, and muscle tension, that mimic coronary events and are interpreted as such (68). The presence of such a path should not be surprising, as intense affective states can precipitate cardiac events in a compromised cardiovascular system (39). This pathway can also be symmetrical, i.e., cardiac symptoms can be misinterpreted as signs of emotional distress and gastrointestinal upset rather than as symptoms of coronary disease (31). These interpretive processes can affect what is included and what is excluded in a symptom report.

The second hypothesis is that active emotional states may activate episodic memories of illness that lead subjects to reexperience somatic sensations of illness. Our initial formulation of this hypothesis (46,51), which posits that episodic memories have pain sensations incorporated within them (47), was influenced by the case studies describing the role of emotional reactions in the activation of phantom pain (60). These case studies suggested that the connection of affect and pain memories is symmetrical. Thus, in addition to activating pain memories, emotional states such as fear and anxiety can lead a pained and injured person to attend to and dwell on his or her injury and its associated pain, facilitating the formation of a pain memory (9). The amputation following this ``rehearsal'' process removes the normal sensory input that could override the pain memory.

Other possible examples of emotional activation enhancing pain report can be seen in studies by Croyle and Uretsky (21) and Salovey and Birnbaum (75). These investigators conducted laboratory studies in which subjects were exposed to emotionally provocative stimuli, e.g., reviewing distressing life episodes, and found an increase in symptom reports in subjects exposed to this emotional activation but no increase in symptom reports in control subjects.

Emotion and the Generation of Disease

The third, psychosomatic, hypothesis suggests that emotional states can be antecedents of disease and disease-produced symptoms. For fairly obvious reasons, this hypothesis has generated considerable excitement among investigators and laypersons as direct effects of this type, if they exist, create hope that disease onset and progression can be halted by exerting control over one's emotional life. The evidence for direct effects is, however, ambiguous (85), although in some areas the evidence is quite robust, e.g., studies of type A behavior pattern suggest that chronic hostility is causally related to the development of hypertension and cardiovascular disease and, therefore, to the appearance of cardiac symptoms (for reviews, see refs. 15, 57, 85).

The extended time frame for the development of hostility-induced cardiovascular disease, however, is less germane to our quest for direct effects of emotion on symptom reports. Short-term effects, e.g., emotional states affecting disease risk and symptom reports within days, weeks, or (at most a few) months, is of greater relevance. The evidence for such associations is ambiguous for three reasons. First, there has been insufficient attention to possible differences in the physiological response patterns for different emotions (26) and the impact of these differences on the physiological processes underlying specific diseases. Second, the physiological response patterns associated with specific emotions may reflect the coping procedures brought into play by the emotions. Thus, the impact of emotion on disease may be due to activity in the coping part of the emotion system (see Figure 1) rather than an emotion itself (22,30,54,85). From the perspective of our self-regulation model, these effects, emotion affecting coping and coping affecting disease, hence symptoms, would define an indirect pathway. Finally, longitudinal data with independent measures of emotion (state and trait), disease, and symptoms will be needed to strengthen statements that emotion can produce symptoms by first producing disease (55,84).

Few studies examining the effects of emotion on symptoms via the disease route have used differentiated measures of emotion and measures of specific diseases and related symptoms. Therefore, it is still unclear whether the data will support the general hypothesis, emotional stress causing disease, or suggest hypotheses relating specific emotions to specific illnesses. In their excellent review of the literature on cardiac events, Kamarck and Jennings (39) suggest that strong arousal of any emotion (anger, fear, or pleasure) may provoke a coronary event in a compromised cardiovascular system. In this instance, therefore, the data support a general hypothesis, as it appears unnecessary to differentiate the negative affects and/or the negative from positive affects in predicting coronary events for coronary compromised individuals.

The usefulness of differentiation, however, is suggested when we examine the relationship of emotional states to the common cold. Confirmatory factor analyses differentiate negative emotional states into their components of anxiety, anger, and depression, and our preliminary analyses using structural modeling, showed that depression, and neither anxiety nor anger, was most likely to be the important factor that could predict increases in symptoms over time (51a). These data supported a differentiation hypothesis, i.e., only depression predicted increased symptoms, even though at each of the five measurement points all three negative affects were correlated with symptoms.

