UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Negative affective tendencies predict precursors of metabolic syndrome in physically healthy young women Blackwell, Ekin 2010

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2010_fall_blackwell_ekin.pdf [ 9.75MB ]
Metadata
JSON: 24-1.0069941.json
JSON-LD: 24-1.0069941-ld.json
RDF/XML (Pretty): 24-1.0069941-rdf.xml
RDF/JSON: 24-1.0069941-rdf.json
Turtle: 24-1.0069941-turtle.txt
N-Triples: 24-1.0069941-rdf-ntriples.txt
Original Record: 24-1.0069941-source.json
Full Text
24-1.0069941-fulltext.txt
Citation
24-1.0069941.ris

Full Text

NEGATIVE AFFECTIVE TENDENCIES PREDICT PRECURSORS OF METABOLIC SYNDROME IN PHYSICALLY HEALTHY YOUNG WOMEN  by  Ekin Blackwell  B.Sc., Technical University of Istanbul, 1989 M.A.Sc., Dalhousie University, 1992 M.A., University of British Columbia, 2003  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in The Faculty of Graduate Studies (Psychology)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) April 2010 © Ekin Blackwell, 2010  ii ABSTRACT Accumulating evidence indicates that negative affect (NA) and positive affect (PA) are linked to a variety of physical health outcomes. However, there are a number of significant gaps in this literature. First of all, many commonly employed instruments assess broad psychological constructs rather than basic mood states. Second, these measures do not distinguish between valence and activation, two theoretically important dimensions of affect. This makes it difficult to sort out which key ingredients are involved in the affect and health connection. Third, with all the studies of medical patients, it is not evident whether moods matter causally or if they are confounded with disease severity. Finally, the biological mechanisms through which different moods might influence health are largely unknown. A critical question is whether NA and PA work through different pathways. The current research project was designed to help address these problems. The study employed a circumplex model of affect to track patterns of mood in physically healthy young women over the course of 6 months. Biological measures representing preclinical markers of metabolic disease were obtained at the end of this period. This methodology enabled us to examine the independent and joint contributions of different affective tendencies, varying on dimensions of valence and activation, to biological mechanisms involved in disease development. The findings revealed that the tendency to experience chronic negative emotions, regardless of their associated level of activation, predicted less optimal metabolic symptoms. In contrast, positive affective tendencies had neither a favorable nor detrimental effect on these outcomes. However, we found no evidence that NA and PA have unique biological correlates. The implications of these findings for models of emotion and health are discussed, and several potential avenues for future research are suggested.  iii TABLE OF CONTENTS  ABSTRACT………………………………………………………………………………………ii TABLE OF CONTENTS………………………………………………………………………. iii LIST OF TABLES……………………………………………………………………………..... v LIST OF FIGURES…………………………………………………………………………….. vi ACKNOWLEDGEMENTS…………………………………………………………………….vii INTRODUCTION………………………………………………………………………………. 1 Negative Affect and Health…………………………………………………………………………... 3 Depression…………………………………………………………………………………………... 3 Anxiety…………………………………………………………………………………………........7 Anger/Hostility……………………………………………………………………………………… 8 Other Negative Moods…………………………………………………………………………….. 10 Positive Affect and Health………………………………………………………………………….. 14 Positive Affect and Mortality……………………………………………………………………… 15 Positive Affect and Morbidity……………………………………………………………………... 17 Summary of Limitations of Previous Work……………………………………………………….. 19 Theories of Affect……………………………………………………………………………………. 23 Negative Affect…………………………………………………………………………………….. 23 Positive Affect……………………………………………………………………………………... 23 Valence and Arousal: Two Core Dimensions of Affect…………………………………………... 24 Are NA and PA Independent? ……………………………………………………………………. 26 Summary of Current Study………………………………………………………………………… 28 Hypotheses…………………………………………………………………………………………... 30 Relative Contributions of NA and PA…………………………………………………………….. .30 The Role of Activation…………………………………………………………………………….. 30 NA-specific and PA-specific Pathways…………………………………………………………… 31 The Role of Affective Variability…………………………………………………………………. 32  METHOD………………………………………………………………………………………. 34 Participants………………………………………………………………………………………….. 34 Procedure……………………………………………………………………………………………. 35 Mood Diary……………………………………………………………………..…………………… 37 Biological Outcome Measures……………………………………………………………………… 37 Systolic and Diastolic Blood Pressure………………………………………………………..…… 37 Waist-hip Ratio……………………………………………………………………………………. 38 Glucose………………………………………………………………………………………..…… 38 Insulin……………………………………………………………………………………………… 38 High Density Lipoprotein (HDL) ………………………………………………………………… .38 Triglycerides…………………………………………………………………………………….… .38  iv Control Measures………………………………………………………………………………….... 38  RESULTS…………………………………………………………………………………...….. 40 Comparison of Participants vs. Non-participants………………………………………………… 40 Pre-Analysis Data Screening………………………………………………………………..……… 40 Outliers…………………………………………………………………………………………….. 40 Missing Data…………………………………………………………………………………….… 40 Examination of Residual Scatterplots………………………………………………………….….. 41 Data Reduction……………………………………………………………………………………… 42 Creation of Composite Biological Outcome Score………………………………………………... 42 Creation of Composite Mood Variables……………………………………………………….…... 43 Descriptive Findings…………………………………………………………………………..…….. 44 Bivariate Correlations Among Composite Mood Variables……………………………….……... 45 Bivariate Correlations Among Biological Measures……………………………………………… 46 Tests of Primary Hypotheses………………………………………………………………..……… 47 Relative Contributions of NA and PA……………………………………………………..………. 49 The Role of Activation…………………………………………………………………………….. 54 NA-specific and PA-specific Pathways…………………………………………………………… 58 The Role of Affective Variability………………………………………………………….……… 67 Secondary Analyses…………………………………………………………………….…………… 68 Presence of Mood Disorder………………………………………………….…………………….. 68 Influence of Number of Observations…………………………………………………………...… 68  DISCUSSION………………………………………………………………………………...… 70 Global NA Effects…………………………………………………………………………………… 71 Global PA Effects…………………………………………………………………………………… .74 Activation Effects for NA…………………………………………………………………...……… .78 Activation Effects for PA…………………………………………………………………………… 82 Evidence for Valence-Specific Pathways………………………………………………...………… 84 Variability in Affect Over Time……………………………………………………………….…… 89 Caveats Regarding Circumplex Structure of Mood……………………………………………… 90 Limitations and Future Directions………………………………………………………………… 91  CONCLUSION………………………………………………………………………………….96 REFERENCES………………………………………………………………………………….97 APPENDIX A: Ethical Approval……………………………………………………………. 119  v LIST OF TABLES Table 1  Descriptive Statistics for Affective Predictors and Biological Outcome Measures………………………………………………………..44  Table 2  Bivariate Correlations Among Composite Mood Variables………….………………48  Table 3  Bivariate Correlations Among Biological Measures………………………................ 49  Table 4  Hierarchical Multiple Regressions Predicting the Composite Biological Outcome from Global NA and Global PA................................................................................... 51  Table 5  Hierarchical Multiple Regressions Predicting the Composite Biological Outcome from Negative Mood Vectors Differing in Degree of Activation………………….....55  Table 6  Hierarchical Multiple Regressions Predicting the Composite Biological Outcome from Positive Mood Vectors Differing in Degree of Activation…………...………... 57  Table 7  Hierarchical Multiple Regressions Predicting Individual Biological Outcomes from Global NA and Global PA……………………………………………………... 59  Table 8  Hierarchical Multiple Regressions Predicting Systolic Blood Pressure from Negative Mood Vectors Differing in Degree of Activation ….………………...63  Table 9  Hierarchical Multiple Regressions Predicting Diastolic Blood Pressure from Negative Mood Vectors Differing in Degree of Activation…………………….64  Table 10 Hierarchical Multiple Regressions Predicting Glucose from Negative Mood Vectors Differing in Degree of Activation……..……………...66  vi LIST OF FIGURES Figure 1 Mood Circumplex ………………………….....…………………………………....... 27 Figure 2 Scatterplot of Global NA and Metabolic Symptom Scores………………………….. 53 Figure 3 Scatterplot of Global PA and Metabolic Symptom Scores………………………….. 53  vii ACKNOWLEDGEMENTS I would like to express my gratitude to the many individuals who have supported me through my doctoral studies. To my research supervisor, Gregory Miller, thank you for encouraging me to continuously challenge myself and think in unconventional ways. I consider myself fortunate to have benefited from your mentorship. My heartfelt appreciation also goes out to Jeremy Biesanz for so generously dedicating his time and expertise during the data analysis phase of this project. I would also like to thank the other members of my core committee (Liz Dunn and Scott Carlson) for their constructive feedback and valuable insights at various stages of my dissertation. I would like to acknowledge our lab coordinator Tara Martin for her help in setting up the web survey, overseeing the data collection process, and tolerating my endless requests. I also wish to extend my sincere gratitude to Ralph Hakstian, who has acted as an informal adviser to me since I first decided to embark on my new career path nearly a decade ago. To all my friends and family, I am indebted to you for your endless support, patience, and understanding during my long journey in this program. I am especially grateful to my husband Shawn, whose confidence in me has often exceeded my own. Finally, I am grateful for the love and affection of my two vibrant little daughters, Sage and Tala, who remind me every day what is truly important in life.  1 INTRODUCTION For centuries, people have believed that mood can impact physical health. The common conception is that negative moods are bad for one’s health whereas positive moods are healthenhancing. A substantial body of empirical evidence now exists in support of this idea. Although the bulk of this research has focused on constructs of negative affect (NA), accumulating evidence also suggests a protective role for positive affect (PA). Prospective studies that look at objective markers of disease provide the strongest evidence because they minimize the possibility that health status influences mood, and avoid the problem of mood being confounded with subjective health outcomes. In the NA domain, depression, anger/hostility, and anxiety have been identified as key risk factors for disease. Depression has been implicated in the long-term development of many medical conditions in initially healthy adults - chiefly cardiovascular disease, diabetes, and cancer. Evidence also suggests that depression predicts poorer prognosis in patients with HIV, cancer, and diabetes; and may trigger recurrent coronary events in patients with advanced cardiovascular disease. Dispositional anger and hostility show reliable associations with cardiovascular disease and all-cause mortality in adults with no known prior disease, and recent evidence suggests that these traits may also heighten risk for stroke and type 2 diabetes. Although anxiety has been less well studied, there is some evidence that it promotes the development of cardiovascular disease in initially healthy persons (Kubzansky & Kawachi, 2000; Suls & Bunde, 2005). Although there has been a growing interest in the link between PA and well-being, few studies have documented links between PA and objective indicators of disease. Existing evidence generally supports the notion that PA is associated with better health outcomes and  2 greater longevity (e.g. Kawamoto & Doi, 2002; Maier & Smith, 1999; Ostir, Markides, Black, & Goodwin, 2000), yet there are also notable exceptions. Specifically, PA may worsen prognosis in very ill patients and the institutionalized elderly (Janoff-Bulman & Marshall, 1982; Stones, Dornan, & Kozma, 1989). In healthy young persons, extreme PA has been shown to predict a shorter life span (Friedman, Tucker, Tomlinson-Keasey, Schwartz, Wingard, & Criqui, 1993). Despite significant advances in our understanding of the affect and health link, there are several lingering problems with the evidence. To begin with, many commonly employed measures of affect are too broad, making it difficult to sort out whether the effects are due to mood or related cognitive or behavioral processes. Second, these measures do not distinguish between valence and activation, two theoretically important dimensions of affect. Valence refers to the degree of pleasantness (or unpleasantness) of a mood state, whereas activation refers to its level of physiological arousal, as experienced subjectively. It is thus unclear whether the effects are due to NA or PA, valence or activation, or if these components act independently or jointly. Third, with all the studies of medical patients, it is not evident whether affect matters causally or if it is confounded with disease severity. Finally, even if affect does act causally, our understanding of plausible mechanisms is limited. One prominent question is whether NA and PA work through different pathways. This research project was designed to overcome these problems. The study employed a theory-based dimensional measure of affect to track patterns of mood in physically healthy adolescent females over a 6-month period. Biological measures representing preclinical markers of disease were obtained at 6-month follow-up. This approach afforded the opportunity to analyze the independent and joint contributions of different affective tendencies, varying on dimensions of valence and activation, to biological mechanisms involved in disease  3 development. Because the sample was comprised of healthy young women, the likelihood of confounding by disease was minimal. Negative Affect and Health Considerable attention has been devoted to the role of NA in the onset and progression of disease, and risk for premature death. The literature is vast, and several qualitative and quantitative reviews have been conducted on this topic. Research in this area has converged on hostility, depression, and anxiety as the key constructs in this association. Although cardiovascular disease has been the most widely studied health outcome, links to other diseases such as type 2 diabetes, cancer, and infectious diseases (e.g. HIV, viral colds) have also been reported. Depression. Numerous prospective investigations have examined the relationship between depression and disease onset or progression, particularly in the cardiovascular realm. Researchers typically conceive depression as either the presence of a depressive disorder based on diagnostic criteria, or ask participants to rate their depressive symptoms on a continuum with standardized scales such as the Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961) or the Center for Epidemiological Studies – Depression Scale (CES-D; Radloff, 1977). As a psychological construct, depression encompasses affective (e.g. sadness), cognitive (e.g. worthlessness), behavioral (e.g. social isolation), and physical (e.g. fatigue, insomnia) symptoms. Some researchers also measure concepts that are closely related to depression, such as hopelessness, pessimism, and vital exhaustion. The consensus from reviews of studies with coronary heart disease (CHD) as the endpoint is that depression is a strong risk factor for CHD morbidity and mortality, controlling for well-known risk factors such as smoking, high blood pressure, obesity, and cholesterol. A  4 meta-analysis based on 11 prospective studies of initially healthy individuals reported an overall relative risk ratio (RR) of 1.64 (Rugulies, 2002), concluding that evidence favoring a link between depression and CHD onset is strong. In an earlier qualitative review, Kubzansky and Kawachi (2000) also speculated that depression may be an antecedent to CHD, but cautioned that further work is needed to support this hypothesis. Both reviews only included studies that assessed clinical depression or depressive mood. Studies that measured related concepts like vital exhaustion or hopelessness were not considered. In a more recent review of the literature on NA and cardiovascular disease, Suls and Bunde (2005) identified 76 prospective studies that assessed depression as a risk factor. These reviewers used a less stringent definition for depression, also including studies that measured concepts like hopelessness and vital exhaustion. Of these studies, 24 used initially healthy samples, and 56 used samples with pre-existing CHD. A link between depression and CHD was most evident in physically healthy people, with only 2 studies reporting null findings and 7 reporting marginally significant effects. In contrast, a mixture of positive and negative findings were reported in studies of samples with known CHD at baseline, leading Suls and Bunde (2005) to conclude that the evidence from these samples is inconclusive. However, others have speculated that depression may be an especially strong risk factor for cardiac death in patients with advanced coronary artery disease (CAD), especially following a recent myocardial infarction (MI) (e.g Carney & Freedland, 2003). According to this view, depression acts as an acute trigger of clinical symptoms in vulnerable patients. There is also evidence that depression may influence the onset and progression of other diseases. Spiegel and Giese-Davis (2003) reviewed 11 studies examining depression in relation to cancer incidence, progression, and mortality, concluding that evidence was strongest for a link  5 between sustained depression after a diagnosis and cancer progression or mortality. Once again, this is a case where depression may act as a short-term trigger in vulnerable patients. However, they also acknowledge that some symptoms of cancer could mimic symptoms of depression. In other words, rather than depression influencing cancer mortality, more advanced disease may worsen depressive symptoms. The authors also note that several studies assessed depression only once at baseline in order to predict cancer incidence several years later. Because one-time measurements may reflect transient mood changes that later resolve, it is unlikely that they would reliably predict future disease incidence. There is also evidence that depression plays a role in HIV disease progression, defined as time to death, time to AIDS diagnosis, or alterations in immune markers such as CD4 and natural killer (NK) cells. Leserman (2003) reviewed evidence based on 14 longitudinal studies, 9 of which reported positive findings. For example, in a 7-year prospective study of 402 HIVinfected gay men, those with high depressive symptoms at every visit had a 1.7 times greater chance of death compared to those who never displayed elevated depressive symptoms (Mayne, Vittinghoff, Chesney, Barrett, & Coates, 1996). A 5.5-year follow-up study of 277 gay men did not find an association between baseline depression and AIDS or mortality, but did find that depression predicted more rapid declines in CD4 counts (Burrack, Barrett, Stall, Chesney, Ekstrand, & Coates, 1993). Leserman and colleagues conducted a series of studies based on a cohort of 96 initially asymptomatic HIV-infected gay men. Depressive symptoms predicted greater declines in CD16+ NK cells, CD56+ NK cells, and CD8+ T cells over 2 years (Leserman, Petitto, Perkins, Folds, Golden, & Evans, 1997). Furthermore, higher cumulative depression scores increased the risk of AIDS over 5.5 years (Leserman, Jackson, Petitto, Golden, Silva, Perkins, et al., 1999) as well as the risk of developing an AIDS related clinical condition over 9  6 years (Leserman, Petitto, Gu, Gaynes, Barroso, Golden, et al., 2002). According to Leserman, the most reliable evidence comes from studies that had long follow-up periods (2 years or more), and assessed depression as a cumulative average across multiple time points. Short-term studies have tended to report null findings (e.g. Rabkin, Williams, Remien, Goetz, Dertzner, & Gorman, 1991; Vedhara, Schifitto, & McDermott, 1999). However, participants in the Vedhara et al. (1999) study were in the advanced stage of their illnesses, so it is possible that depression is less important at this stage. A small but growing database suggests that depression may also increase subsequent risk of type 2 diabetes, even after adjusting for traditional risk factors such as age, race, and body weight. Two large-scale prospective studies have reported that depression more than doubles the risk of developing type 2 diabetes in nominally healthy samples (Eaton, Armenian, Gallo, Pratt, & Ford, 1996; Kawakami, Tkatsuka, Shimuzu, & Ishibashi, 1999). Another study compared people with a history of depression and never-depressed controls, selected from a general practice database. Both groups were followed for a diagnosis of type 2 diabetes. Although no overall relation was detected, depressed males under 50 had a 78% greater risk of developing type 2 diabetes (van den Akker, Schuurman, Metsemakers, & Buntinx, 2004). According to a meta-analysis of 24 studies, depression is also associated with poor glucose control in both type 1 and type 2 diabetes, with effect sizes ranging from small to moderate (de Groot, Anderson, Freedland, Clouse, 0& Lustman, 2001). Another meta-analysis of 27 studies reported similar associations between depression and various complications in patients with type 1 and type 2 diabetes (Lustman, Anderson, Freedland, de Groot, Carney, & Clouse, 2000). Unfortunately, the direction of these relationships could not be established in either case because the studies were cross-sectional.  7 All in all, there is evidence that depression may influence a variety of short-term and long-term disease outcomes, but several negative findings have also been reported. These inconsistencies may reflect the heterogeneous nature of depression. For some individuals, depression is a transient response to short-term life stress, whereas for others, it is a chronic and debilitating disorder. Temporary bouts of depression may have no long-term health consequences for initially healthy individuals, but may worsen clinical outcomes in susceptible patients. On the other hand, depression may serve as a long-term promoter of disease processes in those who experience chronic depressive symptoms. It is also possible that certain profiles of symptoms, such as those involving PA or NA, explain links with morbidity and mortality. Variability in results may thus reflect differences in symptom profiles across patients. Studies that unpack these constructs could help to pinpoint which aspects of depression are critical to health. Anxiety. Accumulating evidence suggests that anxiety may be a potent risk factor for cardiovascular disease. Like depression, anxiety is sometimes assessed as a form of psychopathology (e.g. generalized anxiety disorder, phobia, or panic disorder) using diagnostic interviews, and sometimes measured as the tendency to experience chronic fear, worry, or apprehension using self-report rating scales. These measures contain a mixture of affective (e.g. anxious feelings), cognitive (e.g. worry), behavioral (e.g. avoidance), and physical (e.g. racing heart) components. Anxiety has been most frequently examined in relation to coronary-related problems. Suls & Bunde (2005) reviewed 12 prospective studies of samples free of CHD at baseline, and 17 prospective studies of samples with established CHD at baseline. Only those studies that used hard health outcomes such as mortality, incident CHD, recurrent cardiac events such as a new MI, or sudden cardiac death, were included. Follow-up periods ranged from 2  8 years to 35 years. Collectively, evidence from healthy samples suggested a role for anxiety in the development of cardiovascular disease, with RRs ratios ranging from 2.4 to 7.8, after adjusting for relevant demographic, medical, and behavioral risk factors. However, there was little evidence for a positive relationship between anxiety and cardiac death or disease exacerbation in high-risk samples. A similar conclusion was reached by Kubzansky and Kawachi (2000) in an earlier review. It thus appears that anxiety acts as a long-term promoter of the atherosclerotic process rather than a short-term trigger of cardiac events in vulnerable patients. These authors further suggested that among anxiety, anger, and depression, evidence for a relationship to CHD onset was strongest for anxiety. A temporal association between anxiety and CHD has been demonstrated in some recent studies as well. For example, in a report based on the Framingham Offspring Study, tension predicted 10-year incident CHD, total mortality, and development of arterial fibrillation in men; anxiety was identified as a risk factor for total mortality in both men and women. Relative risk ratios after adjusting for relevant biomedical and behavioral risk factors were in the range of 1.25 (Eaker, Sullivan, Kelly-Hayes, D’Agostino, & Benjamin, 2005). Another study of a large cohort selected from a managed care database compared the prospective risk of CHD in patients with and without panic disorder (PD). Those with pre-existing cardiovascular disease were excluded from the study. After adjusting for several demographic and behavioral risk factors as well as use of medications, the PD group was found to have an 87% greater risk of developing CHD compared to the non-PD group. In PD patients with comorbid depression, the risk was 2.5 times greater (Caminero, Blumentals, Russo, Brown, & Castilla-Puentes, 2005). Anger/Hostility. Anger, anger expression and hostility are three closely related dimensions of the anger complex. Hostility is a cognitive style defined by cynical and  9 antagonistic attitudes