Emotional Effects on Symptom Representation and Coping

We will address three issues in this section. The first concerns the effects of active (emotionally provocative) life stressors on symptom interpretation; the second, the relationship of traits of negative affect to the reporting of symptom changes; the third, the many possible relationships among depression, symptom reporting, and coping.

Stress, Symptom Interpretation, and Coping

The prevailing outlook, ably summarized by Cohen and Williamson (14), suggests that the presence of stressors and negative affect encourages the representation of somatic sensations as disease symptoms and the combination of symptoms and negative affect increases the probability of using medical care (71). However, it is overly simplistic to expect that emotion-generated somatic sensations will necessarily be interpreted and represented as symptoms. Our self-regulation model and the sharing of somatic sensations by disease and emotion suggest a more complex relationship. Specifically, the model suggests that people can recognize that emotionally stressful situations have somatic effects and that, when they are under stress, they may attribute somatic sensations to emotion rather than to illness. As representations direct coping procedures, attributing somatic sensations to emotional stress and discounting them as signs of illness should reduce the likelihood of utilizing medical care.

Both laboratory and field data support the more complex view that stressors may alter the way sensations are interpreted, thereby affecting the procedures used to manage them. For example, in a simple laboratory study, groups of students were asked to imagine what they would think if they awoke the next morning and experienced a set of six specific symptoms (4). They were then asked to rate whether they would judge the symptoms to be signs of stress or signs of illness. Some of the subjects made these judgments after reading six symptoms of diabetes, others after reading six symptoms of mononucleosis, and others after reading a randomly selected set of six symptoms. Half of each of these three groups of subjects made their judgments the day before a midterm examination, a stressful day, whereas the other half did so on a Friday, the day before an exam- and class-free weekend. The data showed that the same somatic signs would be interpreted as indicators of illness, that is, as symptoms, if they appeared on a Saturday morning, and as signs of stress if they appeared the morning of a midterm examination. These effects held for judgments of the randomly selected symptoms and the symptoms of diabetes, as neither set was a clear and/or a familiar picture of an illness for these young subjects. On the other hand, the subjects were familiar with the symptoms of mononucleosis, and judgments of those were unaffected by the day on which the judgment was made. The presence of stress has also been shown to alter the interpretation of symptoms of hypertension, depending on subjects' beliefs respecting whether the disease is in fact caused by stress (4,18).

These laboratory findings conducted under controlled circumstances make clear that the presence of an emotionally provocative stressor can lead somatic sensations to be interpreted as signs of stress rather than signs of illness. We are now preparing a report from a field study showing the very same effect. Specifically, symptoms experienced in the presence of a (nonillness) stressor are likely to be interpreted as signs of stress rather than as signs of illness. They are unlikely to result in care seeking when the stressor is of recent onset, i.e., beginning less than 3 weeks prior to the interview. Rather than report these symptoms to a medical practitioner, subjects dealing with both a recent stressor and a symptom appear to adopt a ``wait and see'' strategy to determine whether the sensation is of medical significance. If the stressor is chronic, its presence will facilitate care seeking as a means of reducing the individual's overall burden of stress. In summary, the data do not support a simple (main effect) relationship among stress, symptom interpretation, and coping. Stress can either enhance or reduce the likelihood of a somatic sensation's being interpreted as a sign of illness and facilitate or reduce the likelihood of coping by using medical care. The outcome depends upon the presence of conditions that allow subjects to attribute or define the somatic sensations as signs of stress and whether the stress is acute (recent) or chronic.