towards others. The most common measure of hostility is the CookMedley Hostility (Ho) scale (Cook & Medley, 1954). Despite its popularity, this measure has been criticized for its poor internal reliability and heterogeneous item content. More recent instruments measure behavioral ratings of hostility based on structured interviews such as the Interpersonal Hostility Assessment Technique (IHAT) (Haney, Maynard, Houseworth, Scherwitz, Williams, & Barefoot, 1996). Anger refers to an unpleasant emotional state, which can range from mild annoyance to full-blown rage. Trait anger refers to the tendency to experience chronically high levels of this emotion. The Spielberger Trait Anger Scale (Spielberger, Jacobs, Russell, & Crane, 1983) is the most popular measure of this dimension and has good psychometric properties. Finally, anger style is the tendency to either verbally or physically express angry feelings (anger-out), or to inhibit these feelings (anger-in). The Spielberger Anger Expression Questionnaire (Spielberger, Johnson, Russell, Crane, Jacobs, & Worden, 1985) is frequently used to measure these anger coping styles. In general, however, assessment instruments tend to blur distinctions between hostility, anger, and anger expression, and other conceptually similar constructs such as aggression and dominance. In a comprehensive meta-analysis of 45 studies, hostility was identified as an independent risk factor for CHD and all-cause mortality (Miller, Smith, Turner, Guijarro, & Hallet, 1996). Suls & Bunde (2005) reached a similar conclusion in their qualitative review of 38 prospective studies examining the relationship between dimensions of the anger complex and CHD. Of these studies, 23 relied on initially healthy samples and 15 relied on samples with established CHD at baseline. These authors concluded that on the whole, evidence for heightened CHD risk was strongest for hostility and anger expression, with RRs ranging from 1.4 to 9.60 for hostility, and 1.46 to 6.4 for anger expression in healthy populations. Evidence  10 for a link between trait anger and CHD risk was less reliable, although there were comparatively fewer studies in this category. A study of over 23,000 male health professionals, which was not included in this review, suggests a more intricate association between anger and cardiovascular disease. Anger expression did not predict risk for coronary or cardiovascular disease in the overall sample, but emerged as a protective factor for those under 65. On the other hand, more frequent feelings of self-reported anger increased risk for recurrent problems in men with preexisting coronary disease (Eng, Fitzmaurice, Kubzansky, Rimm, & Kawachi, 2003). A few studies have also linked anger and hostility to other disease outcomes. For example, Everson and colleagues found that expressed anger (anger-out) doubled the risk of a first-time stroke over 8 years (Everson, Kaplan, Goldberg, Lakka, Sivenius, & Salonen, 1999), adjusting for several other risk factors. However, higher anger-in scores were unrelated to this risk. In two separate studies based on a community sample of over 12,000 initially healthy black and white adults, anger was associated with a two-fold increase in risk for stroke over 6 years (Williams, Nieto, Sanford, Couper, & Tyroler, 2002), and a 31% increase in risk for type 2 diabetes over 6 years (Golden, Williams, Ford, Yeh, Sanford, Nieto, et al., 2006). Other Negative Moods. Very few studies have examined the impact of NA on objective health outcomes using more pure measures of affect. This is cause for concern, as the constructs assessed are so broad that it is difficult to discern what the critical components are. In one study that utilized a more narrow definition of NA, healthy participants were exposed to a cold virus. Prior to exposure, they rated their state and trait levels of NA using 12 adjectives from the Profile of Mood States scale (POMS; McNair, Lorr, & Droppleman, 1971). State and trait NA were independently associated with subjective cold symptoms, however only state NA was associated with an objective marker of illness (mucus weights) (Cohen, Doyle, Skoner, Fireman,  11 Gwaltney, & Newsom, 1995). These findings suggest that state NA fosters actual disease processes whereas trait NA increases symptom reporting. However, the fact that trait NA was not correlated with pre-exposure symptoms indicated that the latter effect was due to heightened sensitivity to illness rather than a hypochondriacal response. Overall, existing evidence suggests that NA may contribute to disease onset in initially healthy people. Results from patient samples are somewhat mixed, although there is good evidence that depression can worsen clinical outcomes in patients with advanced CHD. Some have suggested that weak or negative results from patient samples may be a consequence of range restriction (e.g. Miller et al., 1996). Other factors further complicate interpretation of findings based on persons with established disease. Disease symptoms may be indistinguishable from NA, particularly during hospitalization (Suls & Bunde, 2005). Somatic symptoms of diagnostic assessment tools often overlap with disease symptoms. For example, some symptoms of panic or anxiety mimic features of cardiovascular disease. Significant overlap between symptoms of depression and HIV has also been documented (Kalichman, Sikkema, & Somlai, 1995). Furthermore, what is thought to be chronic NA may actually reflect the normal and transient emotional response to being hospitalized or learning about a diagnosis. Recently, a number of researchers have questioned whether depression, anxiety and anger/hostility uniquely contribute to disease risk, or whether disease-proneness stems from a general disposition towards experiencing aversive emotions (e.g. Suls & Bunde, 2005). This question deserves further scrutiny as reports linking all of these constructs to cardiovascular disease continue to accumulate. Most researchers have examined these affective dispositions in isolation, ignoring their substantial conceptual and measurement overlap (Suls & Bunde, 2005). Measurement overlap refers to the high correlations among instruments. Conceptual overlap  12 refers to the psychological relatedness among theoretical constructs. For example, depression and anxiety have many shared symptoms, and comorbidity between the two disorders is high. Anger and hostility are also related to anxiety and depression in meaningful ways. For instance, depression often manifests as irritability. Hierarchical models of affect indicate that all of these constructs can be subsumed under a common negative affectivity factor (Clark & Watson, 1991). The substantial overlap among negative affective constructs calls into question the assumption that each has unique and independent effects on health. Failure to consider this overlap makes it impossible to identify whether individual affective states or a core set of features common to all are responsible for relationships with health outcomes. These affective constructs may also have additive effects, such as the combination of depression and anger being more potent than depression alone. Studying each construct individually may lead to underestimation of effects. The issue of construct overlap has been recognized to some extent in studies of patients. For example, in a study led by Mendes de Leon, researchers examined trait anger and vital exhaustion in relation to cardiac events in patients undergoing angioplasty, and found that both made independent contributions (Mendes de Leon, Kop, Stuart, Bar, & Appels, 1996). A study of post-MI patients examining anger expression, depression, and anxiety found that only depression predicted cardiac-related morbidity and mortality (Ahern, Gorkin, Anderson, Tierney, Hallstrom, Ewart, et al., 1990). Two other studies of post-MI by FrasureSmith et al. (1995; 2003) also warrant attention. The first study examined the relative importance of depression, anxiety, anger-in, and anger-out in predicting cardiac events in hospitalized patients over 12 months. The impact of depressive symptoms, history of major depression, and anxiety on recurrent cardiac events were independent of each other and of  13 cardiac disease severity (Frasure-Smith, Lesperance, & Talajic, 1995). In a 5-year follow-up study of a larger sample of MI patients, the same research team examined several psychological measures of depression, anxiety, anger, psychological distress, and social support in relation to cardiac mortality (Frasure-Smith & Lesperance, 2003). Factor analysis revealed 3 underlying factors: NA, overt anger, and social support. Examined separately, depression, anxiety, and distress were all predictive of 5-year mortality, although only depression remained a significant predictor after adjusting for cardiac disease severity. When examined in conjunction, taking into account the shared NA factor, only depression and NA remained significant predictors. In the only study that considered the issue of construct overlap in healthy individuals, researchers tested the unique vs. shared predictive strength of anxiety, depression, anger, and general distress in relation to 10-year incident CHD. When considered separately, all four measures contributed to the prediction. However, when considered together, only anxiety and general distress emerged as significant predictors (Kubzansky, Cole, Kawachi, Vokonas, & Sparrow, 2006). Suls & Bunde (2005) have highlighted the need for a systematic set of studies that concurrently test the effects of depression, anxiety, and anger/hostility on risk for disease in initially healthy samples. This may be the only way to establish which construct is the culprit, and to compare the relative importance of their independent, shared, and/or additive effects. The current proposal rests on the assumption that it is the affective aspect of these constructs that drive changes in health. According to this position, the action is not with broad mood states like depression, anxiety, and hostility, but rather with the fundamental affective constituents of these states, like PA, NA and arousal. Although it is possible that the behavioral or cognitive aspects of these constructs are responsible for changes in health, existing evidence favors the view that the affective component is the driving force behind this association.  14 Positive Affect and Health In contrast to the extensive literature on the relationship between NA and clinical markers of disease, relatively few studies have examined PA in relation to clinical health outcomes. In a thorough qualitative review of this literature, Pressman and Cohen (2005) highlight several consistent patterns, and identify a number of unresolved issues. Since the publication of this seminal article, two more reviews have assessed this literature using metaanalytic methods. The first one by Howell, Kern, & Lyubomirsky (2007) concluded that PA was associated with several short- and long-term outcomes, including markers of normal functioning. The second study by Chida & Steptoe (2008) focused on 70 studies investigating relations between PA and morbidity in healthy and disease populations. Their analysis revealed favorable effects of PA on longevity. Specifically, PA was associated with reduced cardiovascular and allcause mortality in healthy populations, and with reduced death rates in patients with renal failure and HIV. In the remainder of this section, we draw on Pressman and Cohen’s (2005) earlier review because it provides a more fine-grained analysis of issues that are relevant to the present study. Unlike studies of NA, which have focused on depression, anxiety, and anger/hostility as separate constructs, studies of PA have tended not to draw clear boundaries between different kinds of positive moods. This has led to substantial variability in the way PA has been measured. One line of work has focused on the link between broad psychological constructs like optimism, self-esteem, well-being, or life satisfaction. These broad constructs have both affective and cognitive features, which researchers have not typically differentiated in relation to health outcomes. Among studies that used more narrowly defined measures of PA, some  15 assessed single mood states like happiness whereas others utilized multiple-item adjective scales. Finally, a few studies have used observer-rated indices of PA such as laughter or excitement. Positive Affect and Mortality. Mortality studies typically focus on older populations. These studies track individuals who are healthy at study onset or control for existing medical conditions. Sometimes, elderly individuals are recruited from nursing homes or other institutional settings. In a majority of these studies, PA was assessed as a trait, with happiness, well-being, and life satisfaction being the most commonly measured traits. Several prospective studies of adults from the general population have reported that PA is associated with greater longevity. Most studies in this category focused on noninstitutionalized older persons, and all found that higher trait PA predicted increased longevity (e.g. Kawamoto & Doi, 2002; Levy, Slade, Kunkel, & Kasl, 2002; Maier & Smith, 1999; Ostir, et al., 2000; Parker, Thorslund, & Nordstrom, 1992). Similar findings have been reported for interviewer-rated life satisfaction and happiness in two separate studies of individuals over 60, who were followed for 15 and 2 years, respectively (Palmore, 1969; Zuckerman, Kasl, & Ostfeld, 1984). Another study tracked over 22,000 Finnish twins, aged 18-64 at onset, for 20 years. The trait PA measure was an index of life satisfaction, which included items such as interest in life, happiness, and loneliness. Lower life satisfaction predicted increased mortality due to disease, injury, and suicide (Koivumaa-Honkanen, Honkanen, Viinamaeki, Heikkila, Kaprio, & Koskenvuo, 2000). However, an earlier study of nearly 7000 community members within an even broader age range found no effect of trait PA, measured by general positive morale and feelings of happiness (Kaplan & Camacho, 1983). In an intriguing study by Danner, Snowdon, and Friesen (2001), the autobiographies of 180 nuns (mean age = 22) entering a convent were examined for positive and negative emotion content. They were then followed up  16 for a period of about 60 years. Greater positive emotion content was associated with greater longevity, whereas negative emotions were not related to longevity. Although a majority of studies based on older community samples have found that higher trait PA predicts greater longevity, a number of well-designed studies have found the reverse. In a 65-year prospective study of a large sample of gifted children (mean age = 11), higher levels of parent- and teacher-rated cheerfulness and optimism, predicted decreased longevity (Friedman et al., 1993). However, this risk pertained only to those who scored at the extreme end of the positivity scale. In a study of 30 nursing home residents, higher baseline levels of well-being, assessed as a combination of happiness, interest, disappointment, and life satisfaction, predicted decreased longevity over 2.5 years after adjusting for baseline health status (Janoff-Bulman & Marshall, 1982). A similar study of 156 institutionalized elderly found that subjective ratings of happiness over the past month predicted decreased longevity over 5 years (Stones et al., 1989). However, a study of 129 institutionalized individuals found no discernable effect for life satisfaction predicting longevity over 4 years (O’Connor & Vallerand, 1998). Taken together, mortality studies suggest that PA may be important for sustained wellbeing in healthy older individuals. However, high levels of PA may pose a risk for institutionalized individuals, who may have chronic conditions or age-related declines in health. Two possible interpretations have been offered for this effect. One is that extreme happiness actually reflects denial of loss of control over health, which could lead to poor health behaviors or reduced compliance with medical treatments. Alternatively, low levels of PA may signal a fighting spirit (Pressman & Cohen, 2005). The study of gifted children is consistent with the idea that extreme positivity, especially commencing at an early age, may signify a sense of invulnerability that culminates in risk-taking or negative health behaviors. Of course, it is  17 possible that intense levels of PA influence health more directly through underlying physiological processes. Positive Affect and Morbidity. In general, studies that assessed broad psychological traits like optimism or self-esteem have shown beneficial or protective effects of these dispositions. For instance, optimism has been associated with lower incidence of heart disease and better pulmonary function in male war veterans (Kubzansky, Sparrow, Vokonas, & Kawachi, 2001), as well as better recovery from coronary artery bypass surgery in patients with cardiovascular disease (Scheier, Matthews, Owens, Magovern, Lefebvre, Abbott, et al., 1989). In HIV patients in the middle stages of their disease, optimism predicted slower disease progression over 2 years, as indicated by slower declines in CD4 cells and viral loads. Those scoring in the lower quartile of optimism lost CD4 cells at a 1.55 times greater rate than those scoring the upper quartile (Ironson, Balbin, & Stuetzle, 2005). Another trait that is closely tied to optimism is hope, which is defined by some as a forward-looking emotion and by others as a mode of future-oriented thinking. A prospective study of patients that were recruited from primary care facilities found that higher levels of hope predicted lower likelihood of having or developing hypertension (HT), diabetes mellitus (DM), and respiratory tract infections (RTIs) (Richman, Kubzansky, Maselko, Kawachi, Choo, & Bauer, 2005). Controlling for trait anger and trait anxiety did not alter these results, suggesting that the effects could be attributed to the presence of this positive emotional trait rather than the absence of negative emotional traits. However, it is not clear how much of these health benefits are specific to the affective components of these traits compared to their cognitive components. Parallel findings have emerged from prospective studies that utilized more narrowly defined measures of dispositional PA. In a study of patients with cardiovascular disease aged 55  18 and older, Middleton & Byrd (1996) found that those who retrospectively reported greater levels of happiness at baseline had lower hospital readmission rates 90 days later, controlling for factors like initial health status, length of initial stay, and hope for the future. However, this study confounds PA and NA because the measure of happiness was based on a combined score (PA minus NA). Another study of healthy seniors revealed that lower levels of PA, but not higher levels of NA, both assessed at baseline using items from the CES-D, predicted greater stroke occurrence over 6 years (Ostir, Markides, Peek, & Goodwin, 2001). Associations with PA held constant after controlling for NA. One study that examined PA and NA in relation to cold symptoms deserves special mention because of the way affect was conceptualized and measured. In this study, researchers exposed 334 healthy adults to one of two cold viruses. Prior to viral exposure, participants were asked to rate their daily mood over 7 days, using 9 positive and 9 negative adjectives. Adjectives were chosen to represent 3 subcategories of PA (vigor, well-being, and calm) and 3 subcategories of NA (depression, anxiety, and hostility). Summary measures of trait PA and trait NA were obtained by averaging these ratings across the 7 days. Aggregating multiple measures of state affect over time is a reliable and recommended method for assessing stable affective tendencies. Results revealed that higher levels of trait PA were associated with lower rates of clinical infection, even after controlling for NA and other relevant factors such as age, sex, and initial antibody levels. Analyses of individual affect subscales revealed that the association was significant for vigor and well-being only. This finding indicates that not all types of PA have similar influences on health. However, neither trait NA nor the interaction between trait NA and trait PA were associated with rates of clinical infection (Cohen, Doyle, Turner, Alper, & Skoner, 2003). Interestingly, trait PA was associated with fewer, and trait NA  19 with greater, self-reported symptoms. When both affective styles were entered together, neither was significant. This latter finding highlights the overlapping variance between PA and NA and the importance of considering their joint influences. Most other investigations in the affect and health field have studied NA and PA in isolation under the implicit assumption that they are uncorrelated. Because mood adjectives refer to pure feeling states, it is unlikely that the observed effects were due to components other than affect. As in studies of NA, many commonly used measures in the PA literature tap complex phenotypes that include behavior, affect, and cognition. Components other than affect could therefore play a role in health. Secondly, many studies confounded PA and NA by employing bipolar scales to assess PA (e.g. Janoff-Bulman & Marshall, 1982; Kawamoto & Doi, 2002; Koivumaa-Honkanen et al., 2000; Parker et al., 1992). Hence, they cannot rule out NA as an alternative explanation. Finally, Pressman and Cohen (2005) point out that some common PA adjectives like active, energetic, and peppy might be problematic for use with older individuals because they could be tightly coupled with perceived health. Summary of Limitations of Previous Work The most critical issues with the affect and health literature can be captured in four main points. First, it remains unclear whether the associations are due to moods or related cognitive or behavioral processes. Second, assuming that moods do play a role, it remains to be sorted whether the effects are due to NA or PA, valence or activation, or whether these components operate independently or together. Third, due to the vast number of studies relying on medical patients, it is difficult to know whether moods contribute causally to health or if they are merely a proxy for disease severity. Fourth, assuming that moods do act causally, we do not know what  20 the plausible mechanisms are; one outstanding question is whether NA and PA work through different pathways. These problems are discussed in turn below. The first issue in the literature has to do with the choice of instruments that have been used to measure affect. Several commonly utilized measures fail to discriminate genuine feeling states from related personality traits, cognitive states, or symptoms of psychopathology. This is especially true for studies of NA, which have mainly focused on complex psychological constructs like depression or hostility. In general, investigations of PA have employed more refined indices of affect, yet a significant number of studies have assessed broader constructs like optimism or life satisfaction. Even if moods are involved in associations with health, issues of independence and overlap further complicate interpretation of the literature, making it difficult to sort out which affective ingredients are responsible for these links. Identifying the key ingredients involved in the affect and health link is the first step in developing an accurate theoretical framework to guide future work. At a more practical level, this knowledge could lead to the development of more appropriate interventions to improve the health trajectories of individuals at risk. In investigations of NA, researchers are beginning to recognize that despite their distinctive qualities, depression, anxiety, and anger/hostility share much in common. Pressman and Cohen (2005) have likewise questioned whether it is important to distinguish between various positive affects such as happiness, joy, and elation, or whether they are comparable enough to have similar influences on health. The problem of dependency is not simply a concern for same-valence moods. NA and PA are also related, though this correlation tends to weaken across longer time intervals (Diener & Emmons, 1984; Diener, Smith, & Fujita, 1995). When NA and PA are not studied separately,  21 it is impossible to ascertain whether health outcomes are due to high levels of NA; low levels of PA; or both. This may be especially relevant to understanding the depression and health link, as depression is thought to be a disorder of high NA and low PA. Although researchers have distinguished moods largely on the basis of their valence, few have differentiated them according to their level of activation. Pressman and Cohen (2005) speculated that activated pleasant moods (e.g. joy) and unactivated pleasant moods (e.g. calm) may influence health differently. The issue of activation must also be considered for negative moods. To date, only one study has specifically examined the contributions of arousal vs. valence using a dimensional model of affect. Results revealed that mood states with a high arousal component were associated with objective asthma symptoms, whereas mood states with a high pleasantness component were associated with subjective symptom reports (Affleck, Apter, Tennen, Reisine, Barrows, Willard, et al., 2000). A third problem that remains to be addressed is whether moods are causally related to health outcomes. Studies of patients cannot rule out the possibility that moods reflect disease severity. In other words, those with more advanced disease may experience more negative moods or fewer positive moods, either in response to their prognosis or as a direct result of their medical condition. It is also possible that unmeasured, preclinical disease processes contribute to affective changes in some healthy individuals. Some researchers try to deal with this issue by statistically controlling for disease severity at baseline. However, getting accurate indicators of disease severity is often difficult, and the success of this approach depends entirely on the reliability and validity of these indicators. Lastly, the biological pathways through which moods might influence health are largely unknown. One unsettled question in the literature is whether there are NA-specific and PA-  22 specific pathways to health. Several biological mechanisms have been proposed to mediate the path from NA to disease, including cardiovascular reactivity (e.g. Smith & Ruiz, 2002), HPA dysregulation (e.g. Stetler & Miller, 2005; Yehuda, Teicher, Trestman, Levengood, & Siever, 1996), and inflammatory processes (e.g. Gallo & Matthews, 2003; Kiecolt-Glaser, McGuire, Robles, & Glaser, 2002; Miller & Blackwell, 2006; Miller, Stetler, Carney, Freedland, & Banks, 2002). On the other hand, the pathways linking PA to health are less well understood. In some cases, PA has been associated with biological alterations in the same direction as NA, whereas others have documented changes in the opposite direction, or shown no association at all. There is some speculation that growth hormone and prolactin are uniquely responsive to PA, however there is very limited evidence to support this contention (e.g. Berk, Tan, Fry, Napier, Lee, Hubbard, et al., 1993; Codispoti, Gerra, Montebarocci, Zaimovic, Raggi, & Baldaro, 2003). To reiterate, there is a demand for studies that use uncontaminated, “pure” indices of affect that simultaneously assess a broad spectrum of moods differing in both valence and level of activation. In addition, more studies of healthy young individuals are needed in order to sort out issues of mood being an artifact of existing or silent disease, by way of looking at outcome measures that reflect biological precursors of illness. This approach offers a viable way to examine the unique vs. joint contributions of different affective tendencies, varying in valence and activation, to biological mechanisms involved in disease development. Moreover, although research suggests that acute emotions may trigger disease events in individuals with chronic illnesses, enduring patterns of affect are likely more important for long-term health outcomes, especially in healthy individuals. Such affective tendencies are best studied by assessing mood over extended time periods.  