Traits of Negative Affect and Symptom Reporting

Although the evidence is increasingly clear that active states of emotional stress can affect interpretation and coping with somatic sensations, it is less clear that such effects can be found when emotions are indexed by trait measures. The effects can be found when emotions are indexed by trait measures. The consistent association between trait measures of negative emotions and reports of somatic symptoms introduces ambiguities in the interpretation of epidemiological data in which stress is said to predict disease when disease is measured by symptom reports. The trait of negative affect introduces a possible confound in these data that suggests a need for caution in the concluding that stress causes disease, as the data may reflect the presence of the trait of negative affect (83). While the basic finding and need for caution are sound, there is the false implication that symptom reports are little more than a second measure of ``negative affect'' and invalid as measures of disease. There are two reasons to doubt this latter conclusion. The first is methodological. The correlational effects reported in virtually all of the newer studies are from cross-sectional rather than from longitudinal data (16,83). Thus, the relationship of negative affect to symptom reports that is clearly seen in cross-sectional data may be independent of the relationship of negative affect to changes in symptom reports: the former a background association, the latter a dynamic and possibly causal one (84).

The second issue is conceptual; the association of symptom reports with disease is real, but it varies with specific illness. Thus, the less than perfect association is due not only to failure of symptom reports to reflect specific conditions, that is, to deficits in specificity, but to failures in sensitivity; symptom reports are not always sensitive to the presence of disease because diseases are not always symptomatic and/or may be symptomatic at different points in their natural history. Examples of this are cardiovascular disease and many cancers, conditions that do not provoke a symptom-generating immune response and are indeed characteristically asymptomatic until they are sufficiently advanced to disrupt physiological function. Stronger relationships should obtain between symptoms and infectious illness, as the immune response to invading pathogens is a major source of the somatic changes that lead to symptom reporting.⁴

More recent data examining the relationship of trait measures of negative affect to change in symptoms present a different picture. Long-term (10-year time frames) longitudinal studies show virtually no relationships among baseline measures of trait, negative affect, and increments of physical or psychological symptoms (79). Studies recently completed in our laboratory present similar findings. For example, changes in symptoms from a preinoculation baseline to a 24- or 72-hr postinoculation (with flu, tetanus, and neo-antigen inoculations) follow-up measure showed increases in local symptoms (sore arms) for all subjects regardless of their level of trait negative affect (23). All subjects also showed a decrease in reports of general symptomatology from before to after the active inoculants. There were no such changes from before to after three placebo inoculations. The results for change did not affect the cross-sectional findings; subjects high on trait negative affect reported more symptoms. This relationship was statistically significant but small in magnitude. In sum, the data suggest that emotional traits have a static rather than a causal or dynamic relationship to symptom reports.

Depression, Symptom Reporting, and Coping

The data in support of a dynamic, i.e., causal relationship between symptoms and emotion are more favorable when emotion is defined as active states. The initial analyses of our data which suggest that depressed mood states are related to increases in symptom reports over five interviews completed at 3-month intervals (51a), and the absence of similar relationships for anxious and angry moods, implicate affective states, specifically depression, as the emotion likely to be involved in symptom generation. The relationship of depression to changes in symptom reports is of special interest because it is symmetrical, shows a special relationship to pain reports, and can reflect at least two different underlying physiological mechanisms. Symmetry is apparent, as both illness and pain can generate depression as seen in the occurrence of depressive affect during and immediately subsequent to many viral and bacterial infections (2,33) and subsequent to protracted pain from arthritis (8). With respect to the underlying mechanisms, the data suggest that both psychological factors, i.e., uncontrollable stressors such as death of a loved one, and biological factors, i.e., viral and bacterial disease, affect a common somatic system that generates the mental state and somatic symptoms (loss of appetite, withdrawal from overt activity, fatigue, and depressed mood) that can be interpreted as depression or as illness (33,54). When illness is the source of depressed mood, the underlying mechanism appears to involve release of interleukin-1 by the immune system, which leads to physiological changes such as fever and decreased blood iron levels conducive to reducing the rate of pathogen replication. The somatic experiences and behavioral changes accompanying these physiological effects are both felt and interpreted as depression.

Depression has also been implicated in pain from coronary ischemia. Sheps et al. (76) report that ischemic attacks as measured by suppression of the ST wave are more frequently nonpainful or ``silent'' among nondepressed than depressed patients. They hypothesize that the presence of normally elevated levels of beta-endorphin block pain experience in nondepressed patients and the absence of normal levels result in awareness of pain linked to ST-wave suppression in depressed individuals. Unfortunately, the data supporting this hypothesis are somewhat inconsistent (e.g., ref. 34). We suspect the inconsistency arises because the investigators have failed to adequately control the patients' representations and coping strategies to ensure that different emotional states would be activated in their experimental situations for depressed and nondepressed patients.