23 Theories of Affect Negative Affect. Historically, interest in NA predominated affect theory and research. This is in part due to the common conception that NA is bad for one’s well-being. Excessive NA has been associated with a plethora of psychological and physiological problems, whereas problems stemming from excessive PA are rare. Another reason why NA has driven affect theory is that some commonly experienced states like fear and anger have obvious survival value. Central to models of NA is the idea that negative emotions create specific action tendencies, urging the organism to respond in organized ways (Lazarus, 1991). For example, anxiety and fear prepare the body to respond to threat or emergency by triggering a cascade of physiological changes. In contrast, positive emotions are not associated with such an organized physiological response. Finally, negative emotions are more differentiated than positive emotions, whether measured with self-report scales or with objective measures like facial expression (e.g. Ellsworth & Smith, 1988a; Ellsworth & Smith, 1988b). Positive Affect. The relatively recent interest in PA has been fueled by Fredrickson’s (1998) influential broaden-and-build theory. According to this model, feelings like joy, interest and contentment function to correct or undo the aftereffects of negative feelings. Supporting this conjecture, a number of studies have shown that positive emotions aid physiological recovery following a stressful emotional experience (Fredrickson & Levenson, 1998; Tugade & Fredrickson, 2004). The model also posits that the effects of positive emotions are cumulative. Feeling good in the present also increase the likelihood that one will feel good in the future, leading to an upward spiral effect. Whereas negative emotions are important in emergency situations and produce quick physiological responses, the influence of positive emotions may  24 unfold more gradually. In other words, to reap the benefits of positive emotions, one must experience them frequently over longer durations. Valence and Arousal: Two Core Dimensions of Affect. Years of research on affect has led to the general consensus that subjective affective experience is best captured by a circumplex structure, with mood items arranged in a circular space. This structure was derived from analyses of the observed relationships between mood items. In Russell’s (1980) model, each mood state is represented as a linear combination of two independent dimensions representing valence and arousal. The valence dimension determines the subjective degree of pleasantness or unpleasantness of a mood state, whereas the arousal dimension determines the degree of physiological activation or stimulation associated with a particular mood state. A similar model has also been proposed by Larsen and Diener (1992). Mood states that lie directly on the valence dimension are neutral with respect to level of arousal, and mood states that lie directly on the arousal dimension are neutral with respect to valence. All other mood states are a mixture of these two underlying dimensions. Although some researchers disagree that valence and arousal are the fundamental components of the circumplex structure, all other labels that have been proposed are merely different rotations of these two orthogonal dimensions. For example, the two focal dimensions of Watson and Tellegen’s (1985) circumplex, which they labeled PA and NA, represent 45-degree rotations of the valence and arousal dimensions. However, because Watson and Tellegen emphasized the activated poles of these dimensions, the PA and NA subscales of their PANAS scale (Watson, Clark, & Tellegen, 1988) only include items that capture mood states of high activation. Watson and colleagues later decided to rename the PANAS PA and NA scales to Positive Activation and Negative Activation to more accurately reflect this aspect (Watson, Wiese, Vaidya, & Tellegen, 1999).  25 Studies in affective neuroscience also support the existence of two distinct neural circuitries for valence and arousal (Posner, Russell, & Peterson, 2005). The valence circuitry involves the mesolimbic dopamine system, extending from the ventral tegmental area to the nucleus accumbens, which further projects to the hippocampus, amygdala, and prefrontal cortex (Depue, Luciana, Arbisi, Collins, & Leon, 1994). Although this system primarily processes reward and pleasure, studies of drug withdrawal indicate that hypoactivation of this area is associated with a variety of negative emotions (e.g. Diana, Pistis, Muntoni, & Gessa, 1996). A growing body of literature also indicates that positive and negative emotions are accompanied by different patterns of prefrontal cortical activation. Specifically, left-sided activity accompanies positive emotions, whereas right-sided activity accompanies negative emotions. Stable differences in prefrontal activation asymmetry have also been linked to temperamental dispositions towards PA and NA (Tomarken, Davidson, Wheeler, & Doss, 1992; Davidson, 1998). Finally, PA and NA have been associated with two separable motivational systems in the brain that guide approach and withdrawal behavior. NA, which serves to keep the organism out of trouble by promoting vigilant and apprehensive behavior, is linked to the behavioral inhibition system (BIS). PA, which simultaneously motivates and rewards approach related behavior, is linked to the behavioral approach system (BAS) (Gray, 1987; Carver & White, 1994). The pathways of the arousal network are thought to include the thalamus, amygdala, and the reticular system (Jones, 2003). These regions respond to both aversive and appetitive excitatory stimuli, with increased activation in these regions being associated with states of extreme emotional arousal such as panic or mania (e.g. Rauch, Shin, & Wright, 2003).  26 Evidence for the operation of two separable circuitries for valence and arousal suggests that when considering the effects of PA and NA, it is imperative to distinguish between mood states within a particular valence that differ with respect to degree of arousal. Are NA and PA Independent? Research on self-reported affect initially started out with the idea that positive and negative mood are bipolar opposites. However, this proved difficult to demonstrate empirically, with some researchers claiming independence and others claiming bipolarity. In a recent analysis of this vexing problem, Russell and Carroll (1999) outline several factors that can influence the observed correlation between PA and NA. Measurement error plays a critical role, as do item selection and response format. Measurement intervals can also bias this correlation towards bipolarity or independence. When reporting intervals are short (e.g. asking participants to rate how they feel at the moment), PA and NA will tend to be negative correlated; at longer response intervals or when ratings are averaged across time, the correlations will tend to weaken (Diener & Emmons, 1984; Diener, Smith, & Fujita, 1995). The circumplex model of affect offers a reconciliation of the independence vs. bipolarity debate. This structure is purported to capture the full range of affective experience by considering not only the valence of a mood state but also its level of activation. Consider the circumplex structure depicted in Figure 1, which is based on Larsen & Diener’s (1992) model. The configuration consists of two core affective dimensions, activation and pleasantness, and two intermediate dimensions that reflect a mixture of activation and pleasantness. Terms that are on orthogonal dimensions (i.e. 90 degrees apart) are semantically unrelated and should thus show low correlations (r = 0 in theory). Passive-blue, stimulated-nervous, and relaxed-drowsy are examples of such pairs. On the other hand, terms that are on opposite ends of the same dimension are opposite in meaning and should be highly negatively correlated (r = -1.00 in  27 theory). Happy-sad, active-passive, anxious-calm, and peppy-drowsy are examples of this kind. However, as mentioned previously, this dependence weakens when ratings are aggregated across time. So although one cannot be happy and sad in the same instant, over the long term, one can experience states of both happiness and sadness.  Figure 1 Mood Circumplex  ACTIVATED active, lively, activated ACTIVATED UNPLEASANT anxious, nervous, irritable  ACTIVATED PLEASANT peppy, stimulated, enthusiastic  UNPLEASANT sad, blue, dissatisfied  PLEASANT happy, cheerful, pleased  UNACTIVATED UNPLEASANT drowsy, tired, sluggish  UNACTIVATED PLEASANT calm, relaxed, peaceful UNACTIVATED passive, quiet, at rest  28 Summary of Current Study The current investigation differs from prior studies of the affect and health link in several key aspects. First, it employed the circumplex model of mood assessment in order to measure basic affective tendencies. This measure is superior to previously utilized measures of affect because it is grounded in affect theory, allowing a broad spectrum of commonly experienced negative and positive mood states to be assessed simultaneously. The model also makes specific predictions regarding the degree of overlap between different mood states. Importantly, it differentiates between valence and activation, two core dimensions of the affective experience that are believed to reflect the operation of separate neural circuitries. The model also separates PA from NA, making it possible to assess their independent and overlapping contributions. Another novel feature of this study is that it was based on data obtained from physically healthy female adolescents. This population was of interest for several reasons. First, choosing young and healthy individuals minimized the possibility of hidden disease processes contributing to affective experience. Second, a majority of studies have focused on male samples, so the generalizability of findings to females is questionable. Third, adolescence is a critical developmental period accompanied by intense emotional experience, especially for girls. It is also a time when the behavioral and biological processes involved in many chronic diseases are initiated (Matthews, 2005). The present investigation was linked to an ongoing larger study examining the longitudinal relationship between the emergence of depressive symptoms and changes in health in adolescent females at risk for depression based on family history or cognitive style. Although the decision to focus on this particular sample was based on practical reasons, it presented an opportunity to examine characteristic patterns of NA and PA prior to the onset of depression.  29 While the obvious downside of this approach was limited generalizability, a likely benefit was that it increased the potential to see more variability in mood than would be present in the general population. Although participants were selected on the basis of being at higher risk for depression, they were required to be in otherwise excellent health to get into the study, so confounding by disease was not an issue. This investigation examined a collection of biological symptoms referred to as the metabolic syndrome. This syndrome is considered a critical intermediate condition on the way to diabetes mellitus and cardiac disease. We chose to focus on this outcome because most of the findings in the affect and health literature relate to these conditions. The contribution of the metabolic syndrome to premature CHD is on par with risk factors such as smoking (Wilson, D’Agostino, Levy, Belanger, Silbershatz, & Kannel, 1998). Indicators of the metabolic syndrome include central adiposity, insulin resistance, elevated BP, low high-density lipoproteins (HDL), and elevated triglycerides. Epidemiological studies have established that the constellation of risk factors comprising the metabolic syndrome ultimately forecast cardiovascular morbidity and mortality events (DeFronzo & Ferrannini, 1991; National Cholesterol Education Program, 2001). Metabolic symptoms begin to appear as early as childhood and adolescence (Arslanian & Suprasongsin, 1996; Chen, Srinivasan, Elkasabany, & Berenson, 1999; Csabi, Török, Jeges, & Molnar, 2000; Weiss, Dziura, Burgert, Tamborlane, Taksali, Yeckel, et al., 2004), making them a good candidate outcome for a study that focuses on younger persons. Available evidence also suggests that metabolic symptoms intensify with age (Chen, Bao, Begum, Elkasabany, Srinivasan, & Berenson, 2000), and track from childhood into adulthood (Katzmarzyk, Perusse, Malina, Bergeron, Despres, & Bouchard, 2001), i.e. youth who have high levels of metabolic symptoms also tend to as adults, and vice-versa. The assessment  30 and detection of the metabolic syndrome in adolescents is based on guidelines that have been established for adults (National Cholesterol Education Program, 2001), with cut points adjusted to reflect age-appropriate norms (Chen et al., 2000; Weiss et al., 2004). Hypotheses Relative Contributions of NA and PA. Although most of the biological outcomes of interest in this study have been traditionally tied to NA, the confounding role of PA has rarely been taken into account. Therefore, the first goal of the study was to test the relative contributions of global NA and global PA on the measured biological outcomes. In the circumplex structure, the left half represents global NA, and the right half represents global PA. Our main hypothesis was that indicators of metabolic syndrome would show moderately strong univariate associations with global NA and global PA. However, we expected stronger associations with global NA. This is because anxiety, anger, and depression can all be subsumed under a global NA factor, whereas depression is the only affective construct characterized by low levels of PA. We also expected that, when global NA and global PA are entered simultaneously in the prediction model, their unique effects would diminish due to their overlapping variance. The Role of Activation. An additional hypothesis was that the effects of NA and PA would vary depending on level of activation. For negatively valenced moods, we hypothesized that stronger levels of arousal would predict greater risk for metabolic syndrome. The opposite pattern was expected for positively valenced moods, such that lower levels of arousal would predict lower risk for metabolic syndrome. In the circumplex model, the activated unpleasant dimension (anxious/nervous/irritable) captures moods that are closely connected to anger and anxiety, which share high arousal as a common element. Based on the consistency of evidence linking trait anger and trait anxiety with  31 risk for CHD onset in healthy adults, we expected this dimension to show the strongest association with risk status in our sample. Because the unactivated pleasant dimension (calm/relaxed/peaceful) represents the semantic opposite of the activated pleasant dimension, we expected higher mean levels on this dimension to predict lower risk status. We expected the unpleasant (sad/blue/dissatisfied) and pleasant (happy/cheerful/pleased) dimensions, which reflect moderate levels of arousal, to bare more modest relations with biological risk. Higher means on the unpleasant dimension suggest mild depressive states, whereas higher means on the pleasant dimension suggest the absence of such states. Higher means on the unactivated unpleasant dimension (drowsy/tired/sluggish) suggest more intense depressive states typified by low vigor and disengagement, whereas higher means on the activated pleasant (peppy/stimulated/enthusiastic) dimension suggest greater vigor and engagement with life. Although there is considerable evidence that depression may lead to deteriorating health in vulnerable patient populations, evidence for depression as a long-term promoter of disease processes in initially healthy individuals is less consistent. In addition, the direction of the relationship between depression and disease is not as clearly established as it is for other negative affects. As such, we anticipated that these two dimensions would be weak predictors of biological risk. Because we view affective valence as a necessary element in disease proneness, we did not predict the pure activated and pure unactivated dimensions of the circumplex to have consequences for health. As such, we hypothesized that these two dimensions would be unrelated to biological risk. NA-specific and PA-specific Pathways. As mentioned previously, stable individual differences in positive and negative affectivity have been linked to different motivational  32 systems in the brain and different patterns of prefrontal activity. If trait PA and trait NA have different neurobiological substrates, then it is reasonable to expect some degree of specificity in the peripheral biological pathways through which they influence health. As such, we anticipated that some biological risk factors would be uniquely associated with general trait NA and others would be uniquely associated with general trait PA. We also considered the possibility that more specific relations will be detected for activated vs. unactivated dimensions of these traits. However, these analyses were largely exploratory as there is insufficient evidence to explicitly predict which individual risk factors will be uniquely associated with specific dimensions of affect. The Role of Affective Variability. Whereas the previous hypotheses are concerned with differences in average levels of affective intensity, the repeated measures design of this study afforded the opportunity to explore another aspect of individual functioning, namely variability in affective intensities over time (Larsen, 1987; Russell & Carroll, 1999). In past research, emotional variability has been linked to high conflict and imagination (Wessman & Ricks, 1966) and low self-complexity (Linville, 1982), among other correlates. A more recent study revealed that individuals with borderline personality disorder showed more intraindividual variability in overall affect intensity as well as intensity of pleasant affect compared to normal controls (Russell, Moskowitz, Zuroff, Sookman, & Paris, 2007). Consistency of attributes across time or situations is considered by some to be an important dimension of personality functioning (e.g. Larsen, 1987; Shoda, Mischel, & Wright, 1994) and may be predictive of physical health outcomes. For example, in a study of older adults in a retirement community, individual differences in week-to-week variability in perceived control over a 7-month period predicted 5year mortality, whereas average levels did not (Eizenman, Nesselroade, Featherman, & Rowe,  33 1997). To our knowledge, only one study to date has examined affect variability in relation to a physical health outcome. In this study of healthy older adults, PA variability (defined as the within-person standard deviation) in the afternoon hours was significantly associated with higher evening cortisol (r = 0.38) (Simpson, McConville, Rae, O’Connor, Stewart-Knox, Coudray, et al., 2008). In other words, participants with fluctuating afternoon PA may have had difficulty “turning off” the stress response at the end of the day. The relevance of intra-individual affect variability to physical health outcomes has otherwise been overlooked. We predicted that fluctuating patterns of affect would be associated with increased risk status compared to more stable patterns of affect, holding mean levels constant. As these analyses were exploratory, no specific predictions were made with respect to the relative importance of specific dimensions of affect.  34 METHOD The present investigation was undertaken as part of a larger prospective study examining the relationship between stress and health in female adolescents at risk for depression. Since its inception in 2004, 157 young women have enrolled in this study, which entailed a total of 6 scheduled visits separated by 6-month intervals. Although each laboratory session included a standard set of measures, the specific combination of interviews, questionnaires, tests, and biological measures varied from visit to visit. Each visit lasted approximately 2-3 hours, depending on the specific set of measures included in that session. A summary of procedures and measures that are pertinent to the present investigation are presented below. Participants Adolescent females aged 15 to 19 from the greater Vancouver area were recruited for the 2.5-year long prospective study through advertisements in local newspapers, highschools, and community colleges. Individuals who responded to these advertisements were directed to the study website, where they were asked to complete two questionnaires – one to assess cognitive vulnerability to depression and another to assess family history of depression. Those who responded affirmatively to the second questionnaire were contacted by a research assistant, who asked them follow-up questions in order to ascertain whether their family member might be reasonably classified as meeting DSM-IV criteria for depression. Individuals who were deemed to have a first-degree relative with a history of clinical depression, or scored in the upper quintile of the population distribution on cognitive vulnerability, were considered eligible for participation. Prospective participants were also required to be free of any past or present medical conditions necessitating hospitalization or surgery, and not currently on any medications (other than oral contraceptives) that could interfere with biological measurements. Individuals  35 with chronic conditions were excluded from the study only if they were required to take medication in order to manage their symptoms (e.g. asthma). However, those with relatively minor chronic conditions that did not require medical interventions (e.g. seasonal allergies) were considered eligible for study entry. At the time of their first laboratory visit, participants were additionally screened for the presence of any lifetime Axis I disorders using the Structured Clinical Interview for DSM-IV, Non-Patient version (SCID-NP; First, Spitzer, Gibson, & Williams, 2002). Participants who met diagnostic criteria for current or lifetime major depression, mania, dysthymia, psychosis, drug/alcohol abuse or eating disorders at study entry were excluded from the study. Data collection for the present investigation occurred between visits 4 and 5, which will hereon be referred to as the baseline and follow-up visits. Of the 128 participants who completed the baseline visit in time to be included in the present study, 32 did not attempt any mood diaries. These participants were excluded from analyses. Eight of the remaining 96 participants did not return for their follow-up session, and hence did not have any biological data available for analysis. This resulted in a final sample size of 88. A comparison of participants vs. non-participants is given on p. 38 of the Results section. The age of participants at the time of the follow-up session ranged from 17 to 21, with a mean of 19.26 years (SD = 1.42). The sample was ethnically diverse, including 46.6% European, 37.5% Chinese, 6.8% Other Asian, 1.1% African, 3.4% Aboriginal, and 4.5% reporting other ethnicities. Procedure At the baseline visit, participants were assessed for the presence of current or lifetime Axis I disorders using the SCID-NP. This diagnostic information was used to ensure that observed associations between mood and metabolic syndrome were not simply an artifact of  36 their shared variance with disordered affect. Participants subsequently provided various biological measures, and completed an array of interviews, questionnaires, and tasks that were not pertinent to the present investigation. At the end of the session, they were instructed on how to complete the web-based mood diary. Participants received $50 for time spent at the laboratory and $20 advance payment for the web diary. They were also reimbursed for transportation and/or parking fees. Between the baseline and follow-up visits, participants were asked to complete a weekly web-based diary to assess their mood. Participants were encouraged to vary the days and times that they complete the diary from week to week in order to capture as much variability in their moods as possible and to avoid time of day effects. The time and date stamp feature of the webbased survey made it possible to keep track of precise diary entry times. In order to facilitate compliance, these records were monitored on a weekly basis, and email reminders were issued to participants who failed to complete a diary within 2 weeks of their prior entries. If participants failed to complete within one week of this reminder, another reminder was issued through a phone call. At the follow-up visit, participants once again underwent a structured psychiatric interview to assess the presence of any current or lifetime Axis I disorders. Biological outcome measures that are relevant to the present investigation were also collected at this time. Overnight fasting blood samples were collected via blood draws performed by a trained phlebotomist. These blood samples were used to quantify serum levels of glucose, insulin, high-density lipoprotein (HDL), and triglycerides. Measures of systolic and diastolic blood pressure, as well as waist and hip circumference were also obtained. At the end of the study, participants were paid $50 for the visit, and were reimbursed for transportation and/or parking fees.  37 Mood Diary Each week, participants were asked to rate how well each of 24 adjectives described the way they had been feeling in the last 24 hours on a scale from 0 (not at all) to 5 (extremely). By asking them to report on their mood in the preceding day rather than the past week, we aimed to minimize the effects of retrospective bias. Sixteen items for the mood measure were drawn from Larsen and Diener’s (1992) mood circumplex. This instrument consists of eight octants representing different combinations of valence and arousal, with 2 adjectives in each octant. A third adjective was added to each octant in order to broaden the range of relevant mood descriptors, yielding a total of 8 new items. Two of these additional items were taken from Watson et al.’s (1988) PANAS scales and the remaining 6 were chosen from a set of adjectives proposed by Feldman-Barrett and Russell (1998). The adjectives corresponding to each octant of the circumplex can be found in Figure 1. Biological Outcome Measures Systolic and Diastolic Blood Pressure. To measure resting SBP and DBP, participants were seated in a chair with an occluding cuff on one arm. Following a 5-minute rest period, 4 blood pressure readings, spaced 2 minutes apart, were collected using an automated oscillometric device (BpTRUTM BPM-100, VSM MedTech, Coquitlam, BC). Average SBP and DPB were based on the last 3 readings. Blood pressure is expressed in mm Hg. The BpTRUTM has been tested in a population group ranging in age from 3 to 83 years. This non-invasive blood pressure monitoring device yields data that are as precise as those obtained from assessments performed in a physician’s office, and its overall accuracy is comparable to the recognized standard, auscultatory mercury sphygmomanometer (Mattu, Heran, & Wright, 2004).  38 Waist-hip Ratio. Central adiposity was estimated by the waist-hip ratio (WHR). Measurements were obtained using a fabric tape measure. Waist circumference was measured at the midpoint between the upper iliac crest and lower costal margin at the midaxillary line. Hip circumference was measured at the maximum width of the buttocks. During measurement, participants were instructed to stand with their feet together and adopt a relaxed body position. Glucose. Glucose assays were performed on an ADVIA 1650 Chemistry System (Bayer Diagnostics, Tarrytown, NY) using enzymatic techniques with hexokinase and glucose-6phosphate dehydrogenase. The inter-assay coefficient of variation for this method is 1.2%. Glucose is expressed in mmol/L. Insulin. Insulin was measured using a chemiluminescent technique on an IMMULITE 2000 (Diagnostic Products Corporation, Los Angeles, CA). The inter-assay coefficient of variation for this method is 3.1%. Insulin is expressed in pmol/L. High Density Lipoprotein (HDL). Quantification of HDL was achieved through enzymatic techniques that utilized cholesterol esterase and cholesterol oxidase, after very-low density, low-density, and intermediate-density lipoproteins were removed through centrifugation. The inter-assay coefficient of variation for this method is 5.1%. HDL is expressed in mmol/L. Triglycerides. Triglycerides were measured on a Hitachi 747 instrument (Kyowa Medex, Japan) using enzymatic techniques. The inter-assay coefficient of variation for this method is 1.1%. Triglycerides are expressed in mmol/L. Control Measures Measures of age, ethnicity, and socioeconomic status were entered as control variables in all regression analyses predicting biological outcomes, because these factors are known  39 predictors of metabolic syndrome and are likely correlates of affect. Age (in years) was defined as age at the time of the follow-up visit. Ethnicity was defined on the basis of participants’ selfidentified racial/cultural status. For the purpose of regression analyses, this discrete variable was converted into two dichotomous variables through dummy coding, with one variable representing European background (European=1 vs. non-European=0) and the other representing Asian background (Asian=1 vs. non-Asian=0). Thus the reference group represented individuals from non-European and non-Asian backgrounds. Socioeconomic status (SES) was defined as the highest level of education (in years) attained by the participant’s most educated parent or legal guardian.  40 RESULTS Comparison of Participants vs. Non-Participants One-way ANOVAs were used to test whether participants (N = 88) and non-participants (N = 32) differed on the continuous outcome variables (biological outcomes, SES, and age). ’  Due to unequal sample sizes between these two groups, Welch’s F statistic was calculated to obtain a more accurate p-value for these outcomes. Chi-square analysis was used to test whether the two groups differed on the categorical ethnicity variable. WHR was the only outcome on ’  which the two groups differed significantly (F  = 6.20, p = .02).  The groups did not differ on  ’  the remaining variables (F s < 2.07, p’s > .17; ethnicity χ 2(6, 128) = 5.33, p = .50). Pre-Analysis Data Screening Outliers. Inspection of univariate distributions of the biological variables revealed a small number of outliers, defined as more than 3 standard deviations above the sample mean. These included one outlier for SBP and DBP (this individual had consistently high blood pressure measurements across all laboratory visits), one outlier for triglycerides (high value), two outliers for glucose (low values), and four outliers (high values) for insulin. However, as all of these values fell within the normal physiological range, they were retained in all analyses. No outlier cases were identified for the composite biological outcome or the set of aggregated mood variables used as predictors. The calculation of these indices is explained in detail below. Missing Data. As previously mentioned, 32 of the 128 participants were excluded from analyses as they had completely missing diaries. The number of weekly diary entries ranged from 2 to 28, with a mean of 15.28 and a standard deviation of 6.21. A majority of participants (88.6%) had at least 8 weeks of diary data, and 56.2% had at least 3 months (approximately 14  41 weeks) worth of entries, which translates into half of the expected number of entries for a 6month timeframe. Examination of the weekly diary entries for the final sample revealed a small percentage of missing mood ratings. This percentage was estimated by summing the number of missing adjective ratings across all weeks of attempted diary entries, then dividing this sum by the total expected number of ratings. An “attempted” diary entry was identified by the presence of a date stamp. Given that there were 1423 attempted diary entries and 24 adjectives per entry, 34,152 ratings would be expected in total. With 855 adjective ratings omitted, the percentage of missing mood scores was estimated as 2.5%. Due to the unlikelihood of encountering problems as a result of this small percentage of missing observations, these data were treated as missing at random without replacement. Of the 88 participants included in the analysis, 78 had complete data on all 7 biological measures, with the remaining 10 participants missing values on one to 4 measures. While this led to a loss of power in regression analyses involving individual biological outcomes, as explained below, cases with partial biological data were retained when the predicted outcome was the composite biological risk score. Examination of Residual Scatterplots. As all analyses were based on linear regression techniques, residual plots of predicted outcome scores against errors of prediction were visually inspected to ensure that assumptions of linearity, normality and homoscedasticity were not violated. Examination of these plots indicated that all of these assumptions were met for each regression test. Inspection of loess curves further indicated that the relationships between the mood predictors and biological outcomes were best represented by linear rather than curvilinear functions.  42 Data Reduction Creation of Composite Biological Outcome Score. While there are no definitive criteria for the diagnosis of metabolic syndrome, the National Cholesterol Education Program Adult Treatment Panel III (ATP III, 2001) and the World Health Organization have published a set of risk factors based on clinical cut-off scores. For present purposes, metabolic syndrome was defined on the basis of these guidelines as the presence of 3 or more of the following indicators: (1) Abdominal obesity, defined as WHR > 0.8, (2) High SBP and DBP, defined as >=130/85 mmHg, (3) Signs of insulin resistance, defined as fasting insulin levels >= 108 pmol/L or fasting glucose levels >= 5.5 mmol/L, (4) Elevated triglycerides, defined as >= 1.7 mmol/L, (5) Low HDL cholesterol, defined as < 1.3 mmol/L. As would be expected in a young and physically healthy sample, few participants exceeded these clinical cut-off scores. Of the total sample, 4.5% (N=4) had a high WHR, 1.1% (N=1) had high blood pressure, 14.8% (N=13) had high fasting insulin, 9.1% (N=8) had elevated triglycerides, and 9.1% (N=8) had low HDL. None of the participants qualified as having high levels of fasting glucose, and only 4.5% of the entire sample (N=4) met criteria for metabolic syndrome. Given the relatively good health of these young women, a cumulative risk index was created for each individual by averaging her z-scores on the biological measures of interest (i.e. SBP, DBP, WHR, HDL, fasting glucose, fasting insulin, and triglycerides), with higher scores indicative of less optimal health, or increased risk for metabolic syndrome. The linear combination approach based on continuous rather than categorical risk factors has previously been used to define metabolic syndrome in adolescent and young adult populations (Ravaja, Keltikangas-Jarvinen, & Keskivaara, 1996; Raikkonen, Matthews, & Salomon, 2003). Because higher HDL scores imply lower risk, z-scores on this measure were multiplied by -1 before inclusion in the cumulative score. This strategy served to  43 maximize variability of the biological outcome variable. Using the mean of the z-scores rather than their sum also allowed for individuals with missing data on one or more of the variables to be included in the analysis. Creation of Composite Mood Variables. Ratings for the 24 adjectives were reduced to 8 unipolar mood vectors based on their theoretical placements on the circumplex, with each vector comprising the average of 3 adjective ratings across time. Thus, moving counter-clockwise around the circumplex, the 8 vectors were formulated as follows: (1) average ratings of happy/cheerful/pleased formed the pleasant mood (neutral activation) vector, (2) average ratings of peppy/stimulated/enthusiastic formed the activated pleasant mood vector, (3) average ratings of active/lively/activated formed the activated mood (neutral valence) vector, (4) average ratings of anxious/nervous/irritable formed the activated unpleasant mood vector, (5) average ratings of sad/blue/dissatisfied formed the unpleasant mood vector, (6) average ratings of drowsy/tired/sluggish formed the unpleasant unactivated mood vector, (7) average ratings of passive/quiet/at rest formed the unactivated mood (neutral valence) vector, and (8) average ratings of calm/relaxed/peaceful formed the unactivated pleasant mood vector. Following this data reduction scheme, global NA was defined as the average of the 3 negatively valenced vectors, and global PA was defined as the average of the 3 positively valenced vectors. Stability of vector scores was assessed by computing the proportion of total variance in mood ratings attributable to between-person differences. Hierarchical linear modeling was used to partition the within-person and between-person variance in mood ratings for each of the eight vectors. Estimates of temporal stability ranged from a low of .41 for the pleasant vector, to a high of .49 for the activated unpleasant vector, indicating adequate stability and justifying the aggregation scheme.  44 Descriptive Findings The minimum and maximum values, means, and standard deviations for the composite mood variables and indicators of biological health are listed in Table 1. Average ratings for negatively valenced moods were predominantly low, leading to distributions that were positively skewed to varying degrees, with distributions for unpleasant mood and global NA showing the most substantial skew. While there are no distributional assumptions for predictors in regression analyses, predictions are usually enhanced if these variables are normally distributed. However, our overall results were unaltered when we applied data transformation strategies to improve the distribution of these variables. In contrast, average ratings for positively and neutrally valenced moods had nearly normal distributions. As compared to other mood dimensions, average ratings for the affectively neutral unactivated mood dimension showed lower variability.  Table 1 Descriptive Statistics for Affective Predictors and Biological Outcome Measures N  Minimum  Maximum  Mean  SD  Affective Predictors Pleasant  88  .55  4.11  2.3  .82  Activated pleasant  88  .09  3.64  1.78  .76  Activated  88  .25  3.53  1.94  .78  Activated unpleasant  88  .10  3.33  1.22  .65  Unpleasant  88  .03  4.17  .95  .73  Unactivated unpleasant  88  .27  4.00  1.63  .71  Unactivated  88  .48  2.83  1.44  .46  Unactivated pleasant  88  .48  4.08  1.93  .75  45 N  Minimum  Maximum  Mean  SD  Global NA  88  .29  3.11  1.26  .61  Global PA  88  .39  3.91  2.00  .72  Mean SBP (mm Hg)  86  82.33  167.00  103.22  10.24  Mean DBP (mm Hg)  86  53.50  124.00  67.24  9.33  Glucose (mmol/L)  83  3.50  5.40  4.63  .40  Insulin (pmol/L)  81  14.99  253.00  70.67  45.54  HDL (mmol/L)  81  1.02  2.49  1.63  .31  Triglycerides (mmol/L)  82  .40  3.52  .99  .51  WHR  86  .65  .88  .75  .05  Composite Outcome Score  88  -1.02  1.93  .00  .52  Biological Outcomes  Note. SBP = Systolic Blood Pressure; DBP = Diastolic Blood Pressure; HDL = High Density Lipoprotein; WHR = Waist to Hip Ratio; NA = Negative Affect; PA = Positive Affect.  Not unexpectedly, four of the seven biological measures (namely SBP, DBP, triglycerides, and insulin) had positively skewed distributions, with most scores falling on the low end of the range (indicative of better health) and not many cases with high scores. The distribution of the composite biological outcome measure was somewhat positively skewed, though to a lesser extent compared to the individual measures. On the other hand, distributions for glucose, HDL, and WHR were roughly normal. Bivariate Correlations Among Composite Mood Variables Zero-order correlations between the composite mood variables are provided in Table 2. The pattern of correlations revealed some unusual and largely unexpected relationships. Correlations among positively valenced mood vectors were substantial, ranging from .62 to .86;  46 correlations among negatively valenced moods were also sizeable, ranging from .59 to .76. Correlations between negatively and positively valenced moods were negative, ranging from a low of -.02 (n.s.) between activated pleasant and activated unpleasant mood, to a high of -.42 between pleasant and unpleasant mood. Not surprisingly, positively valenced moods were strongly correlated with global PA, and negatively valenced moods were strongly correlated with global NA. The correlation between global PA and global NA was -.31, indicating that they shared less than 10% of their variance. In sum, the obtained pattern of correlations revealed a higher than expected overlap between similarly valenced moods; this was especially true for positive moods. On the other hand, oppositely valenced mood vectors exhibited small to moderate negative correlations. Perhaps the most perplexing finding was the small positive correlation between the theoretically bipolar activated and unactivated mood vectors. As well, the activated mood vector had strong positive correlations with positively valenced mood vectors and small negative correlations with negatively valenced mood vectors. On the other hand, the unactivated mood vector showed modest positive correlations with all mood vectors, irrespective of valence. The meaning of these findings and their implications for the study of affect extended over time is considered in more detail in the Discussion section. Bivariate Correlations Among Biological Measures Zero-order correlations between the individual biological measures and composite risk index are provided in Table 3. The composite outcome showed the strongest association with SBP, followed by DBP, triglycerides, and insulin; its relationships with HDL, WHR, and glucose were more modest, but still significant. As would be expected, SBP and DBP were  47 highly correlated; SBP was also significantly related to glucose. Insulin, triglycerides, and WHR were significantly associated with each other but unrelated to any of the other measures. Tests of Primary Hypotheses All hypotheses were tested through a series of hierarchical multiple regression analyses, with the biological outcome of interest (cumulative risk index and individual measures) modeled as a function of affective tendency, controlling for demographic variables (age, ethnicity, and SES). For each regression, demographic variables were entered together as a block in step one. The affective tendency predictor pertinent to the hypothesis being tested was entered in step two. All assumptions of multiple regression were met for each analysis, including normality, linearity, and homoscedasticity, as determined through visual inspection of residual scatterplots.  Table 2 Bivariate Correlations Among Composite Mood Variables Pleasant  Activated Activated Activated pleasant unpleasant  Unpleasant Unactivated Unactivated Unactivated Global NA unpleasant pleasant  Pleasant  --  Activated pleasant  .86**  --  Activated  .88**  .92**  --  Activated unpleasant  -.30**  -.02  -.12  --  Unpleasant  -.42**  -.18  -.24*  .76**  --  Unactivated unpleasant  -.29**  -.16  -.21*  .62**  .59**  --  Unactivated  .25*  .19  .25*  .19  .23*  .34**  --  Unactivated pleasant  .78**  .62**  .68*  -.34**  -.29**  -.23*  .59**  --  Global NA  -.39**  -.14  -.22*  .90**  .89**  .84**  .29**  -.33**  --  Global PA  .97**  .90**  .91**  -.24*  -.33**  -.25*  .38**  .87**  -.31**  Note. NA = Negative Affect; PA = Positive Affect; *p < .05, **p < .01 (2-tailed). 48  49 Table 3 Bivariate Correlations Among Biological Measures SBP  DBP  Glucose Triglycerides HDL  Insulin WHR  SBP  --  DBP  .80**  --  Glucose  .27*  .19  --  Triglycerides  .10  .04  .04  --  HDL  -.21  -.14  .12  -.05  --  Insulin  .15  -.04  .15  .34**  -.12  --  WHR  -.04  -.18  -.18  .44**  -.16  .29*  --  Composite Risk Index  .72**  .58**  .38**  .55**  -.44**  .54**  .39**  Note. SBP = Systolic Blood Pressure; DBP = Diastolic Blood Pressure; HDL = High Density Lipoprotein; WHR = Waist to Hip Ratio; sample sizes range from 79 to 88; *p < .05, **p < .01 (2-tailed).  Relative Contributions of NA and PA. One of the key objectives of this study was to test the relative contributions of NA and PA to overall metabolic health. Based on past research, both global NA and global PA were expected to predict our biological outcome score when tested individually, though global NA was expected to be a stronger predictor. In addition, taking into account the likely overlap between global NA and global PA, their relative contributions were expected to diminish when these predictors were entered simultaneously in the analysis.  50 As seen in Table 4, none of the demographic variables were significant predictors of biological health. After controlling for age, ethnicity, and SES, global NA was a significant predictor of metabolic symptoms, with each one-point increase in NA scores predicting a .23 increase in z-scores on the composite. Also evident is that global NA accounted for approximately 6% of the variance in metabolic symptoms. Contrary to expectations, global PA did not emerge as a significant predictor of the composite. Scatterplots depicting the relationships between global affect and metabolic symptom scores are presented in Figures 2 and 3, along with their corresponding regression lines. In these graphs, all predictors have been mean-centered, and the ethnicity variables have been coded to reflect European background (i.e. European = 1 and Asian = 0). Thus the regression lines represent the predicted relationship between global affect and metabolic symptom scores for individuals of European background, aged 19.26 years, with a parental education of 16.26 years. With global PA accounting for near-zero variance in outcome, it was not surprising that inclusion of both affective measures in the regression model had almost no impact on the predictive strength of global NA, clearly demonstrating that the relation between metabolic symptoms and NA is unique and not due to its overlapping variance with PA.  51 Table 4 Hierarchical Multiple Regressions Predicting the Composite Biological Outcome from Global NA and Global PA B  SE B  β  p  ∆R2  ∆adjR2  ∆F  p  .02  -.02  .49  .82  .07  .06  5.86  .02  .02  -.03  .39  .82  .01  -.01  .53  .47  Model 1 Step 1 Constant  -1.35  .98  .17  Age  .06  .04  .15  .21  European ethnicity  .01  .22  .01  .96  Asian ethnicity  -.12  .22  -.12  .58  SES  .00  .02  .02  .84  Step 2 Global NA  .23  .09  .27  .02  Model 2 Step 1 Constant  -.99  1.02  .33  Age  .05  .05  .15  .23  European ethnicity  .01  .22  .01  .97  Asian ethnicity  -.10  .23  -.09  .67  SES  .01  .02  .05  .65  Step 2 Global PA  -.06  .09  -.09  .47  52 B  SE B  β  p  ∆R2  ∆adjR2  ∆F  p  .02  -.03  .39  .82  .07  .05  2.89  .06  .02  -.03  .39  .82  .07  .04  2.05  .11  Model 3 Step 1 Constant  -1.34  1.00  .19  Age  .06  .04  .15  .21  European ethnicity  .01  .22  .01  .96  Asian ethnicity  -.12  .22  -.12  .59  SES  .00  .02  .02  .83  Step 2 Global NA  .22  .10  .26  .03  Global PA  .00  .09  .00  .97  Model 4 Step 1 Constant  -.99  1.02  .16  Age  .05  .04  .14  .25  European ethnicity  .01  .22  .01  .98  Asian ethnicity  -.12  .22  -.12  .58  SES  .00  .02  .02  .85  Step 2 Global NA  .20  .11  .24  .06  Global PA  -.04  .10  -.05  .73  NA*PA interaction  -.10  .16  -.08  .52  Note. SES = Socioeconomic Status; European ethnicity (1 = European; 0 = Non-European); Asian ethnicity (1 = Asian; 0 = NA = non-Asian); Negative Affect; PA = Positive Affect.  53 Figure 2 Scatterplot of Global NA and Metabolic Symptom Scores  Figure 3 Scatterplot of Global PA and Metabolic Symptom Scores  54 In order to explore the possibility that the effect of global PA was moderated by global NA, thus masking its relationship with our outcome, an interaction term was created by multiplying the mean-centered values of global NA and global PA. This product term was then added as a third predictor in the regression model1. However, as indicated in Table 5, this analysis yielded no evidence for an interaction effect, though inclusion of the interaction term somewhat weakened the relationship between global NA and the outcome. According to models of NA and health, negative emotions may indirectly influence physical health by promoting poor health practices such as smoking, drinking, and lack of exercise. We tested this idea by examining whether any of these behaviors mediated the relationship between global NA and metabolic symptoms in our sample. For this purpose, we relied on three self-report measures, all obtained at the follow-up visit and referring to health practices in the past 6 months: Smoking status (0 = no, 1 = yes), alcohol consumption (average number of drinks consumed in a typical week), and exercise (total minutes of exercise per week). Not only were these health behaviors unrelated to the biological outcome (all p’s > .25), controlling for them in the regression model had no impact on the strength of the relationship between global NA and metabolic risk. The Role of Activation. A further aim of this study was to find out whether the relationship between mood and metabolic symptoms would vary depending on level of activation. Specifically, the predictive strength of negative mood was expected to increase with increasing levels of arousal. In contrast, the predictive strength of positive mood was expected to increase with diminishing levels of arousal. These hypotheses were not supported. The regression coefficients for the three negative affect dimensions were largely similar (see Table 1  Centering the predictors prior to forming the interaction term is recommended as it reduces their correlation with the interaction term.  55 5), implying that it is the shared rather than unique variance between these dimensions that explains their relationship with biological risk. As evident from Table 6, the regression coefficients for the positive affect dimensions were not significant, indicating that regardless of activation, PA is unrelated to metabolic risk in this sample. Consistent with our expectations, dimensions representing pure unactivated and pure activated mood states were not significant predictors of the biological outcome (β = .17, p = .13, and β = -.06, p = .62, respectively).  