In summary, studies examining the relationship between depressed emotion and symptoms report positive findings when they examine the relationship between emotional states and symptom experience, and negative findings when they examine the relationship between emotional ``traits'' and symptom reporting. The findings also suggest the need for caution in attributing symptoms to the personality attributes of the patient. Ignoring the symptom complaints of a neurotic or hypochondriacal patient is a risky course of action if one is unfamiliar with the patient's history and prior complaints. Changes in complaints may be as diagnostic for the worrier as for the stoic.

CONCLUSION

The central point of the systems model that has guided our research on symptom reporting is that somatic sensations are embedded in cognitive, affective, and behavioral procedures on the way to becoming symptoms. As we stated earlier, the basis for a parallel view of affective and cognitive processes was formed by our early, social-psychological research on fear as a motive for changing health attitudes and health behaviors (45). This model then served as the stimulus to our later studies on the reduction of the distress but not the sensory component of pain during noxious medical procedures (35,36). These studies showed that preparatory information allowed patients to become familiar with the somatic sensations of noxious experiences and interpret them as benign, and made a distinctive contribution to distress reduction. Instructions and practice in coping procedures, e.g., how to swallow an endoscopic tube (36) and abdominal breathing and pushing during childbirth (44), can add to and interact with the sensory information. The collective results of these studies strongly suggested a substantial degree of independence between the underlying mechanisms for processing the sensory and affective components of stimulus information.

The above conclusion was reinforced by models of the neurophysiological structures involved in the pain system. These models indicate that the sensory and emotional-motivational components of pain are processed independently from the periphery (the distinction between A and C fiber systems) to the central nervous system (the motivational system represented in cingulate cortex, the sensory in the primary and secondary sensory areas [80a]). But although the underlying mechanisms are separable, their products are less so; pain experience typically blends the sensory and affective components, as the underlying mechanisms are activated in unison and are in constant interaction.

REFERENCES

1. Allport G, Pettigrew TF. Cultural influence on the perception of movement: the trapezoidal illusion among Zulus. J Abnorm Soc Psychol 1957;55:104-113.

2. Aneshensel CS, Frerichs RR, Huba GJ. Depression and physical illness: a multiwave, nonrecursive causal model. J Health Soc Behav 1984;25:350-371.

3. Barsky AJ, Klerman GL. Overview: hypochondriasis, bodily complaints, and somatic styles. Am J Psychiatry 1983;140:273-283.

4. Baumann L, Cameron LD, Zimmerman R, Leventhal H. Illness representations and matching labels with symptoms. Health Psychol 1989;8:449-469.

5. Baumann L, Leventhal H. ``I can tell when my blood pressure is up, can't I?'' Health Psychol 1985;4:203-218.

6. Bishop GD, Converse SA. Illness representations: a prototype approach. Health Psychol 1986;5:95-114.

7. Blumhagen D. Hyper-tension: a folk illness with a medical name. Culture Med Psychiatry 1980;4:197-227.

8. Brown GK. A casual analysis of chronic pain and depression. J Abnorm Psychol 1990;99:127-137.

9. Brownlee S, Leventhal H, Balaban M. Autonomic correlates of illness imagery. Psychophysiology 1991;29:1-13.

10. Cacioppo JT, Andersen BL, Turnquist DC, Petty RE. Psychophysiological comparison processes: interpreting cancer symptoms. In: Andersen BL, ed. Women with cancer: psychological perspectives. New York: Springer-Verlag, 1985:141-171.

11. Cameron L, Leventhal EA, Leventhal H. Symptom representations and affect as determinants of care seeking. Health Psychol (in press).

12. Casey KL. Pain: a current review of neural mechanisms. Am Sci 1973;61:194-200.

13. Cohen S, Tyrrell DAJ, Smith AP. Psychological stress and susceptibility to the common cold. N Engl J Med 1991;325:599-605.