Table 5 Hierarchical Multiple Regressions Predicting the Composite Biological Outcome from Negative Mood Vectors Differing in Degree of Activation B  SE B  β  p  ∆R2 ∆adjR2  ∆F  p  .02  -.03  .39  .82  .05  .04  4.26  .04  Model 1 Step 1 Constant  -1.5  1.00  .14  Age  .06  .04  .16  .18  European ethnicity  .02  .22  .02  .94  Asian ethnicity  -.09  .22  -.09  .67  SES  .01  .02  .05  .65  Step 2 Unactivated unpleasant  .17  .08  .23  .04  56 B  SE B  β  p  ∆R2 ∆adjR2  ∆F  p  .02  -.03  .39  .82  .05  .04  3.97  .05  .02  -.03  .39  .82  .06  .05  5.22  .03  Model 2 Step 1 Constant  -1.05  .99  .29  Age  .05  .04  .13  .28  European ethnicity  .00  .22  .00  .99  Asian ethnicity  -.10  .22  -.10  .64  SES  .01  .02  .06  .58  Step 2 Unpleasant  .16  .08  .22  .05  Model 3 Step 1 Constant  -1.35  .98  .17  Age  .06  .04  .16  .19  European ethnicity  .03  .22  .02  .91  Asian ethnicity  -.12  .22  -.12  .58  SES  .00  .02  .02  .88  Step 2 Activated unpleasant  .20  .09  .26  .03  Note. SES = Socioeconomic Status; European ethnicity (1 = European; 0 = Non-European); Asian ethnicity (1 = Asian; 0 = NA = non-Asian); NA = Negative Affect; PA = Positive Affect.  57 Table 6 Hierarchical Multiple Regressions Predicting the Composite Biological Outcome from Positive Mood Vectors Differing in Degree of Activation B  SE B  β  p  ∆R2  ∆adjR2  ∆F  p  .02  -.03  .39  .82  .02  .00  1.28  .26  .02  -.03  .39  .82  .00  -.01  .29  .59  Model 1 Step 1 Constant  -.86  1.02  .40  Age  .05  .04  .14  .27  European ethnicity  -.01  .22  -.01  .98  Asian ethnicity  -.12  .23  -.11  .61  SES  .01  .02  .06  .61  Step 2 Unactivated pleasant  -.09  .08  -.13  .26  Model 2 Step 1 Constant  -1.02  1.02  .32  Age  .05  .05  .15  .23  European ethnicity  .01  .22  .01  .97  Asian ethnicity  -.09  .23  -.09  .70  SES  .01  .02  .05  .66  Step 2 Pleasant  -.04  .08  -.07  .59  58 B  SE B  β  p  ∆R2  ∆adjR2  ∆F  p  .02  -.03  .39  .82  .00  -.01  .12  .73  Model 3 Step 1 Constant  -1.08  1.01  .29  Age  .05  .05  .15  .23  European ethnicity  .02  .23  .02  .95  Asian ethnicity  -.07  .23  -.07  .76  SES  .01  .02  .05  .67  Step 2 Activated pleasant  -.03  .08  -.04  .73  Note. SES = Socioeconomic Status; European ethnicity (1 = European; 0 = Non-European); Asian ethnicity (1 = Asian; 0 = NA = non-Asian); NA = Negative Affect; PA = Positive Affect.  NA-specific and PA-specific Pathways. An additional purpose of this study was to examine whether global NA and global PA would relate to different biological outcomes. However, these analyses were viewed as exploratory because we did not have strong a priori predictions as to which outcomes would be uniquely associated with NA versus PA. This issue was explored by running separate regression models for each of the 7 biological outcomes, with global NA and global PA entered simultaneously in each model in order to provide a more stringent estimate of their unique effects. The results of these analyses are listed in Table 7. Global NA emerged as a significant predictor of SBP and glucose; it was also a weak predictor of DBP. Its relationship with the remaining four outcomes did not reach statistically significant levels. Global PA, on the other hand, was unrelated to any outcome. In these analyses, higher age was marginally related to higher DBP, and Asian ethnicity marginally predicted lower glucose levels compared to other ethnic groups.  59 Table 7 Hierarchical Multiple Regressions Predicting Individual Biological Measures from Global NA and Global PA B  SE B  β  p  ∆R2  ∆adjR2  ∆F  p  .04  -.01  .77  .55  .07  .05  3.12  .05  .04  -.01  .85  .50  .06  .04  2.53  .09  Outcome: SBP Step 1 Constant  86.17  19.64  .00  Age  .77  .86  .11  .37  European ethnicity  3.32  4.49  .16  .46  Asian ethnicity  1.71  4.60  .08  .71  SES  -.49  .34  -.16  .15  Step 2 Global NA  4.78  1.92  .29  .02  Global PA  1.06  1.77  .08  .55  Outcome: DBP Step 1 Constant  38.70  18.04  .04  Age  1.50  .79  .23  .06  European ethnicity  .36  4.12  .02  .93  Asian ethnicity  -3.06  4.22  -.16  .47  SES  -.018  .32  -.01  .95  Step 2 Global NA  3.09  1.76  .20  .08  Global PA  -1.37  1.62  -.10  .40  60 B  SE B  β  p  ∆R2  ∆adjR2  ∆F  p  .08  .03  1.54  .20  .07  .05  2.82  .07  .03  -.02  .65  .63  .00  -.03  .14  .87  Outcome: Glucose Step 1 Constant  4.23  .78  .00  Age  .01  .03  .02  .89  European ethnicity  -.27  .16  -.34  .10  Asian ethnicity  -.30  .17  -.37  .08  SES  .02  .01  .14  .21  Step 2 Global NA  .18  .08  .28  .02  Global PA  .04  .07  .07  .59  Outcome: Insulin Step 1 Constant  181.74 94.64  .06  Age  -4.25  .4.10  -.13  .30  European ethnicity  -2.09  19.96  -.02  .92  Asian ethnicity  -.25  20.63  .00  .99  SES  -2.24  1.72  -.16  .20  Step 2 Global NA  4.80  9.22  .07  .60  Global PA  1.40  8.70  .02  .87  61 B  SE B  β  p  ∆R2  ∆adjR2  ∆F  p  .01  -.04  .27  .90  .02  -.01  .68  .51  .04  -.01  .78  .54  .02  -.01  .79  .46  Outcome: HDL Step 1 Constant  1.93  .64  .04  Age  .00  .03  -.13  .30  European ethnicity  -.07  .14  .02  .88  Asian ethnicity  -.01  .14  SES  .00  .01  -.16  .18  Step 2 Global NA  -.06  .06  -.13  .32  Global PA  -.05  .06  -.12  .38  Outcome: Triglycerides Step 1 Constant  -.18  1.05  .86  Age  .06  .05  .15  .23  European ethnicity  -.02  .22  -.02  .92  Asian ethnicity  -.11  .23  -.11  .64  SES  .03  .02  .17  .17  Step 2 Global NA  .00  .10  .01  .97  Global PA  -.13  .10  -.16  .24  62 B  SE B  β  p  ∆R2  ∆adjR2  ∆F  p  .03  -.02  .54  .71  .01  -.02  .37  .70  Outcome: WHR Step 1 Constant  .66  .10  .00  Age  .00  .00  .09  .48  European ethnicity  .02  .02  .18  .43  Asian ethnicity  .02  .02  .18  .44  SES  .00  .00  .14  .25  Step 2 Global NA  -.01  .01  -.07  .55  Global PA  -.01  .01  -.10  .45  Note. SES = Socioeconomic Status; European ethnicity (1 = European; 0 = Non-European); Asian ethnicity (1 = Asian; 0 = NA = non-Asian); NA = Negative Affect; PA = Positive Affect; SBP = Systolic Blood Pressure; DBP = Diastolic Blood Pressure; HDL = High Density Lipoprotein; WHR = Waist to Hip Ratio; sample sizes range from 81 to 88.  Follow-up regression analyses were conducted to further elucidate which of the negative mood dimensions were the most robust predictors of SBP, glucose, and DBP (see Tables 8-10). Whereas the activated unpleasant and unpleasant dimensions were significant predictors for all three outcomes, the unactivated unpleasant dimension was a marginally significant predictor for SBP but unrelated to DBP or glucose. Thus, in contrast to the earlier finding that the effects of negative mood on the composite biological index did not differ by level of activation, in the case of individual biological measures, higher-activation moods appeared to be stronger predictors.  63 Table 8 Hierarchical Multiple Regressions Predicting Systolic Blood Pressure from Negative Mood Vectors Differing in Degree of Activation B  SE B  β  p  ∆R2  ∆adjR2  ∆F  p  .04  -.01  .77  .55  .04  .02  3.03  .09  .04  -.01  .77  .55  .05  .04  4.41  .04  Model 1 Step 1 Constant  86.77  19.81  .00  Age  .83  .87  .12  .34  European ethnicity  3.57  4.55  .18  .44  Asian ethnicity  1.97  4.55  .10  .67  SES  -.40  .35  -.13  .25  Step 2 Unactivated unpleasant  2.79  1.60  .19  .09  Model 2 Step 1 Constant  94.22  19.29  .00  Age  .64  .86  .09  .46  European ethnicity  2.92  4.52  .14  .52  Asian ethnicity  1.35  4.53  .07  .77  SES  -.50  .35  -.16  .15  Step 2 Unpleasant  3.20  1.52  .21  .04  64 B  SE B  β  p  ∆R2  ∆adjR2  ∆F  p  .04  -.01  .77  .55  .08  .07  6.69  .01  Model 3 Step 1 Constant  87.66  19.15  .00  Age  .84  .85  .12  .32  European ethnicity  3.58  4.45  .18  .42  Asian ethnicity  .91  4.48  .04  .84  SES  -.50  .34  -.16  .14  Step 2 Activated unpleasant  4.50  1.74  .29  .01  Note. SES = Socioeconomic Status; European ethnicity (1 = European; 0 = Non-European); Asian ethnicity (1 = Asian; 0 = NA = non-Asian).  Table 9 Hierarchical Multiple Regressions Predicting Diastolic Blood Pressure from Negative Mood Vectors Differing in Degree of Activation B  SE B  β  p  ∆R2  ∆adjR2  ∆F  p  .04  -.01  .85  .50  Model 1 Step 1 Constant  35.53  18.24  Age  1.51  .80  .23  .06  European ethnicity  .47  4.19  .03  .91  -1.67  4.19  -.09  .69  .02  .32  .01  .95  Asian ethnicity SES  .06  65 B  SE B  β  p  Step 2 Unactivated unpleasant  1.81  1.47  .14  ∆R2  ∆adjR2  ∆F  p  .02  .01  1.51  .22  .04  -.01  .85  .50  .05  .04  3.97  .05  .04  -.01  .85  .50  .06  .05  5.36  .02  .22  Model 2 Step 1 Constant  40.55  17.63  .02  Age  1.37  .79  .21  .09  European ethnicity  -.05  4.13  .00  .99  Asian ethnicity  -2.32  4.14  -.12  .58  SES  -.07  .32  -.02  .83  Step 2 Unpleasant  2.78  1.39  .22  .05  Model 3 Step 1 Constant  35.11  17.60  .05  Age  1.54  .78  .23  .05  European ethnicity  .52  4.09  .03  .90  Asian ethnicity  -2.63  4.2  -.14  .53  SES  -.07  .31  -.02  .84  Step 2 Activated unpleasant  3.70  1.60  .26  .02  Note. SES = Socioeconomic Status; European ethnicity (1 = European; 0 = Non-European); Asian ethnicity (1 = Asian; 0 = NA = non-Asian).  66 Table 10 Hierarchical Multiple Regressions Predicting Glucose from Negative Mood Vectors Differing in Degree of Activation B  SE B  β  p  ∆R2 ∆adjR2  ∆F  p  .08  .03  1.54  .20  .03  .02  2.48  .12  .08  .03  1.54  .20  .06  .05  4.91  .03  Model 1 Step 1 Constant  4.28  .79  .00  Age  .01  .03  .02  .84  European ethnicity  -.27  .17  -.34  .11  Asian ethnicity  -.30  .17  -.37  .08  SES  .02  .01  .17  .14  Step 2 Unactivated unpleasant  .10  .06  .18  .12  Model 2 Step 1 Constant  4.55  .75  .00  Age  .00  .03  .00  .99  European ethnicity  -.29  .16  -.36  .08  Asian ethnicity  -.31  .17  -.39  .06  SES  .02  .01  .13  .25  Step 2 Unpleasant  .13  .06  .24  .03  67 B  SE B  β  p  ∆R2 ∆adjR2  ∆F  p  .08  .03  1.54  .20  .06  .05  5.28  .02  Model 3 Step 1 Constant  4.33  .76  .00  Age  .01  .03  .02  .85  European ethnicity  -.27  .16  -.33  .11  Asian ethnicity  -.32  .16  -.39  .06  SES  .02  .01  .14  .22  Step 2 Activated unpleasant  .15  .07  .26  .02  Note. SES = Socioeconomic Status; European ethnicity (1 = European; 0 = Non-European); Asian ethnicity (1 = Asian; 0 = NA = non-Asian).  Though not directly related to our hypotheses, we were interested to know whether the overall extremity of an individual’s affective style, regardless of associated valence or activation state, predicted metabolic symptoms. This predictor was computed by averaging mood ratings on all 8 dimensions of the circumplex. However, average extremity of response was unrelated to our biological outcome (β = .12, p > .25). The Role of Affective Variability. While the idea of behavioral instability is not new in personality research, few studies to date have examined intraindividual variability in affect using repeated measures data. As such, we were curious to know whether fluctuating patterns of affect would be associated with greater metabolic symptoms compared to more stable patterns of affect, holding mean levels constant. In the present investigation, three separate measures of affective instability were created for global NA, global PA, and the affect intensity measure just  68 described above. The standard deviations of these variables around each person’s mean were then entered as predictors, controlling for mean values, in three separate regression equations predicting the composite biological index. None of these measures of affective variability emerged as significant predictors (all p’s > .50). Secondary Analyses Presence of Mood Disorder. In order to rule out the possibility that the observed effects were due to emotional disturbances caused by depression rather than negative affective tendencies, the regression models were re-run with depression status entered as a control variable. Participants were classified as depressed if they met DSM-IV criteria for a minor or major depressive episode at the time of the baseline or follow-up visit (current depression), or had experienced such an episode at any point since their last visit that was resolved (past depression). Of the 88 participants who were included in this study, only 7 received a diagnosis of either past or current depression. When depression status (1 = current or past episode, 0 = no history of depression) was entered as a covariate in the regression models, the results remained essentially unchanged, indicating that the reported effects were not attributable to the presence of a mood disorder. Influence of Number of Observations. As the number of completed diary entries varied considerably between participants, we considered the possibility that the obtained results were influenced by the frequency of mood ratings contributing to estimations of mean affect. That is, if participants were more likely to complete diaries during periods of high distress rather than periods of elation, the potential for detecting NA versus PA effects may have increased as a result of greater variance in negative compared to positive mood ratings. However, in direct contradiction to this possibility, the number of diary entries was significantly negatively  69 correlated with global NA (r = -.30, p = .004), but was unrelated to global PA (p = .50), indicating that individuals with more observations tended to have lower negative mood ratings. Because the overall effect of this would be to restrict the range of these ratings, a more conservative test of NA effects seems likely, increasing our confidence in the reported findings. We also considered the possibility that frequency of entries was a proxy for an unmeasured external variable such as mental health, life stress, or an aspect of personality like conscientiousness, all of which could be related to NA and inflate its link with the outcome. However, the effects of global NA on the outcome persisted after controlling for number of diary records, allaying this concern.  70 DISCUSSION In recent years, there has been considerable progress in our understanding of the link between affect and health. Prospective studies have implicated broad NA states like depression, anxiety, and hostility in the development and progression of a variety of medical conditions. There is mounting evidence that PA, too, plays a role in some conditions. Despite these advances, several questions remain unanswered. First, researchers have typically relied on instruments that measure complex psychological or personality constructs, which confound mood with cognitive and behavioral processes. Second, it is unclear whether the observed effects are due to NA or PA, valence or activation, or if these components act independently or jointly. Third, reliance on patients or older populations introduces the possibility that affect is confounded with disease severity or undetected conditions. Finally, it is unclear whether NA and PA exert their influences through distinct pathways. The present investigation was designed to help address these key issues. Using a theorybased dimensional measure of affect, the study tracked patterns of mood in physically healthy young women over the course of 6 months, at which time biological measures reflecting preclinical markers of metabolic disease were collected. The repeated measures design was based on the premise that enduring patterns of affect, rather than acute emotions, are likely to promote disease processes in healthy individuals. With this approach, we were able to analyze the relative contributions of core affective tendencies to biological mechanisms involved in disease development. This section will outline findings in relation to each of the hypotheses tested in this study, and place them in the context of prior research. It will also discuss the implications of these findings for models of health, and more generally for dimensional models of emotion. The  71 generalizability of the findings and limitations of the study design will be considered next. The section will conclude with recommendations for future work. Global NA Effects The results of this investigation indicated that participants who tended to experience more intense negative emotions over the course of 6 months had more metabolic symptoms compared to those who experienced these emotions to a lesser degree. On the whole, our findings cohere well with the vast research literature implicating broad negative affective states like anger, depression, and anxiety in the development of cardiovascular disease and diabetes. They are also consistent with the sparse set of prospective studies showing that negative affective states precede the development of the metabolic syndrome or its key components, insulin resistance and central adiposity. In a recent review of this literature, Goldbacher and Matthews (2007) concluded that depression, anger, and hostility predict the metabolic syndrome or its key components, although the direction of the effect is less clear for anxiety. They also note that the effects for depression and anger are particularly strong for women. More recently, in a study of over 3,000 healthy men and women, Toker, Shirom, and Melamed (2008) found that depression among women, but not men, was associated with a 1.94-fold risk of having the metabolic syndrome. As the present investigation was conducted with an exclusively female sample, our findings are especially relevant to reports of sex-dependent effects. The findings of the present investigation imply that the overlap between various negative emotions may be more critical for disease risk than their specificity. This draws attention to the possibility that some of the depression, anxiety, and anger effects that have been reported in the literature may be due to features they share. Our findings concur with arguments made by Suls & Bunde (2005) that these psychological constructs may not have distinctive effects, but  72 promote risk for CHD because of a common underlying tendency to experience chronic negative emotions. If true, these results suggest a different approach to construing and assessing psychological risk for disease. At a minimum, they suggest that considering affective constructs in isolation may hide important dimensions underlying their relationships with health outcomes. What these constructs share in common may also give clues to mechanisms of action, limiting the number of potential mediators. If the critical factor truly is NA, then it is likely that behavioral or biological mechanisms, rather than cognitive mechanisms, are liable. These results also stress the need to focus more on the role of core dispositional affects, as risk for disease may reach beyond clinically significant psychological impairment. This knowledge is essential not only for model building, but also for clinical practice, where most of the focus is on screening for clinical depression or anxiety rather than for dispositional tendencies to experience subclinical negative affect. Our findings also contribute to the literature by showing that the connection between NA and health is not limited to older individuals, but may take root at a relatively young age, instigating subtle alterations in biological functioning that could, over time, lead to more widespread physiological dysregulation. Another reason our findings are significant is that in middle-aged or elderly persons, mood may be an artifact of unobserved health conditions, whereas in a young and healthy sample, the likelihood of NA being confounded by underlying disease is minimal. As such, the direction of the observed associations is likely in the direction of NA to metabolic symptoms rather than the other way around (though we were unable to test this in the present investigation because the biological outcomes were assessed only once, at the end of the diary period).  73 Assuming NA does promote metabolic symptoms, how might it do so? The most likely explanation is that it works via its effects on the sympathetic nervous system (SNS) and hypothalamic pituitary adrenal (HPA) axis. SNS activity triggers the release of epinephrine and norepineprhine, which initiate a series of biological changes that prepare the body for immediate action. As a result, blood pressure and heart rate go up, fat and glucose are metabolized for instant energy, and non-essential activities like digestion are slowed down. These immediate changes are backed by cortisol, which is released in response to HPA activity. While these responses are adaptive in the short term, prolonged exposure to them can lead to wear and tear and failure of normal physiological functions. Over time, the fatty acids and glucose that are mobilized in response to SNS and HPA activity begin to accumulate underneath the lining of blood vessels. Chronic elevations in these stress hormones also make cells resistant to insulin, further exacerbating this problem. Additionally, cortisol directly effects fat storage and weight gain, especially in the abdominal region. These disruptions can, in turn, lead to the development of metabolic syndrome, which can set the stage for heart disease and diabetes (Sapolsky, 1998). Negative emotions can also adversely influence health by discouraging positive health practices like sleep, diet, and exercise, and encouraging deleterious behaviors such as smoking and alcohol consumption, all of which have been implicated in the development of the metabolic syndrome, heart disease, and diabetes (Goldbacher & Matthews, 2007). However, we did not find that health behaviors mediated the relationship between NA and health in the present sample, suggesting that the influence of NA on metabolic symptoms is more likely to occur through direct physiological responses rather than indirect behavioral pathways. In sum, the results of our study suggest that a general disposition to experience negative emotions may be an early risk factor for metabolic syndrome, which is itself a known risk factor  74 for medical conditions like CHD and type 2 diabetes. This knowledge also has practical significance, as it suggests that interventions targeting negative affective tendencies and related lifestyle alterations at an early age could have the potential to avert negative health trajectories and reduce future risk for disease. Global PA Effects Contrary to expectations, the propensity to experience higher levels of positive emotions over time did not predict lower metabolic symptom scores in our study sample. This finding seemingly contradicts the literature showing generally favorable effects of PA on health. However, whereas prior studies in this realm have almost exclusively focused on older adults and long-term outcomes, our study focused on younger people and proximal outcomes. We will next consider why age and proximity of outcome may have obscured potentially beneficial effects of PA in this sample. One possibility is that the influence of PA on health entails subtle changes that are not immediately apparent but accumulate gradually over time. This may be especially true if PA influences on health occur primarily through psychological processes rather than through direct physiological mechanisms. Such indirect mechanisms are described in the stress-buffering model proposed by Pressman and Cohen (2005), which attributes the health effects of PA chiefly to its ability to allay the detrimental impact of stress on the body. The stress-buffering view is also consistent with Frederickson’s (1998) broaden-and-build theory of positive emotions. Numerous studies have demonstrated that positive emotions temporarily broaden the scope of attention and cognition (e.g. Basso, Schefft, Ris, & Dember, 1996; Fredrickson & Branigan, 2005; Isen, Daubman, & Nowicki, 1987; Isen, Johnson, Mertz, & Robinson, 1985), inspiring Frederickson’s idea that positive emotions loosen the grip that negative emotions have on one’s  75 thinking. In addition to expanding intellectual resources, she has also theorized that positive emotions create resilience through enhanced social interactions and exploratory behaviors. These cognitive and behavioral changes could, over time, buffer individuals from future negative emotions, reducing the likelihood of downstream health risks. Fredrickson’s theory implies that it is necessary to experience positive emotions repeatedly over time for cognitive and behavioral changes to yield enduring effects. Smith & Baum (2003) have similarly suggested that PA may protect individuals from negative emotional and cognitive responses to stress by encouraging restorative health behaviors like sleep, vacation, relaxation, and exercise, and promoting secretion of endogenous opioids, all of which have the potential to curb the physiological stress response. To the extent that PA operates in the subtle and gradual ways these models imply, our study could have overlooked its benefits by limiting assessment of metabolic symptoms to one occasion shortly after diary monitoring was completed. One study by Steptoe & Wardle (2005) underscores this point by showing that trait PA (ratings of happiness averaged over the working day) was unrelated to same-day ambulatory SBP but was significantly inversely related to SBP at 3-year follow-up, even after adjusting for negative affect. There is also evidence that some of PA’s health benefits arise because of its ability to undo the physiological consequences of NA. We tested this possibility in the current sample, but found no evidence for such a moderating influence of PA on metabolic symptoms. These findings are inconsistent with recent studies demonstrating that positive emotions may speed cardiovascular and immune recovery from negative emotional arousal (e.g. Fredrickson & Levenson, 1998; Ong & Allaire, 2005; Tugade & Fredrickson, 2004; Valdimarsdottir & Bovbjerg, 1997). However, in these studies mood was either manipulated through a laboratory stressor, or was experienced as a transitory state. To our knowledge, the only naturalistic study  76 that has tested this hypothesis is the viral exposure study by Cohen et al. (2003), which examined whether trait NA and trait PA predicted rates of clinical infection in participants who were exposed to a cold virus. While trait PA was associated with lower rates of infection, there was no evidence that it interacted with trait NA. It also bears noting that all of the laboratory studies on undoing have focused on PA’s ability to soothe physiology after brief bouts of emotional arousal (typically less than an hour), but it may take considerably longer for PA to “undo” the effects of persistent NA. This may explain why PA did not moderate the association between NA and metabolic risk in our sample. Evolutionary perspectives also suggest that the effects of PA on health are unlikely to be as potent as NA effects. As the primary purpose of NA is to protect the organism from danger by activating a coordinated, intensive biological response, it seems plausible that repeated exposure to NA would exhaust bodily resources relatively quickly. On the other hand, the primary purpose of PA is to motivate approach behaviors (e.g. eating, drinking, socializing). Watson et al. (1999) have speculated that the natural rhythm of PA is to “ebb and flow with the daily tide of events”, suggesting that variations in PA tend to be subtle. In addition, whereas we can prolong the effects of negative events through cognitive processes such as worry and rumination, the effects of positive events, when they do occur, may be relatively short-lived. Even exceptionally happy events such as winning the lottery do not alter one’s average level of PA for long (Brickman, Coates, & Janoff-Bulman, 1978). Research also suggests that the cognitive process by which people attempt to make sense of positive events leads to the paradoxical effect of diminishing pleasure (Wilson, Centerbar, Kermer, & Gilbert, 2005). All of these observations further support the view that it may be necessary to experience PA consistently over long periods to reap its benefits.  77 Another reason why our study may have failed to detect PA effects is that younger individuals may have generally higher levels of PA compared to older individuals, who may have more variability in their range of positive emotions. However, in this case a negatively skewed distribution would be expected, with a clustering of scores in the higher ranges, which was not true for our study sample. If anything, there was a noticeable absence of scores on the high end of the scale. This suggests a somewhat different possibility, which is that the health benefits of PA emerge at intensities that were not found in this sample. Unfortunately, we were unable to find published data enabling us to compare the extremity of PA scores in our sample to those obtained in other studies that used circumplex-based instruments to assess mood. However, there are two reasons to rule out range restriction as a factor in the absence of PA effects. First, the overall shape of the distribution obtained in our sample conforms well to observations made by Watson et al. (1999) regarding the typical distribution of PA scores. Specifically, they note that “PA scores show a roughly symmetrical and slightly platykurtic distribution”, with “substantial variability across a broad range”. Second, when we computed the disattenuated correlation coefficient to correct for range restriction, the resulting estimate did not differ substantially from the original. A final speculation relates to the makeup of our sample. Given that these young women were characterized by a depressogenic cognitive style, or exposed to first-degree relatives with a depressive disorder, they may have been better at recognizing negative mood states, and less adept at recognizing positive mood states, which may have enhanced NA effects while masking PA effects. An alternative explanation is that they were biologically predisposed to react more strongly to NA, or less strongly to PA, or both. As such, they may be more sensitive to the biological influences of NA, and less sensitive to the biological influences of PA, compared to  78 others. On the other hand, if the primary influence of PA on health is driven by a stressbuffering effect, it is possible that despite experiencing subjectively high levels of PA, these young women continued to think or act in maladaptive ways that undermined their ability to cultivate their coping resources. An additional possibility is that the restorative mechanisms involved in recovery from stress-induced physiological activation (whether or not these mechanisms are mediated by PA) were in some way impaired. It is also plausible that many of our participants were in the midst of major life changes that tend to occur during the transition into adulthood, resulting in heightened susceptibility to NA and diminished receptivity to PA. To sum, our results indicate that dispositional patterns of PA have neither a protective nor detrimental effect in a young and healthy sample. While the absence of potentially harmful effects was predictable, the absence of potentially beneficial effects was unexpected. In light of prior theory and research, we have considered various reasons why this prediction was not supported in our data set. Exploring how age moderates the effects of PA on both proximal and distal indicators of health would be a particularly meaningful avenue for future work. Activation Effects for NA Our analyses found generally similar effects for negatively valenced mood vectors on metabolic risk, regardless of degree of activation. Furthermore, neither mood vector was a significant predictor after controlling for the other two, which rules out independent effects of mood vectors differing in activation. These observations provide some substantiation for the idea that the critical psychological ingredient associated with disease risk is the propensity to experience negative emotions of all kinds, regardless of level of activation. To our knowledge, only three other studies have explicitly compared the effects of negative emotional tendencies differing in activation on health outcomes. The first study by  79 Pollard & Schwartz (2003) analyzed the longitudinal relationship of mood with blood pressure and total cholesterol, each assessed four times over an 18-month period in a sample of healthy government employees. Their results indicated that tense arousal (defined as the difference between terms corresponding to the activated unpleasant and unactivated pleasant vectors in our study) and hedonic tone (defined as the difference between terms corresponding to the pleasant and unpleasant vectors in our study) had similar effects on blood pressure. As was the case in our study, neither mood type was a significant predictor after controlling for the other. In another study by Brosschot & Thayer (2003), heart rate data were collected from healthy participants who reported their emotional arousal and emotional valence at study onset. While both emotional arousal and valence predicted initial HR, prolonged activation (at 5 minutes) was solely predicted by emotional valence (negative affect). The results of a third study are less straightforward. In this study, measures of negative affect (average of sad, frustrated, stressed) and arousal (average of alert and tired) were both associated with higher ambulatory blood pressure (Kamarck, Shiffman, Smithline, Goodie, Paty, Gnys, et al., 1998). Although the authors attribute these findings to an emotional activation effect (as opposed to negative emotionality), these results are not so easily interpreted because the measure of negative affect contained a mixture of low and high activation terms, and because the term “tired” implies a negative emotional state rather than a neutral one. An important difference between our study and the ones by Kamarck et al. (1998) and Pollard & Schwartz (2003) is that their studies examined relations between within-person changes in mood and BP, whereas our study explored whether stable individual differences in affect prospectively predict biological functioning. The fact that all negatively valenced moods, regardless of their associated arousal state, showed significant associations with metabolic symptoms in our study is consistent with the  80 notion that a general disposition towards experiencing aversive emotions is responsible for the NA and health link. There has been increasing interest in testing the validity of this idea. As reviewed earlier, studies examining long-term clinical health outcomes in patients and healthy samples have generated some support for this prospect (e.g. Ahern et al., 1990; Frasure-Smith & Lesperance, 2003; Mendes de Leon et al., 1996; Kubzansky et al., 2006). At least two investigations that examined the relation between trait NA and short-term physical outcomes are also worthy of mention. Ewart & Kolodner (1994) examined whether trait anger and trait negative affect in adolescents predicted increases in ambulatory BP across the school day. The measure of negative affect was derived from a factor analysis of trait anxiety and trait depression. Their results revealed that both negative affect and trait anger predicted BP elevations at 4-month follow-up, although negative affect was a stronger predictor. In a more recent study, multiple measures of depression, anxiety, anger, and NA (conceptualized as the shared variance between all three emotions) in a large sample of community volunteers were examined in relation to high-frequency heart rate variability (HF-HRV), an indicator of cardiac autonomic function that has been linked to cardiovascular morbidity and mortality. After adjusting for critical confounds, the authors found that depression, anxiety, and NA were inversely related to HF-HRV, whereas anger was not, leading them to speculate that NA may be the unifying construct linking these different negative emotions to cardiac autonomic dysregulation (Bleil, Gianaros, Jennings, Flory, & Manuck, 2008). Although our results suggest that variations in arousal do not contribute to the link between NA and health, it is also possible that the observed deviations from circumplex structure in this study concealed potentially significant differences in the effects of like-valenced vectors. In contrast to the apparent independence between vectors of opposite hedonic tone, we  81 found that vectors of the same hedonic tone were highly correlated, suggesting that participants did not clearly distinguish between mood adjectives differing in degree of activation. Consistent with our observation, Fisher, Heise, Bohrnstedt and Lucke (1985) demonstrated that, as one moves from state-like to trait-like measures (i.e. average ratings), the circumplex configuration becomes more elliptical, with like-valence ratings clustering more tightly together. They attributed this to the strong role that systematic bias plays in averaged ratings. Alternatively, it is possible that over time, the emotional experience becomes simplified, such that people attend more heavily to differences in hedonic tone but fail to discriminate between low and high arousal states to the same degree. Differences in activation may also become attenuated because as the time period extends, the probability of experiencing like-valence moods will increase. A number of studies have also found that the structure of affect is highly sensitive to individual difference variables. For example, using a within-subjects design, Feldman (1995) found substantial differences in the emphasis that people place on the valence versus arousal dimensions of affective terms. She used the term “valence focused” to describe persons who attend more to the hedonic tone of emotional experiences, and “arousal focused” to describe persons who attend more to the subjective level of arousal of emotional experiences. Similarly, Terracciano, McCrae, Hagemann, and Costa (2003) found that the classic circumplex structure did not fit well with individuals who they deemed to be less affectively differentiated. To reiterate, our hypothesis that the predictive strength of NA would be greater for higher levels of arousal was not well supported. While the high correlation between likevalenced moods in this study may have made it difficult to tease apart their unique effects, it is just as likely that such unique effects do not exist at the trait level, because, as argued earlier, aggregation across time may dilute differences in activation and enhance differences in valence.  82 This does not imply that dispositional affect is devoid of arousal, only that differences in arousal may become less meaningful as one moves away from state to trait-like measures. Activation Effects for PA Despite the fact that global PA was unrelated to the composite metabolic risk in our sample, we considered the possibility that one or more of its constituents could be a significant predictor. This could occur, for example, if a strong relationship with one vector was obscured by its aggregation with two vectors unrelated to the outcome, or if the effects in question were in opposing directions. However, our analyses did not reveal any significant effects for positively valenced mood vectors associated with different levels of arousal. Thus, our findings call into question the utility of distinguishing between positive moods associated with different levels of physiological arousal. Models of PA and health do not make specific predictions about the kinds of positive moods that are likely to have the greatest impact on health. As such, the possibility that activated and unactivated positive emotions can have divergent influences on health has received little attention in prior work. The only germane studies are those involving laboratory mood inductions. In these studies, the effects of PA have sometimes been in the same direction as NA effects, strongly suggesting a role for physiological arousal. On the basis of these findings, some researchers have surmised that PA states such as intense excitement have the capacity to invoke cardiovascular, HPA, and immune system changes similar to those seen in response to negative emotions and threatening stimuli, whereas states of calm contentment do not (Pressman & Cohen, 2005; Kemeny & Shestyuk, 2008). This may occur because transient increases in PA, especially if sufficiently higher than one’s baseline, pose metabolic demands similar to those incurred by highly activated NA states. This is especially likely to be the case in laboratory  83 paradigms that require participants to be actively involved in generating emotional experiences (Kemeny & Shestyuk, 2008). However, as noted by Pressman & Cohen (2005), naturally occurring PA states are rarely intense enough to provoke these kinds of metabolic demands. And even if such high arousal PA states do occur from time to time, it is unlikely that they would be maintained over the long term. Accordingly, the NA-like effects of transient PA states should not be evident at the trait level. As our study depended on naturally occurring PA states across multiple time points, it is doubtful that we would see the kind of PA-elicited arousal effects that have been reported in studies of laboratory based mood manipulations. Furthermore, as both Pressman & Cohen (2005) and Howell et al. (2007) have observed, in studies reporting increased cardiovascular responses to PA, the magnitude of these responses were generally smaller than those reported for NA. In fact, Howell et al. (2007) computed an overall effect size of r = 0.026 (n.s.) for studies that examined PA in relation to measures of cardiovascular reactivity. Prior empirical work also suggests that positive emotions elicit weaker sympathetic activity than negative emotions (e.g. Larsen, Berntson, Poehlmann, Ito, & Cacioppo, 2008; Taylor, 1991). Taken together, this evidence suggests that PA-elicited arousal effects are likely to be uncommon in naturalistic settings, and when they are found, they are likely to be considerably weaker in comparison to NA-elicited responses. In contrast to studies that manipulated high-arousal PA states such as excitement, those that induced unactivated emotions such as relaxation and calmness have tended to report beneficial effects (Pressman & Cohen, 2005). It is important to bear in mind that these divergent effects of activated and unactivated PA have only been observed for state PA. In contrast, studies of trait PA have not explicitly separated the influences of high versus low arousal PA. As stressed earlier, the beneficial effects of trait PA that have been demonstrated in these studies are  84 limited mostly to older individuals and distal outcomes such as morbidity and survival. For this reason, we cannot be sure that the results of the present investigation can be generalized to these populations and outcomes. Whether differences in activation are important at later stages of the life span remains to be answered. It is also possible that the absence of differentiated effects for activated and unactivated PA were obscured by the high correlations between these components in this study. This finding may be a statistical artifact of our method of measurement or sample characteristics, or as argued previously, it may reflect the tendency for people to become less sensitive to differences between moods of the same valence over extended periods of time, while distinguishing increasingly more strongly between oppositely valenced moods. Apart from the various reasons considered thus far (deviations from circumplex structure, sample characteristics, and focus on short-term outcomes), another reason why the present investigation may have failed to reveal the effects of either global PA or its more differentiated components is that these affective tendencies are associated with biological responses that were not assessed in this study. This possibility is considered further in the next section. Evidence for Valence-Specific Pathways Evolutionary perspectives suggest that negative and positive emotions have different functions. Whereas the main purpose of NA is to prepare the body for immediate action in the face of real or perceived threat (i.e. “fight or flight”), PA encourages approach behaviors such as eating, foraging, and socializing, which may have long-term survival value. There is also some evidence that positive and negative emotions are accompanied by different patterns of brain activity (Tomarken et al., 1992; Davidson, 1998). In light of these considerations, we surmised that NA and PA might exert their influence on health through distinct biological responses. While the low correlation between global NA and global PA in our sample substantiates the idea  85 that these emotional dispositions are more independent than bipolar, we did not find any evidence for the specificity hypothesis. Within the constellation of metabolic markers examined in this study, NA was associated with SBP, DBP, and glucose, whereas PA was unrelated to any measure. The specific NA effects that emerged in the present investigation have been demonstrated in several studies examining the relationship between negative psychological characteristics and indicators of the metabolic syndrome (Goldbacher & Matthews, 2007), and may indeed reflect NA-specific effects. However, due to the dearth of prior investigations on the relationship between PA and metabolic syndrome precursors other than blood pressure, it remains unclear whether there exists PA-specific routes to this set of outcomes. Our results suggest that in a young and healthy population at least, the strongest correlate of metabolic syndrome precursors is NA, and that the presence of PA may be inconsequential for this specific constellation of health outcomes. On the other hand, we cannot rule out the possibility that PA exerts its influence through physiological responses that were not examined in this study. One possibility is that PA affects health mainly by dampening arousal through the PNS. This hypothesis is consistent with either the main-effect or stress-buffering model, as the PNS is thought to be in control during pleasant states like digestion and rest, and acts synergistically with the SNS to curb arousal during emergency fight-and-flight responses. Most evidence in support of this premise comes from experimental studies that found positive associations between induced positive mood and indicators of parasympathetic activity, specifically heart rate variability (HRV) and respiratory sinus arrhythmia (RSA) (e.g. Heponiemi, Ravaja, Elovainio, Naatanen, & Keltikangas-Järvinen, 2006; Matsunaga, Isowa, Kimura, Miyakoshi, Kanayama, Murakami, et al., 2009; McCraty, Atkinson, Tiller, Rein, & Watkins, 1995). Increases in HRV have also been reported in relation to higher levels of PA, but not in relation to BDI symptoms,  86 in patients with CAD (Bacon, Watkins, Babyak, Sherwood, Hayano, Hinderliter, et al., 2004; Bhattacharyya, Whitehead, Rakhit, & Steptoe, 2008). On the other hand, decreased parasympathetic activity (Frazier, Strauss, & Steinhauer, 2004) has also been noted, although this may have been due to heightened arousal given that movies were used to induce feelings of excited happiness and amusement. That the direction of the association between parasympathetic activity and PA may depend on level of arousal is demonstrated in a naturalistic study of patients with cardiovascular disease, who gave ratings on eight basic emotions at hospital entry. Results indicated that acceptance produced higher HRV whereas joy produced lower HRV (CatipovicVeselica, Amidzic, Durijancek, Kozmar, Sram, Glavas, et al., 1999). Because the literature on parasympathetic effects of PA is sparse, it is not possible to draw any firm conclusions. In addition, there have been no studies to date examining the effects of trait PA on parasympathetic measures in healthy samples. As such, the reported effects may be transient responses to acute moods or influenced by disease state. Finally, parasympathetic activity has also been related to negative emotions (e.g. Bacon et al., 2004; McCraty et al., 1995; Miller & Wood, 1997), suggesting that the effects are not necessarily unique to PA. In order to find out whether PA is uniquely associated with parasympathetic activity, we would have had to measure specific indicators like HRV or RSA. Without this information, we cannot rule out the possibility that PA exerts its effects chiefly through these physiological processes. Other potential pathways through which PA may impact health are the endocrine and immune systems (Howell et al., 2007; Pressman & Cohen, 2005). Hormones of the endocrine system that have been studied in relation to PA include the stress hormones cortisol, epinephrine, and norepinephrine. Naturalistic studies generally support associations of state and trait PA with lower levels of stress hormones (Pressman & Cohen, 2005). However, in their meta-analysis,  87 Howell et al. (2007) found cortisol to be the only stress hormone significantly related to PA (r = -0.109). Unfortunately, as is the case for most other biological outcomes, an overwhelming majority of these studies looked at cross-sectional or correlational effects of state PA. There has also been much interest in the hypothesis that PA stimulates release of growth hormone, a molecule that supports repair and growth of bodily tissues, and prolactin, a hormone involved in social bonding, parturition and lactation. However, the accurate assessment of these hormones can be difficult due their cyclical and/or pulsatile release patterns. This is especially true for growth hormone and prolactin, whose peak periods of release occur during sleep (Sassin, Frantz, Weitzman, & Kapen, 1972; Takahashi, Kipnis, & Daughaday, 1968). Our decision not to measure these hormones was largely due to these practical considerations. That said, increases in prolactin (Berk et al., 1989; Turner, Altemus, Yip, Kupferman, Fletcher, Bostrom, et al., 2002) and growth hormone (Codispoti et al., 2003) have been reported in response to induced PA, though trait PA was unrelated to prolactin in one study (Pollock et al., 1979). In addition, many of these hormones are also susceptible to induced NA (e.g. Codispoti et al., 2003; Turner et al., 2002), and have known associations with naturally occurring NA (e.g. Smyth et al., 1998; van Eck, Berkhof, Nicolson, & Sulon, 1996) and mood disturbances, particularly for cortisol (Chida & Steptoe, 2009; Linkowski, 2003; Parker, Schatzberg, & Lyons, 2003; Young, 2004). As such, it is unclear whether they are uniquely associated with either PA or NA. Perhaps some of the strongest evidence for PA effects comes from studies of immune functioning. For these studies, Howell et al. (2007) reported an overall effect size of r = 0.332. While a majority of these studies manipulated or measured state PA, it is noteworthy that several studies reported enhanced immune functioning for trait measures of PA. Several studies have reported that induced PA leads to increases in secretory immunoglobulin A (sIgA), which is an  88 antibody found in the mucosal defense system (Pressman & Cohen, 2005). Some researchers have argued that alterations in total sIgA levels are not a useful measure of immune functioning, because the protein functions in an antigen-specific fashion. Levels of sIgA that are specific to a particular antigen provide a much better functional measure (Stone, Cox, Valdimarsdottir, & Neale, 1987). In two longitudinal studies based on ambulatory data collection, Stone and colleagues found that specific sIgA response to a daily challenge of rabbit albumin was higher on days with high PA and lower on days with high NA (Stone, Cox, Valdimarsdottir, Jandorf, & Neale, 1987; Stone, Neale, Cox, Napoli, Valdimarsdottir, & Kennedy-Moore, 1994). This is quite strong evidence for a pure PA effect, though on an outcome that is simultaneously influenced by NA. Researchers often assume that independence between NA and PA predicts independent effects. However, a low correlation between NA and PA is in and of itself insufficient evidence for independent biological pathways. Whereas perfect bipolarity necessarily implies that NA and PA will show correlations of the same magnitude but in opposing directions, independence between these affects does not dictate unique biological correlates, as two variables can themselves be uncorrelated but show any pattern of correlations with an external variable. In other words, even if trait NA and trait PA are independent, it is possible for them to show a mixed pattern of relationships, i.e. ‘mirrored’ effects for some outcomes and divergent effects for others. Much of the extant evidence suggests that positive and negative emotions elicit physiological responses in the opposite direction, but evidence for distinct biological correlates is minimal. Clearly, there is a need for additional studies that simultaneously examine the effects of trait NA and trait PA on a range of cross-sectional and prospective biological outcomes to shed further light on this question.  