14. Cohen S, Williamson GW. Stress and infectious disease in humans. Psychol Bull 1992;109:5-24.

15. Contrada R, Leventhal H, O'Leary A. Personality and health. In: Pervin L, ed. Handbook of personality: theory of research. New York: Guilford, 1990:638-669.

16. Costa PT, McCrae RR. Somatic complaints in males as a function of age and neuroticism: a longitudinal analysis. J Behav Med 1980;3:245-257.

17. Costa PT, McCrae RR. Hypochondriasis, neuroticism, and aging: when are somatic complaints unfounded? Am Psychol 1985;40:19-28.

18. Croyle RT. Biased appraisal of high blood pressure. Prev Med 1990;19:40-44.

19. Croyle RT, Ditto PH. Illness cognition and behavior. An experimental approach. J Behav Med 1990;13:31-52.

20. Croyle RT, Jemmott JB III. Psychological reactions to risk factor testing. In: Skelton JA, Croyle RT, eds. Mental representation in health and illness. New York: Springer-Verlag, 1991:85-107.

21. Croyle RT, Uretsky MB. Effets of mood on self-appraisal of health status. Health Psychol 1987;6:239-254.

22. Davidson RJ. Prolegomenon to the structure of emotion: gleanings from neuropsychology. Cognition Emotion 1992;6:245-268.

23. Diefenbach M, Leventhal H, Patrick-Miller L, Leventhal EA. Affect and symptom reporting under partial control of somatic sensations (in preparation).

24. Ditto PH, Jemmott JB III, Darley JM. Appraising the threat of illness: a mental representational approach. Health Psychol 1988;7:183-201.

25. Drachman RH, Hochbaum GM, Rosenstock IM. A seroepidemiologic study in two cities. In: Rosenstock IM, Hochbaum GM, Leventhal H, eds. The impact of Asian influenza on community life: a study in five cities. Washington, DC: U.S. Government Printing Office, 1960:27-51.

26. Ekman P. Are there basic emotions? Psychol Rev 1992;99:550-553.

27. Fernandez E, Turk DC. Sensory and affective components of pain: separation and synthesis. Psychol Bull 1992;112:205-217.

28. Fordyce WE. Behavioral methods for chronic pain and illness. St. Louis: CV Mosby, 1976.

29. Gonder-Frederick LA, Cox DJ. Symptom perception, symptom beliefs, and blood glucose discrimination in the self-treatment of insulin-dependent diabetes. In: Skelton JA, Croyle RT, eds. Mental representation in health and illness. New York: Springer-Verlag, 1991:220-246.

30. Graham D. Psychosomatic medicine. In: Greenfield NS, Sternbach RA, eds. Handbook of psychophysiology. New York: Holt, Rinehart, and Winston, 1972.

31. Hackett TC, Cassem NH. Factors contributing to delay in responding to the signs and symptoms of acute myocardial infarction. Am J Cardiol 1969;24:651-658.

32. Hansell S, Leventhal H, Leventhal EA, Diefenbach M, Brownlee S. Negative and positive affect, and subjective assessments of health among older adults (in preparation).

33. Hart BL. Biological basis of the behavior of sick animals. Neurosci Behav Rev 1988;12:123-137.

34. Heller GV, Garber CE, Connolly MJ, et al. Plasma beta-endorphin levels in silent myocardial ischemia induced by exercise. Am J Cardiol 1987;59:735-739.

35. Johnson JE. The effects of accurate expectations about sensations on the sensory and distress components of pain. J Pers Soc Psychol 1973;27:261-275.

36. Johnson JE, Leventhal H. Effects of accurate expectations and behavioral instructions on reactions during a noxious medical examination. J Pers Soc Psychol 1974;29:710-718.