89 Variability in Affect Over Time Emotional variability is considered to be an important dimension of personality that has been studied in relation to a variety of external correlates (e.g. Linville, 1982; Russell et al., 2007; Wessman & Ricks, 1966). However, besides one recent study (Simpson et al., 2008), its relation to physical health has received little attention. We exploited the repeated measures design of our study in order to explore the possibility that fluctuating patterns of affect were associated with greater metabolic symptoms relative to more stable patterns. In accordance with previous work, we defined affective variability as the intra-individual standard deviations of three variables: global PA, global NA, and affect intensity. However, we found no evidence for an association of these variables with metabolic symptoms. This may be due to insufficient differences in intra-individual variability in our sample. Inspection of the distributions of withinpersons standard deviations indicated that fluctuations in participants’ moods across time was considerably smaller compared to between-person differences in average moods. It could be that our weekly assessment method, which asked participants to rate their average mood in the last 24 hours, was not sensitive enough to detect fluctuations across time. Watson (2000) observed that PA has a distinctive daily rhythm, being low in the waking hours, increasing throughout the day, reaching its peak at midday, and then gradually declining in the evening. In contrast, NA levels are relatively stable throughout the day unless disrupted by major life events. This suggests that between-person differences in PA variability may be more evident when measured at multiple time points across the day, whereas between-person differences in NA variability are best captured using multiple assessments across longer time intervals. As such, more frequent assessments or longer time frames may be needed to capture within-person changes in mood. It is also possible that patterns of mood are more stable in younger people compared to older  90 individuals. Finally, the within-person standard deviation may not adequately represent mood variability. Larsen (1987) argues that this measure is limited because it refers to the average extremity of mood rather than frequency of mood shift. For this reason, he has advocated the use of spectral analysis to better represent this aspect of the construct. However, the number of observations in our sample falls short of the recommended minimum of 50 observations (Warner, 1998). Caveats Regarding Circumplex Structure of Mood The present investigation relied on a circumplex model of mood to differentiate between a broad range of possible moods varying in valence and activation. Overall, the pattern of correlations that emerged between the eight mood vectors did not conform well to the circular ordering predicted by a circumplex structure. Critically, vectors that theoretically represented opposite ends of the same continuum did not show the expected large negative correlations indicative of bipolarity. Also contrary to predictions, vectors of the same hedonic tone were highly correlated, suggesting that participants did not clearly distinguish between mood adjectives differing in degree of activation. Furthermore, the activated vector appeared to have positive rather than neutral valence, whereas the unactivated vector was more affectively neutral in comparison. These patterns could have influenced our capacity to detect some of the proposed associations between affect dimensions and metabolic symptoms. There are several factors that could contribute to the observed deviations from circumplex structure. One possibility that comes to mind is that this structure deteriorates when momentary affect ratings are extended over time. Our findings suggest that NA and PA averaged over time are independent rather than bipolar. In other words, a high score on mean NA does not necessarily imply a low score on mean PA. This is consistent with the prevailing  91 view that that the temporal link between NA and PA weakens when ratings are accrued across time (Diener & Emmons, 1984; Russell & Carroll, 1999). Other factors that are believed to influence the correlation between NA and PA include the specific set of items selected to represent these constructs, response format (which will influence whether respondents interpret the scale as being strictly unipolar or ambiguous), as well as random and non-random error (Russell & Carroll, 1999). The relative independence of oppositely valenced vectors and the tight clustering of likevalence terms in our sample suggests a structure that is better represented by two more or less unipolar dimensions corresponding to NA and PA. Unfortunately, unlike the large-sample studies described above, we were unable to perform a factor analysis to more conclusively assess the structure of affect in our small sample. However, the stability of factor solutions may also be susceptible to the influences considered earlier, including item selection, response format and interval, and individual difference variables. Measurement error in particular has been shown to substantially alter the observed correlations between affective ratings (e.g. Green, Goldman, & Salovey, 1993). For this reason, some researchers have advocated the use of structural equation modeling and multiple response formats to control for random and systematic error (e.g. Yik, Russell, & Feldman-Barrett; 1999). Future work could incorporate different questionnaire formats (e.g. adjectives vs. statement lists) with variable response options (e.g. degree of agreement vs. degree to which an item describes one’s feelings), and use confirmatory factor analysis to estimate correlations between latent mood dimensions, while simultaneously estimating the contribution of random and method-specific errors. Limitations and Future Directions The participants in this study were generally representative of the greater Vancouver area  92 in terms of ethnicity and socioeconomic diversity. As such, we have reason to believe that the findings can be generalized to the local population. Research suggests that, despite some culture-specific aspects, the essential structure of emotions is relatively invariant across different cultures (Church, Katigbak, Reyes, & Jensen, 1998; Fontaine, Poortinga, Setiadi, & Markam; 2004; Russell, Lewicka, & Niit, 1989; van Hemert, Poortinga, & van de Vijver, 2007; Watson, Clark, & Tellegen, 1984). On the other hand, the generalizability of our findings may be limited by other characteristics of the current sample. First, since participants in this study were chosen to be at risk for depression, we cannot be sure that the emotional and biological characteristics of our sample are typical of a low-risk population. Thus, our findings may not be pertinent to nonvulnerable individuals. Second, because our study was conducted with a young and exclusively female sample, the results cannot be generalized to males or older adults. Future studies could examine the relationship between mood and biological outcomes in more diverse, mixed-sex samples. Third, restriction of range on biological outcomes due to a young and healthy sample may have contributed to some of our null findings. We have alluded to the possibility that the effects of PA may not transpire until middle or late adulthood, when risk for illness is higher. Recent evidence suggests that some of the relationships between negative emotionality and metabolic symptoms may also be moderated by age, suggesting that these links may take time to develop (Bunde & Suls, 2007). Although our sample comprised of youth at various stages of development, ranging from middle adolescence to young adulthood, the potential moderating effects of these different developmental phases were not examined in the present investigation. Long-term prospective studies that follow youngsters over the course of several years would help to explicate how the relationships between affect and health change over time. In addition, the present investigation assessed a  93 limited set of biological outcomes associated with risk for cardiovascular disease and diabetes. As such, we cannot be sure that other processes relevant to health are more susceptible to the influences of affect, or respond differentially to NA and PA. Future studies could explore these ideas by assessing a more diverse set of outcomes representing a variety of biological systems. The apparent deviations of our mood data from the circumplex structure also posed interpretive challenges. While core affect at a particular slice in time may be adequately described by a two-dimensional structure, other structures may provide a better representation of temperamental affect. As Russell and Feldman-Barrett (1999) have stressed, the psychometrics of mood can become counterintuitive when time is involved. Alternatively, as a consequence of our eligibility criteria for study entry, we may have inadvertently sampled individuals whose subjective experience of mood was not well represented by a circumplex structure. Due to the limited size of our sample, we were unable to utilize factor analysis or other statistical tests of circumplexity to reliably assess the structure of our mood data. While this was not a primary goal of the current investigation, there is a definite need for studies that attempt to clarify the structure of affect extended over time. Elucidating this structure could be instrumental in guiding more specific hypotheses about the role of emotions in health. The results of our study may also be limited by our method of mood assessment. In order to reduce participant burden, we constrained the mood diary to a limited set of adjectives. While our measure was based on a previously validated instrument, it is possible that we missed the key emotions that are important for physical health. Other methods of assessment exist that may yield more variability in mood data without undue burden to participants. For example, the circular mood scale (Jacob, Simons, Manuck, & Rohay, 1989) and the affect grid (Russell, Weiss, & Mendelsohn, 1989) provide participants with a visual representation of the circumplex  94 space on which they can mark the location that best represents their current mood. However, this method forces participants to rate their predominant mood at a single point on the circumplex, whereas multi-item mood scales allow for the simultaneous assessment of multiple moods representing different locations on the circumplex. In addition, analysis of the resultant data set would entail a different methodological approach based on circular statistics. Nevertheless, it would be interesting to compare the results of our investigation with results obtained from these alternate methods of assessment. Our reliance on experience sampling to assess mood repeatedly over time also has its inherent disadvantages. This method is highly dependent on participant cooperation. Due to the time commitment involved, some participants were unwilling or unable to complete their diaries on a regular basis. Event sampling is also susceptible to reactivity and priming effects, which may alter the way in which participants perceive and respond to items. We also relied on participants to decide when to complete their diary entries. Ambulatory data collection based on randomly scheduled signals may have allowed us to capture greater variability in moods and decrease respondent bias, though this method has the potential to be more intrusive. Further, although weekly assessments over a 6-month period provides a more reliable assessment of average affect compared to one-time retrospective reports, more frequent assessments may be necessary to adequately assess the frequency of intra-individual mood shifts. Another key avenue for future work would be to use findings from neurobiological studies to inform hypotheses on potential biological pathways linking emotions to health. Understanding how regional neural activity translates into peripheral biological activity may provide clues to unique and common pathways. The neural mechanisms underlying emotions are likely to be more complex and intertwined than the emotions themselves (Russell &  95 Feldman-Barrett, 1999). As such, it may be that independent pathways are best defined not by activity at a single biological endpoint, but by differences in patterns of activity among multiple biological outcomes. Another limitation of this study is the relatively small sample size of the cohort. While the size of our sample was adequate for detecting effect sizes greater than 0.30 with 80% power, which is the norm in work on affect and health, it was insufficient for detecting effects of lower magnitude. Finally, although our study suggests temporal precedence for mood predicting biological outcomes, it does not establish causality. Unmeasured third variables, such as genes, may contribute both to greater negative emotionality and poorer health status. Experimental studies would be needed to determine whether negative affective tendencies are causally related to metabolic symptoms or other biological measures. Thus an important avenue for future research would be to evaluate the impact of interventions targeting NA on health-related outcomes.  96 CONCLUSION The findings of this research study revealed that negative affective tendencies predict less optimal metabolic symptoms in physically healthy young females, whereas positive affective tendencies have neither a favorable nor detrimental effect on these outcomes. The results also suggest that the relationship between negative affect and health is not moderated by differences in activation, which suggests that negative moods may have common or synergistic rather than unique effects on health. These findings imply that interventions aimed at reducing NA rather than enhancing PA are likely to be more effective in reducing risk for metabolic syndrome. While we found no evidence that NA and PA are associated with different biological correlates, the fact that these affective constructs were relatively independent and did not show mirrored effects is suggestive of the existence of distinctive mechanisms. The current study, though not without its limitations, is a step towards identifying potential pathways through which basic emotional styles in healthy young individuals may set the stage for the development of disease in later life. In view of some largely inexplicable findings, the ideas that have been proposed in this discussion are necessarily speculative and are meant to offer new hypotheses to be pursued in future studies. The present investigation highlights the need for prospective designs that follow young and healthy individuals over an extended period. In addition to collecting a broad range of biological measures at various time points, these studies should ideally re-assess affective tendencies at different life stages, as predominate emotions may change with age and experience. In this way, it may be possible to observe how the relationships between affective tendencies and physiological markers of health evolve over the life span.  97 REFERENCES Abramson, L. Y., Alloy, L. B., Hogan, M. E., Whitehouse, W. G., Donovan, P., Rose, D., Panzarella, C., & Raniere, D. (1999). Cognitive vulnerability to depression: Theory and evidence. Journal of Cognitive Psychotherapy: An International Quarterly, 13, 5-20. Affleck, G., Apter, A., Tennen, H., Reisine, S., Barrows, E., Willard, A., Unger, J., & ZuWallack, R. (2000). Mood states associated with transitory changes in asthma symptoms and peak expiratory flow. Psychosomatic Medicine, 62, 61-68. Ahern, D. K., Gorkin, L., Anderson, J. L., Tierney, C., Hallstrom, A., Ewart, C., Capone, R. J., Schron, E., Kornfeld, D., Herd, J. A., Richardson, D. W., & Follick, M . J. (1990). Biobehavioral variables and mortality or cardiac arrest in the Cardiac Arrhythmia Pilot Study (CAPS). American Journal of Cardiology, 66, 59-62. Arslanian, S., & Suprasongsin, C. (1996). Insulin sensitivity, lipids, and body composition in childhood: Is syndrome X present? Journal of Clinical Endocrinology and Metabolism, 81, 1058–1062. Bacon, S. L., Watkins, L. L., Babyak, M., Sherwood, A., Hayano, J., Hinderliter, A. L., et al. (2004). Effects of daily stress on autonomic cardiac control in patients with coronary artery disease. American Journal of Cardiology, 93, 1292–1294. Basso, M. R., Schefft, B. K., Ris, M.D., & Dember, W. N. (1996). Mood and global-local visual processing. Journal of the International Neuropsychological Society, 2, 249-255. Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory for measuring depression. Archives of General Psychiatry, 4, 561-571.  98 Berk, L. S., Tan, S. A., Fry, W. F., Napier, B. J., Lee, J. W., Hubbard, R. W., et al. (1989). Neuroendocrine and stress hormone changes during mirthful laughter. American Journal of the Medical Sciences, 298, 390–396. Bhattacharyya, M. R., Whitehead, D. L., Rakhit, R., & Steptoe, A. (2008). Depressed mood, positive affect, and heart rate variability in patients with suspected coronary artery disease. Psychosomatic Medicine, 70, 1020-1027. Bleil, M. E., Gianaros, P. J., Jennings, J. R., Flory, J. D., & Manuck, S. B. (2008). Trait negative affect: Toward an integrated model of understanding psychological risk for impairment in cardiac autonomic function. Psychosomatic Medicine, 70, 328-337. Brickman, P., Coates, D., & Janoff-Bulman, R. (1978). Lottery winners and accident victims: Is happiness relative? Journal of Personality and Social Psychology, 36, 917–927. Brown, W. A., Sirota, A. D., Niaura, R., & Engebretson, T. O. (1993). Endocrine correlates of sadness and elation. Psychosomatic Medicine, 55, 458–467. Brosschot, J. F., & Thayer, J. F. (2003). Heart rate response is longer after negative emotions than after positive emotions. Heart rate response is longer after negative emotions than after positive emotions. International Journal of Psychophysiology, 50, 181-187. Bunde, J., & Suls, J. (2006). A quantitative analysis of the relationship between the CookMedley Hostility Scale and traditional coronary artery disease risk factors. Health Psychology, 25, 493-500. Burack, J. H., Barrett, D. C., Stall, R. D., Chesney, M. A., Ekstrand, M. L., & Coates, T. J. (1993). Depressive symptoms and CD4 lymphocyte decline among HIV-infected men. Journal of the American Medical Association, 270, 2568-2573.  99 Caminero, A. G., Blumentals, W. A., Russo, L. J., Brown, R. R., & Castilla-Puentes, R. (2005). Does Panic Disorder Increase the Risk of Coronary Heart Disease? A Cohort Study of a National Managed Care Database. Psychosomatic Medicine, 67, 688-691. Carney, R. M., & Freedland, K. E. (2003). Depression, mortality, and medical morbidity in patients with coronary heart disease. Biological Psychiatry, 54, 241-247. Carver, C. S., & White, T. L. (1994). Behavioral inhibition, behavioral activation, and affective responses to impending reward and punishment: The BIS/BAS Scales. Journal of Personality and Social Psychology, 67, 319-333. Catipovic-Veselica, K., Amidzic, V., Durijancek, J., Kozmar, D., Sram, M., Glavas, B., & Catipovic, B. (1999). Association of heart rate and heart-rate variability with scores on the Emotion Profile Index in patients with acute coronary heart disease. Psychological Reports, 84, 433–442. Chen, W., Bao, W., Begum, S., Elkasabany, A., Srinivasan, S. R., & Berenson, G. S. (2000). Age-related patterns of clustering of cardiovascular risk variables of syndrome X from childhood to young adulthood in a population made up of black and white subjects: The Bogalusa Heart Study. Diabetes, 49, 1042–1048. Chen, W., Srinivasan, S. R., Elkasabany, A., & Berenson, G. S. (1999a). Cardiovascular risk factors clustering features of insulin resistance syndrome (syndrome X) in a biracial (black-white) population of children, adolescents and young adults. American Journal of Epidemiology, 150, 667–674. Chida, Y., & Steptoe, A. (2008) Positive psychological well-being and mortality: A quantitative review of prospective observational studies. Psychosomatic Medicine, 70, 741-756.  100 Church, A. T., Katigbak, M. S., Reyes, J. A. S., & Jensen, S. M. (1998). Language and organization of Filipino emotion concepts: Comparing emotion concepts and dimensions across cultures. Cognition and Emotion, 12, 63-92. Clark, L. A., & Watson, D. (1991). Tripartite model of anxiety and depression: Psychometric evidence and taxonomic implications. Journal of Abnormal Psychology, 100, 316-366. Codispoti, M., Gerra, G., Montebarocci, O., Zaimovic, A., Raggi, M. A., & Baldaro, B. (2003). Emotional perception and neuroendocrine changes. Psychophysiology, 40, 863–868. Cohen, S., Doyle, W. J., Skoner, D. P., Fireman, P., Gwaltney, J. M., & Newsom, J. T. (1995). State and trait negative affect as predictors of objective and subjective symptoms of respiratory viral infections. Journal of Personality and Social Psychology, 68, 159-169. Cohen, S., Doyle, W. J., Turner, R. B., Alper, C. M., & Skoner, D. P. (2003). Emotional style and susceptibility to the common cold. Psychosomatic Medicine, 65, 652-657. Cook, W. W., & Medley, D. M. (1954). Proposed hostility and pharisaic-virtue scales for the MMPI. Journal of Applied Psychology, 38, 414-418. Cook D. G., Mendall, M. A., Whincup, P. H., Carey, I. M., Ballam, L., Morris, J. E., Miller, G. J., Strachan, D. P. (2000). C-reactive protein concentration in children: relationship to adiposity and other cardiovascular risk factors. Atherosclerosis, 149, 139-50. Csabi, G., Török, K., Jeges, S., & Molnar, D. (2000). Presence of metabolic cardiovascular syndrome in obese children. European Journal of Pediatrics, 159, 91–94. Danner, D. D., Snowdon, D. A., & Friesen, W. V. (2001). Positive emotions in early life and longevity: Findings from the nun study. Journal of Personality and Social Psychology, 80, 804-813.  101 Davidson, R. J. (1998). Affective style and affective disorders: Perspectives from affective neuroscience. Cognition and Emotion, 12, 307-330. DeFronzo, R.A., & Ferrannini, E. (1991). Insulin resistance. A multifaceted syndrome responsible for NIDDM, obesity, hypertension, dyslipidemia, and atherosclerotic cardiovascular disease. Diabetes Care, 14, 173–194. de Groot, M., Anderson, R., Freedland, K. E., Clouse, R. E., & Lustman, P. J. (2001). Association of depression and diabetes complications: A meta-analysis. Psychosomatic Medicine, 63, 619-630. Depue, R. A., Luciana, M., Arbisi, P., Collins, P., & Leon, A. (1994). Dopamine and the structure of personality: Relation of agonist-induced dopamine activity to positive emotionality. Journal of Personality and Social Psychology, 67, 485-498. Diana, M., Pistis, M., Muntoni, A., & Gessa, G. (1996). Mesolimbic dopaminergic reduction outlasts ethanol withdrawal syndrome: Evidence of protracted abstinence. Neuroscience, 71, 411-415. Diener, E., & Emmons, R. A. (1984). The Independence of Positive and Negative Affect. Journal of Personality and Social Psychology, 47, 1105-1117. Diener, E., Smith, H., & Fujita (1995). The Personality Structure of Affect. Journal of Personality and Social Psychology, 69, 130-141. Eaker, F. D., Sullivan, L. M., Kelly-Hayes, M., D’Agostino, R. B. Sr., & Benjamin, E. J. (2005). Tension and Anxiety and the Prediction of the 10-Year Incidence of Coronary Heart Disease, Atrial Fibrillation, and Total Mortality: The Framingham Offspring Study. Psychosomatic Medicine, 67, 692-696  102 Eaton, W. W., Armenian, H., Gallo, J., Pratt, L., & Ford, D. E. (1996). Depression and risk for onset of type II diabetes. A prospective population-based study. Diabetes Care, 19, 10971102. Eizenman, D. R., Nesselroade, J. R., Featherman, D. L., & Rowe, J. W. (1997). Intraindividual variability in perceived control in a older sample: The MacArthur successful aging studies. Psychology and Aging, 12, 489-502. Ellsworth, P. C, & Smith, C. A. (1988a). From appraisal to emotion: Differences among unpleasant feelings. Motivation and Emotion, 2, 271-302. Ellsworth, P. C, & Smith, C. A. (1988b). Shades of joy: Patterns of appraisal differentiating pleasant emotions. Cognition and Emotion, 2, 301-331. Eng, P. M., Fitzmaurice, G., Kubzansky, L. D., Rimm, E. B., & Kawachi, I. (2003). Anger expression and risk of stroke and coronary heart disease among male health professionals. Psychosomatic Medicine, 65, 100-110. Everson, S. A., Kaplan, G. A., Goldberg, D. E., Lakka, T. A., Sivenius, J., & Salonen, J. T. (1999). Anger expression and incident stroke: Prospective evidence from the Kuopio ischemic heart disease study. Stroke, 30, 523-528. Ewart, C. K., & Kolodner, K. B. (1994). Negative affect, gender, and expressive style predict elevated ambulatory blood pressure in adolescents. Journal of Personality and Social Psychology, 66, 596-605. Feldman, L. A. (1995). Valence focus and arousal: Individual differences in the structure of affective experience. Journal of Personality and Social Psychology, 69, 153-166. Feldman-Barrett, L., & Russell, J. A. (1998). Independence and bipolarity in the structure of current affect. Journal of Personality and Social Psychology, 74, 967-984.  103 First, M., Spitzer, R., Gibbon, M., & Williams, J. (2002). Structured Clinical Interview for DSMIV-TR Axis I Disorders, Research Version, Non-patient Edition. New York: Biometrics Research. Fisher, G. A., Heise, D. R., Bohrnstedt, G. W., & Lucke, J. F. (1985). Evidence for extending the circumplex model of personality trait language to self-reported moods. Journal of Personality and Social Psychology, 49, 233-242. Fontaine, J. R. J., Poortinga, Y. H., Setiadi, B., & Markam. S. S. (2004). Cognitive structure of emotion terms in Indonesia and The Netherlands. Cognition and Emotion, 16, 61-86. Ford, E. S. (1999). Body mass index, diabetes, and C-reactive protein among U.S. adults. Diabetes Care, 22, 1971-1977. Frasure-Smith, N., Lesperance, F., & Talajic, M. (1995). The impact of negative emotions on prognosis following myocardial infarction: Is it more than depression? Health Psychology, 14, 388-398. Frasure-Smith, N., & Lesperance, F. (2003). Depression and other psychological risks following myocardial infarction. Archives of General Psychiatry, 60, 627-636. Frazier, T. W., Strauss, M. E., & Steinhauer, S. R. (2004). Respiratory sinus arrhythmia as an index of emotional response in young adults. Psychophysiology, 41, 75–83. Fredrickson, B.L. (1998). What good are positive emotions? Review of General Psychology, 3, 300-319. Fredrickson, B. L., & Branigan, C. (2005). Positive emotions broaden the scope of attention and thought-action repertoires. Cognition and Emotion, 18, 313-332. Fredrickson, B. L., & Levenson, R. W. (1998). Positive emotions speed recovery from the cardiovascular sequelae of negative emotions. Cognition andEmotion, 12, 191-220.  104 Friedman, H. S., Tucker, J. S., Tomlinson-Keasey, C., Schwartz, J. E., Wingard, D. L., & Criqui, M. H. (1993). Does childhood personality predict longevity? Journal of Personality and Social Psychology, 65, 176-185. Gallo, L. C., & Matthews, K. (2003). Understanding the association between socioeconomic status and physical health: Do negative motions play a role? Psychological Bulletin, 129, 10-51. Goldbacher, E. M., & Matthews, K. A. (2007). Are psychological characteristics related to risk of the metabolic syndrome? A review of the literature. Annals of Behavioral Medicine, 34, 240-252. Golden, S. H., Williams, J. E., Ford, D. E., Yeh, H., Sanford, C. P., Nieto, F. J., & Brancati, F. L. (2006). Anger temperament is modestly associated with the risk of type 2 diabetes mellitus: The atheroslcerosis risk in communities study. Psychoneuroendocrinology, 31, 325-332. Gray, J. A. (1987). Perspectives on anxiety and impulsivity: A commentary. Journal of Research in Personality, 21, 493-509. Green, D. P., Goldman, S. L., & Salovey, P. (1993). Measurement error masks bipolarity in affect ratings. Journal of Personality and Social Psychology, 64, 1029-1041. Haney, T., Maynard, K., Houseworth, S., Scherwitz, L., Williams, R. B., & Barefoot, J. (1996). Interpersonal hostility assessment technique: Description and validation against the criterion of coronary artery disease. Journal of Personality Assessment, 66, 386-401. Heponiemi, T., Ravaja, N., Elovainio, M., & Keltikangas-Järvinen, L. (2006). Experiencing positive affect and negative affect during stress: Relationships to cardiac reactivity and to facial expressions. Scandinavian Journal of Psychology, 47, 327-337.  105 Howell, R. T., Kern, M. L., & Lyubomorsky, S. (2007). Health Benefits: meta-analytically determining the impact of well-being on objective health outcomes. Health Psychology Review, 1, 83-136. Ironson, G., Balbin, E., & Stuetzle, R. (2005) Dispositional Optimism and the Mechanisms by Which It Predicts Slower Disease Progression in HIV: Proactive Behavior, Avoidant Coping, and Depression. International Journal of Behavioral Medicine, 12, 86-97. Isen, A. M., Daubman, K. A., & Nowicki, G. P. (1987). Positive affect facilitates creative problem solving. Journal of Personality and Social Psychology, 52, 1122-1131. Isen, A. M., Johnson, M. M. S., Mertz, E., & Robinson, G. F. (1985). The influence of positive affect on the unusualness of word associations. Journal of Personality and Social Psychology, 48, 1413-1426. Jacob, R. G., Simons, A. D., Manuck, S. B., & Rohay, J. M. (1989). The Circular Mood Scale: A new technique of measuring ambulatory mood. Journal of Psychopathology and Behavioral Assessment, 11, 153-173. Janoff-Bulman, R., & Marshall, G. (1982). Mortality, well-being, and control: A study of a population of institutionalized aged. Personality & Social Psychology Bulletin, 8, 691698. Jones, B. E. (2003). Arousal systems. Frontiers in Bioscience, 8, 438-451. Kalichman, S. C., Sikkema, K. J., & Somlai, A. (1995). Assessing persons with Human Immunodeficiency Virus (HIV) infection using the Beck Depression Inventory: Disease processes and other potential confounds. Journal of Personality Assessment, 64, 86-100. Kamarck, T., Shiffman, S., Smithline, L., Goodie, J., Paty, J., Gnys, M., & Jong, J. (1998). Effects of task strain, social conflict, and emotional activation on ambulatory  106 cardiovascular activity: Daily life consequences of recurring stress in a multiethnic adult sample. Health Psychology, 17, 17–29. Kaplan, G. A., & Camacho, T. (1983). Perceived health and mortality: A nine-year follow-up of the human population laboratory cohort. American Journal of Epidemiology, 117, 292304. Katzmarzyk, P. T., Perusse, L., Malina, R. M., Bergeron, J., Despres, J.-P., & Bouchard, C. (2001). Stability of indicators of the metabolic syndrome from childhood and adolescence to young adulthood: The Quebec Family Study. Journal of Clinical Epidemiology, 54, 190–195. Kawakami, N., Tkatsuka, N., Shimuzu, H., & Ishibashi, H. (1999). Depressive symptoms and occurrence of type 2 diabetes among Japanese men. Diabetes Care, 22, 1071-1076. Kawamoto, R., & Doi, T. (2002). Self-reported functional ability predicts three-year mobility and mortality in community-dwelling older persons. Geriatrics and Gerontology International, 2, 68-74. Kemeny, M. E., & Shestyuk, A. (2008). Emotions, the neuroendocrine and immune systems, and health. In M. Lewis, J. M. Haviland-Jones, & L. Feldman-Barrett (Eds.), Handbook of Emotions (pp. 661–675). New York, NY: The Guilford Press. Kiecolt-Glaser, J. K., McGure, L., Robles, T. F., & Glaser, R. (2002). Emotions, morbidity, and mortality: New perspectives from psychoneuroimmunology. Annual Reviews of Psychology, 53, 83-107. Koivumaa-Honkanen, H., Honkanen, R., Viinamaeki, H., Heikkila, K., Kaprio, J., & Koskenvuo, M. (2000). Self-reported life satisfaction and 20-year mortality in healthy Finnish adults. American Journal of Epidemiology, 152, 983-991.  107 Kubzansky, L. D., Cole, S. R., Kawachi, I., Vokonas, P., & Sparrow, D. (2006). Shared and unique contributions of anger, anxiety, and depression to coronary heart disease: A prospective study in the normative aging study. Annals of Behavioral Medicine, 31, 2129. Kubzansky, L. D., & Kawachi, I. (2000). Going to the heart of the matter: Do negative emotions cause coronary heart disease? Journal of Psychosomatic Research, 48, 323-337. Kubzansky, L. D., Sparrow, D., Vokonas, P., & Kawachi, I. (2001). Is the glass half empty or half full? A prospective study of optimism and coronary heart disease in the normative aging study. Psychosomatic Medicine,63, 910-916. Larsen, R. J. (1987). The stability of mood variability: A spectral analytic approach to daily mood assessments. Journal of Personality and Social Psychology, 52, 1195-1204. Larsen, J. T., Berntson, G. G., Poehlmann, K. M., Ito, T. A., & Cacioppo, J. T. (2008). The psychophysiology of emotion. In M. Lewis, J. M. Haviland-Jones, & L. Feldman-Barrett (Eds.), Handbook of Emotions (pp.801–195). New York, NY: The Guilford Press. Larsen, R. J., & Diener, E. (1992). Promises and problems with the circumplex model of emotion. In M. S. Clark (Ed.), Emotion (pp. 25–59). Thousand Oaks, CA: Sage. Lazarus, R. S. (1991). Emotion and adaptation. New York: Oxford University Press. Leserman, J. (2003). HIV disease progression: Depression, stress, and possible mechanisms. Biological Psychiatry, 54, 295-306. Leserman, J., Jackson, E. D., Petitto, J. M., Golden, R. N., Silva, S. G., Perkins, D.O., Cai, J., Folds, J. D., & Evans, D. L. (1999). Progression to AIDS: The effects of stress, depressive symptoms, and social support. Psychosomatic Medicine, 61, 397-406.  108 Leserman, J., Petitto, J. M., Gu, H., Gaynes, B. N., Barroso, J., Golden, R. N., Perkins, D. O., Cai, J., Folds, J. D., & Evans, D. L. (2002). Progression to AIDS, a clinical AIDS condition, and mortality: Psychosocial and physiological predictors. Psychological Medicine, 32, 1059-1073. Leserman, J., Petitto, J. M., Perkins, D. O., Folds, J. D., Golden, R. N. & Evans, D.L. (1997). Severe stress, depressive symptoms, and changes in lymphocyte subsets in human immunodeficiency virus-infected men. Archives of General Psychiatry 54, 279-285. Levy, B. R., Slade, M. D., Kunkel, S. R., & Kasl, S. V. (2002). Longevity increased by positive self-perceptions of aging. Journal of Personality and Social Psychology, 83, 261-270. Linkowski, P. (2003). Neuroendocrine profiles in mood disorders. International Journal of Neuropsychopharmacology, 6, 191-197. Linville, P. W. (1982). Affective consequences of complexity regarding the self and others. In M. Clark & S. Fiske (Eds.), Affect and cognition: Seventeenth annual symptosium on cognition (pp. 79-109). Hillsdale, NJ: Erlbaum. Lustman, P. J., Anderson, R. J., Freedland, K. E., de Groot, M., Carney, R. M., & Clouse, R. E. (2000). Depression and poor glycemic control: a meta-analytic review of the literature. Diabetes Care, 23, 934-942. Maier, H., & Smith, J. (1999). Psychological predictors of mortality in old age. Journal of Gerontology: Psychological Sciences, 54B, P44-P54. Matsunaga, M., Isowa, T., Kimura, K., Miyakoshi, M., Kanayama, N., Murakami, H., et al. (2009). Associations among positive mood, brain, and cardiovascular activities in an affectively positive situation. Brain Research, 1263, 93-103.  109 Matthews, K. A. (2005). Psychological perspectives on the development of coronary heart disease. American Psychologist, 60, 783-796. Mayne, T. J., Vittinghoff, E., Chesney, M. A., Barrett, D. C., & Coates, T. J. (1996). Depressive affect and survival among gay and bisexual men infected with HIV. Archives of Internal Medicine, 156, 2233-2238. McCraty, R., Atkinson, M., Tiller, W. A., Rein, G., & Watkins, A. D. (1995). The effects of emotions on short-term power spectrum analysis of heart rate variability. American Journal of Cardiology, 76, 1089– 1093. McNair, D. M., Lorr, M., & Droppleman, L. F. (1971). Profile of Mood States. San Diego, CA: Educational and Industrial Testing Service. Mendes de Leon, C. F., Kop, W. K., de Stuart, H. B., Bar, F. W., & Appels, A. P. (1996). Psychosocial characteristics and recurrent events after percutaneous transluminal coronary angioplasty. American Journal of Cardiology, 77, 252-255. Middleton, R. A., & Byrd, E. K. (1996). Psychosocial factors and hospital readmission status of older persons with cardiovascular disease. Journal of Applied Rehabilitation Counseling, 27, 3-10. Miller, G. E., & Blackwell, E. (2006). Turning up the heat: Inflammation as a mechanism linking chronic stress, depression, and heart disease. Current Directions in Psychological Science, 15, 269-272. Miller, T. Q., Smith, T. W., Turner, C. W., Guijarro, M. L., & Hallet, A. J. (1996). A metaanalytic review on hostility and physical health. Psychological Bulletin, 119, 322-348.  110 Miller, G. E., Stetler, C. A., Carney, R. M., Freedland, K. E., & Banks, W. A. (2002). Clinical depression and inflammatory risk markers for coronary heart disease. American Journal of Cardiology, 90, 1279–1283. Miller, B. D., & Wood, B. L. (1997). Influence of specific emotional states on autonomic reactivity and pulmonary function in asthmatic children. Journal of the American Academy of Child & Adolescent Psychiatry, 36, 669–677. National Cholesterol Education Program. (2001). Detection, evaluation and treatment of high blood cholesterol in adults (Adult treatment panel III). (NIH Publication No. 01–3670). Washington, DC: U.S. Government Printing Office. O’Connor, B. P., & Vallerand, R. J. (1998). Psychological adjustment variables as predictors of mortality among nursing home residents. Psychology and Aging, 13, 368-374. Ong, A. D., & Allaire, J.C., Cardiovascular intraindividual variability in later life: The influence of social connectedness and positive emotions. Psychology of Aging, 20, 476-485. Ostir, G. V., Markides, K. S., Black, S. A., & Goodwin, J. S. (2000). Emotional well-being predicts subsequent functional independence and survival. Journal of the American Geriatrics Society, 48, 473-478. Ostir, G. V., Markides, K. S., Peek, M. K., & Goodwin, J. S. (2001). The association between emotional well-being and the incidence of stroke in older adults. Psychosomatic Medicine, 63, 210-215. Palmore, E. B. (1969). Predicting longevity: A follow-up controlling for age. Gerontologist, 9, 247-250. Parker, K. J., Schatzberg, A. F., & Lyons, D. M. (2003). Neuroendocrine aspects of hypercortisolism in major depression. Hormones and Behavior, 43, 60-66.  111 Parker, M. G., Thorslund, M., & Nordstrom, M. L. (1992). Predictors of mortality for the oldest old. A 4-year follow-up of community-based elderly in Sweden. Archives of Gerontology and Geriatrics, 14, 227-237. Pollard, T. M., & Schwartz, J. E. (2003). Are changes in blood pressure and total cholesterol related to changes in mood? An 18-month study of men and women. Health Psychology, 22, 47-53. Pollock, V., Cho, D. W., Reker, D., & Volavka, J. (1979). Profile of mood states: The factors and their physiological correlates. Journal of Nervous and Mental Disease, 167, 612–614. Posner, J., Russell, J. A., & Peterson, B. S. (2005). The circumplex model of affect: An integrative approach to affective neuroscience, cognitive development, and psychopathology. Development and Psychopathology, 17, 715-734. Pressman, S. D., & Cohen, S. (2005). Does Positive Affect Influence Health? Psychological Bulletin, 131, 925-971. Rabkin, J. G., Williams, J. B. W., Remien, R. H., Goetz, R. R., Dertzner, R., & Gorman, J. M. (1991). Depression, distress, lymphocyte subsets, and human immunodeficiency virus symptoms on two occasions in HIV-positive homosexual men. Archives of General Psychiatry, 48, 111-119. Radloff, L. S. (1977). The CES-D Scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385-401. Rauch, S. L., Shin, L. M., & Wright, C. I. (2003). Neuroimaging studies of amygdale function in anxiety disorders. Annals of the New York Academy of Science, 985, 389-410. Richman, L. S., Kubzansky, L., Maselko, J., Kawachi, I., Choo, P., & Bauer, M. (2005). Positive emotion and health: going beyond the negative. Health Psychology, 24, 422-429.  112 Ridker, P. M., Buring, J. E., Shih, J., Matias, M., Hennekens, C. H. (1998). Prospective study of C-reactive protein and the risk of future cardiovascular events among apparently healthy women. Circulation, 98, 731-733. Ridker, P. M., Glynn, R. J., Hennekens, C. H. (1998). C-reactive protein adds to the predictive value of total and HDL cholesterol in determining risk of first myocardial infarction. Circulation, 97, 2007-2011. Rugulies, R. (2002). Depression as a predictor for coronary heart disease: A review and metaanalysis. American Journal of Preventive Medicine, 23, 51-61. Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39, 1161-1178. Russell, J. A., & Carroll, J. M. (1999). On the bipolarity of positive and negative affect. Psychological Bulletin, 125, 3-30. Russell, J. A., Feldman-Barrett, L. (1999). Core affect, prototypical emotional episodes, and other things called emotion: Dissecting the elephant. Journal of Personality and Social Psychology, 76, May 805-819. Russell, J. A., Lewicka, M., & Niit, T. (1989). A cross-cultural study of a circumplex model of affect. Journal of Personality and Social Psychology, 57, 848-856. Russell, J. A., Weiss, A., & Mendelsohn, G. A. (1989). Affect grid: A single-item scale of pleasure and arousal. Journal of Personality and Social Psychology, 579, 493-502. Russell, J. J., Moskowitz, D. S., Zuroff, D. C., Sookman, D., & Paris, J. (2007). Stability of variability of affective experience and interpersonal behavior in borderline personality disorder. Journal of Abnormal Psychology, 116, 578-588.  113 Sapolsky, R. M. (1998). Why zebras don’t get ulcers: An updated guide to stress, stress-related diseases, and coping. New York, NY: W. H. Freeman and Company. Sassin, J. F., Frantz, A. G., Weitzman, E. D., & Kapen, S. (1972). Human prolactin: 24-hour pattern with increased release during sleep. Science, 177, 1205-1207. Scheier, M. F., Matthews, K. A., Owens, J. F., Magovern, G. J. Sr., Lefebvre, R. C., Abbott, R. A., & Carver, C. S. (1989). Dispositional optimism and recovery from coronary artery bypass surgery: the beneficial effects on physical and psychological well-being. Journal of Personality & Social Psychology, 57, 1024-1040. Segerstrom, S. C., & Miller, G. E. (2004). Psychological stress and the human immune system: A meta-analytic study of 30 years of inquiry. Psychological Bulletin, 130, 601-630. Shoda, Y., Mischel, W., & Wright, J. C. (1994). Intraindividual stability in the organization and patterning of behavior: Incorporating psychological situations into the idiographic analysis of personality. Journal of Personality and Social Psychology, 67, 674-687. Simpson, E. E. A., McConville, C., Rae, G., O’Connor, J. M., Stewart-Knox, B. J., Coudray, C., et al. (2008). Salivary cortisol, stress and mood in healthy older adults: The Zenith study. Biological Psychology, 78, 1-9. Smith, A. W., & Baum, A. (2003). The influence of psychological factors on restorative function in health and illness. In J. Suls & K. A. Wallston (Eds.), Social psychological foundations of health and illness (pp. 431-457). Malden, MA, US: Blackwell Publishers. Smith, T. W., & Ruiz, J. M. (2002). Psychosocial influences on the development and course of coronary heart disease: Current status and implications for research and practice. Journal of Consulting and Clinical Psychology, 70, 548-568. Smyth, J., Ockenfels, M. C., Porter, L., Kirschbaum, C., Hellhammer, D. H., & Stone, A. A.  114 (1998). Stressors and mood measured on a momentary basis are associated with salivary cortisol secretion. Psychoneuroendocrinology, 23, 353–370. Spiegel, D., & Giese-Davis, J. (2003). Depression and cancer: mechanisms and disease progression. Biological Psychiatry, 54, 269-282. Spielberger, C. D., Jacobs, G., Russell, S., & Crane, R. S. (1983). Assessment of anger: The State-Trait Scale. In J. N. Butcher & C. D. Spielberger (Eds.), Advances in personality assessment (Vol. 2, pp. 161-189). Hillsdale, NJ: Erlbaum. Spielberger, C. D., Johnson, E. H., Russell, S. F., Crane, R. S., Jacobs, G. A., & Worden, T. J. (1985). The experience and expression of anger: Construction of an anger expression scale. In M. Chesney & R. H. Rosenman, (Eds.), Anger and hostility in cardiovascular and behavioral disorders (pp. 5-30). New York: Hemisphere/McGraw-Hill. Steptoe, A., & Wardle, J. (2005). Positive affect and biological function in everyday life. Neurobiology of Aging. Special Issue: Aging, Diabetes, Obesity, Mood and Cognition, 26, 108-112. Stetler, C., & Miller, G. E. (2005). Blunted cortisol response to awakening in mild to moderate depression: Regulatory influences of sleep patterns and social contacts.  Journal of  Abnormal Psychology, 114, 697-705. Stone, A. A., Cox, D. S., Valdimarsdottir, H., Jandorf, L., & Neale, J. M. (1987). Evidence that Secretory IgA antibody is associated with daily mood. Journal of Personality and Social Psychology, 52, 988–993. Stone, A. A., Cox, D. S., Valdimarsdottir, H., & Neale, J. M. (1987). Secretory IgA as a measure of immunocompetence. Journal of Human Stress, 13, 136-140. Stone, A. A., Neale, J. M., Cox, D. S., Napoli, A., Valdimarsdottir, H., & Kennedy-Moore, E.  115 (1994). Daily events are associated with a secretory immune response to an oral antigen in men. Health Psychology, 13, 440–446. Stones, M. J., Dornan, B., & Kozma, A. (1989). The prediction of mortality in elderly institution residents. Journals of Gerontology, 44(3), P72-P79. Suls, J., & Bunde, J. (2005). Anger, anxiety, and depression as risk factors for cardiovascular disease: The Problems and implications of overlapping affective dispositions. Psychological Bulletin, 131, 260-300. Takahashi,Y., Kipnis, D. M., & Daughaday, W. H. (1968). Growth hormone secretion during sleep. The Journal of Clinical Investigation, 47, 2079–2090. Taylor, S. E. (1991). Asymmetrical effects of positive and negative events: The mobilizationminimization hypothesis. Psychological Bulletin, 110, 67-85. Terracciano, A., McCrae, R. R., Hagemann, D., & Costa, P. T. (2003). Individual difference variables, affective differentiation, and the structures of affect. (2003). Journal of Personality, 71, 669-703. Toker, S., Shirom, A., & Melamed, S. (2008). Depression and metabolic syndrome: Genderdependent associations. Depression and Anxiety, 25, 661-669. Tomarken, A. J., Davidson, R. J., Wheeler, R. E., & Doss, R. C. (1992). Individual differences in anterior brain asymmetry and fundamental dimensions of emotion. Journal of Personality and Social Psychology, 62, 676-687. Tugade, M. M., & Fredrickson, B. L. (2004). Resilient individuals use positive emotions to bounce back from negative emotional experiences. Journal of Personality and Social Psychology, 86, 320-333. Turner, R. A., Altemus, M., Yip, D. N., Kupferman, E., Fltetcher, D., Bostrom, A., et al. (2002).  116 Effects of emotion on oxytocin, prolactin, and ACTH in women. Stress, 5, 269–276. Valdimarsdottir, H. B., & Bovbjerg, D. H. (1997). Positive and negative mood: Association with natural killer cell activity. Psychology and Health, 12, 319-327. van den Akker, M., Schuurman, A., Metsemakers, J., & Buntinx, F. (2004). Is depression related to subsequent diabetes mellitus? Acta Psychiatrica Scandinavica. 110, 178-183. van Eck, M., Berkhof, H., Nicolson, N., & Sulon, J. (1996). The effects of perceived stress, traits, mood states, and stressful daily events on salivary cortisol. Psychosomatic Medicine, 58, 447–458. van Hemert, D. A., Poortinga, Y. H., van de Vijver, F. J. R. (2007). Emotion and culture: A meta-analysis. Cognition and Emotion, 21, 913-943. Vedhara, K., Schifitto, G., & McDermott, M. (1999). Disease progression in HIV-positive women with moderate to severe immunosuppression: The role of depression. Dana Consortium on Therapy for HIV Dementia and Related Cognitive Disorders. Behavioral Medicine, 25, 43-47. Warner, R. M. (1998). Spectral analysis of time-series data. New York, NY: Guilford Press. Watson, D. (2000). Mood and temperament. New York, NY: Guilford Press. Watson, D., Clark, L. A., & Tellegen, A. (1984). Cross-cultural convergence in the structure of mood: A Japanese replication and a comparison with U.S. findings. Journal of Personality and Social Psychology, 47, 127-144. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063-1070.  117 Watson, D., & Tellegen, A. (1985). Toward a consensual structure of mood. Psychological Bulletin, 98, 219-235. Watson, D., Wiese, D., Vaidya, J., & Tellegen, A. (1999). The two general activation systems of affect: Structural findings, evolutionary considerations, and psychobiological evidence. Journal of Personality and Social Psychology, 76, 820-838. Weiss, R., Dziura, J., Burgert, T. S., Tamborlane, W. V., Taksali, S. E., Yeckel, C. W., et al. (2004). Obesity and the metabolic syndrome in children and adolescents. The New England Journal of Medicine, 350, 2362-2374. Wessman, A. E., & Ricks, D. F. (1966). Mood and Personality. New York: Holt, Rinehart, & Winston. Williams, J. E., Nieto, F. J., Sanford, C. P., Couper, D. J., & Tyroler, H. A. (2002). The association between trait anger and incident stroke risk: The Atherosclerosis Risk in Communities (ARIC) Study. Stroke, 33, 13-9. Wilson, T. D., Centerbar, D. B., Kermer, D. A., & Gilbert, D.T. (2005). The pleasures of uncertainty: Prolonging positive moods in ways people do not anticipate. Journal of Personality and Social Psychology, 88, 5-21. Wilson, P. W. F, D'Agostino, R. B., Levy, D., Belanger, A. M., Silbershatz, H., Kannel, W. B. (1998). Prediction of coronary heart disease using risk factor categories. Circulation, 97, 1837-1847. Yehuda, R., Teicher, M. H., Trestman, R. L., Levengood, R. A. & Siever, L. J. (1996). Cortisol regulation in posttraumatic stress disorder and major depression: A chronobiological analysis. Biological Psychiatry, 40, 79-88.  118 Yik, M. S. M., Russell, J. A., & Feldman-Barrett, L. (1999). Structure of self-reported current affect: Integration and beyond. Journal of Personality and Social Psychology, 77, 600619. Young, A. H. (2004). Cortisol in mood disorders. The International Journal on the Biology of Stress, 7, 205-208. Zuckerman, D. M., Kasl, S. V., & Ostfeld, A. M. (1984). Psychosocial predictors of mortality among the elderly poor. The role of religion, well-being, and social contacts. American Journal of Epidemiology, 119, 410-423.  119  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0069941/manifest

Comment

Related Items