37. Katon W, Kleinman A, Rosen G. Depression and somatization: a review, part I. Am J Med 1982;72:127-135.

38. Kellner R. Somatization and hypochondriasis. New York: Praeger, 1986.

39. Kamarck T, Jennings JR. Biobehavioral factors in sudden cardiac death. Psychol Bull 1991;109:42-75.

40. Kulik JA, Mahler HIM. Effects of preoperative roommate assignment on preoperative anxiety and recovery from coronary-bypass surgery. Health Psychol 1987;6:525-543.

41. Larsen RJ. Neuroticism and selective encoding and recall of symptoms: evidence from a combined concurrent-retrospective study. J Pers Soc Psychol 1992;62:480-488.

42. Lau RR, Hartmann KA. Common sense representations of common illnesses. Health Psychol 1983;2:167-185.

43. Leventhal EA, Leventhal H, Schaefer P, Easterling D. Conservation of energy, uncertainty reduction and swift utilization of medical care among the elderly. J Gerontol (in press).

44. Leventhal EA, Leventhal H, Shachman S, Easterling DV. Active coping reduces reports of pain from childbirth. J Consult Clin Psychol 1989;57:365-371.

45. Leventhal H. Findings and theory in the study of fear communications. Adv Exp Soc Psychol 1970;5:119-186.

46. Leventhal H. The integration of emotion and cognition: a view from the perceptual-motor theory of emotion. In: Clark M, Fiske S, eds. Affect and cognition: the 17th annual Carnegie symposium on cognition. Hillsdale, NJ: Erlbaum, 1982:121-156.

47. Leventhal H. Symptom reporting: a focus on process. In: McHugh S, Vallis TM, eds. Illness behavior: a multidisciplinary model. New York: Plenum Press, 1985:219-238.

48. Leventhal H, Diefenbach M. The active side of illness cognition. In: Skelton JA, Croyle RT, eds. Mental representation in health and illness. New York: Springer-Verlag, 1991:247-272.

49. Leventhal H, Diefenbach M, Leventhal EA. Illness cognition: using common sense to understand treatment adherence and affect cognition interaction. Cognitive Ther Res 1992;16:143-163.

50. Leventhal H, Easterling DV, Coons H, Luchterhand C, Love RR. Adaptation to chemotherapy treatments. In: Andersen B, ed. Women with cancer. New York: Springer-Verlag, 1986:172-203.

51. Leventhal H, Everhart D. Emotion, pain, and physical illness. In: Izard CE, ed. Emotions and psychopathology. New York: Plenum Press, 1979:263-299.

52. Leventhal H, Hansell S, Leventhal EA, Diefenbach M, Brownlee S. Negative affect and specific affects as antecedents of symptom reports (in preparation).

53. Leventhal H, Meyer D, Nerenz D. The common sense representation of illness danger. In: Rachman S, ed. Contributions to medical psychology, vol II. New York: Pergamon Press, 1980:7-30.
    1990;20:132-146.

54. Leventhal H, Patrick-Miller L. Emotion and illness: the mind is in the body. In: Lewis M, Haviland J, eds. Handbook of emotion research (in press).

55. Leventhal H, Tomarken AJ. Stress and illness: perspectives from health psychology. In: Kasl V, Cooper CL, eds. Stress and health: issues in research methodology. Sussex, UK: John Wiley, 1987:27-55.

56. Love RR, Wiebe DA, Newcomb PA, et al. Effects of tamoxifen on cardiovascular risk factors in post menopausal women. Ann Intern Med 1991;115:860-864.

57. Matthews KA. Coronary heart disease and type A behavior: update on an alternative to the Booth-Kewley and Friedman quantitative review. Psychol Bull 1988;104:373-380.

58. Matthews KA, Siegel JM, Kuller LH, Thompson M, Varat M. Determinants of decisions to seek medical treatment by patients with acute myocardial infarction symptoms. J Pers Soc Psychol 1983;44:1144-1156.

59. Mechanic D. Illness behavior: an overview. In: McHugh S, Vallis M, eds. Illness behavior: a multidisciplinary model. New York: Plenum Press, 1986:101-110.

60. Melzack R. The puzzle of pain. New York: Basic Books, 1973.

61. Melzack R. Phantom limbs. Sci Am 1992;266:120-126.

62. Meyer D, Leventhal H, Gutmann M. Common-sense models of illness: the example of hypertension. Health Psychol 1985;4:115-135.

63. Nerenz DR, Leventhal H, Love RR. Factors contributing to emotional distress during cancer chemotherapy. Cancer 1982;50:1020-1027.

64. Pennebaker JW. The psychology of physical symptoms. New York: Springer-Verlag, 1982.

65. Pennebaker JW, Watson D. Blood pressure estimation and beliefs among normotensives and hypertensives. Health Psychol 1988;7:309-328.

66. Pilowsky I. Abnormal illness behavior: a review of the concept and its implications. In: McHugh S, Vallis TM, eds. Illness behavior: a multi-disciplinary model. New York: Plenum Press, 1986:391-395.

67. Prohaska TR, Keller ML, Leventhal EA, Leventhal H. Impact of symptoms and aging attribution on emotions and coping. Health Psychol 1987;6:495-514.

68. Rachman S, Levitt K, Lopatka C. Panic: the links between cognitions and bodily symptoms=mI. Behav Res Ther 1987;25:423-441.

69. Riley MW. Aging from birth to death. Washington, DC: American Association for the Advancement of Science, 1979.

70. Rodin G, Voshart K. Depression in the medically ill: an overview. Am J Psychiatry 1986;143:694-705.

71. Roghman KJ, Haggerty RJ. Daily stress and use of health services in young families. Pediatr Res 1973;7:520.

72. Roffwarg HP, Muzio JN, Dement WC. Ontogenetic development of the human sleep-dream cycle. Science 1966;152:604-619.

73. Ryan D, Kovacic CR. Pain tolerance and athletic participation. Percept Mot Skills 1966;22:383-390.

74. Safer M, Tharps Q, Jackson T, Leventhal H. Determinants of three stages of delay in seeking care at a medical clinic. Med Care 1979;17:11-29.

75. Salovey P, Birnbaum D. Influence of mood on health relevant cognitions. J Pers Soc Psychol 1989;57:539-551.

76. Sheps DS, Bragdon EE, Miller PF, Maixner W, Hinderliter AL, Light KC. The perception of pain in patients with myocardial ischemia. J Myocardial Ischemia 1989;1:29-37.

77. Skelton JA, Croyle RT. Mental representation, health, and illness: an introduction. In: Skelton JA, Croyle RT, eds. Mental representation in health and illness. New York: Springer-Verlag, 1991:1-9.

78. Spiegel D, Bloom JR, Kraemer HC, Gottheil E. Effect of psychosocial treatment on survival of patients with metastatic breast cancer. Lancet 1989;2:888-891.

79. Spiro A, Aldwin CM, Levenson MR, Bosse R. Longitudinal findings from the normative aging study: II. Do emotionality and extroversion predict symptom change? J Gerontol 1990;45:136-144.

80. Suchman EA. Stages of illness and medical care. J Health Hum Behav 1965;6:114-128.
    80a

    Talbot JD, Marrett S, Evans AC, Meyer E, Bushnell MC, Duncan GH. Multiple representations of pain in human's cerebral cortex. Science 1991;251;1355-1358.

81. Taylor S. Asymmetrical effects of positive and negative events: the mobilization-minimization hypothesis. Psychol Bull 1991;110:67-85.

82. Turk DC, Meichenbaum D, Genest M. Pain and behavioral medicine: a cognitive-behavioral perspective. New York: Guilford Press, 1983.

83. Watson D, Pennebaker J. Health complaints, stress, and distress: exploring the central role of negative affectivity. Psychol Rev 1989;96:234-254.

84. Watson D, Pennebaker JW. Situational, dispositional, and genetic bases of symptom reporting. In: Skelton JA, Croyle RT, eds. Mental representation in health and illness. New York: Springer-Verlag, 1991:60-84.

85. Weiner H. Specificity and specification: two continuing problems in psychosomatic research. Psychosom Med 1992;54:567-587.

86. Wooley SC, Blackwe ll B, Wingett CW. A learning theory model of chronic illness behavior: theory, treatment, and research. Psychosom Med 1978;40:379-401.