Positive Affect and Health

Negative affective styles such as anxiety, depression, and hostility have long been accepted as predictors of increased risk for illness and mortality. In contrast, positive affective styles have been relatively ignored in the health literature. Here we highlight consistent patterns of research associating trait positive affect (PA) and physical health. The evidence we review suggests an association of trait PA and lower morbidity and decreased symptoms and pain. PA is also associated with increased longevity among community-dwelling elderly. The association of PA and survival among those with serious illness is less clear and suggests the possibility that PA may be harmful in some situations. We conclude by raising conceptual and methodological reservations about this literature and suggesting directions for future research. KEYWORDS—positive emotion; positive affect; morbidity; mortality; health; symptoms The role of emotions in physical health has been a central topic in health psychology for some time. Emotions are thought to represent the principal pathway linking psychological stress to disease, and enduring affective styles such as anxiety and depression have been found to be associated with greater morbidity and mortality. However, when health psychologists have referred to the roles of emotions and affect in health, they have typically meant negative emotions such as anger, depression, and anxiety. Only recently has there been any serious discussion of the potential effect of positive affect (PA). One challenge in making sense of the literature on PA and health is that there is little agreement on what is meant by PA. We define positive emotion or affect as feelings that reflect a level of pleasurable engagement with the environment, such as happiness, joy, excitement, enthusiasm, and contentment (Clark, Watson, & Leeka, 1989). These can be brief, longer lasting, or more stable trait-like feelings. Importantly, the lack of positive engagement does not necessarily imply negative affect such as anger, anxiety, and depression. REVIEW The strongest links between positive emotions and health are found in studies that examine trait affective style, which reflects a person’s typical emotional experience, rather than state affect, which reflects momentary responses to events. Here we provide short descriptions of the associations between trait PA and mortality (longevity), morbidity (illness onset), survival from life-threatening disease, and reports of symptoms and pain. (For a comprehensive review of this literature see Pressman & Cohen, 2005). The studies we review use prospective designs that help to eliminate the explanation that being sick resulted in lower PA. This is done by measuring PA and health at study onset (baseline) and assessing whether PA predicts changes in health over the follow-up period. Because the measure of PA is given before the change in health, it cannot have been caused by that change. Many, but not all, of the studies also include controls for spurious (third) factors such as age, sex, socioeconomic status, and race/ ethnicity. Overall, the literature reviewed here is provocative, although it suffers from a range of methodological and conceptual limitations. It does however allow us to highlight both consistencies in results as well as the issues that need to be addressed to ultimately determine if a positive affective style is an important predictor of good health. Mortality A study that has received considerable attention evaluated PA by coding autobiographical writing samples collected from a group of nuns when they were in their early twenties (Danner, Snowdon, & Friesen, 2001). The greater the number of positive emotion words and sentences, the greater was the probability (adjusting for age and education) of being alive 60 years later. In contrast, the number of negative emotions reported was not associated with mortality. However, the overall evidence on PA and mortality is more complex. Most (seven) of these studies have been done in elderly persons (average age over 60) living either on their own or with their families. These studies are virtually unanimous in linking positive emotional dispositions to longer life. But positive emotions are not generally associated with increased longevity in studies of other populations. For example, two studies suggest that institutionalized elderly with high PA are at increased risk Address correspondence to Sheldon Cohen, Department of Psychology, Carnegie Mellon University, Pittsburgh, PA 15213; e-mail: [email protected]. CURRENT DIRECTIONS IN PSYCHOLOGICAL SCIENCE 122 Volume 15—Number 3 Copyright r 2006 Association for Psychological Science of mortality (Janoff-Bulman & Marshall, 1982; Stones, Dornan, & Kozma, 1989) and an analysis of a sample of gifted children found that PA during childhood was associated with greater risk for death 65 years later (Friedman et al., 1993). Illness Onset In a study from our own laboratory (Cohen, Doyle, Turner, Alper, & Skoner, 2003), 334 adult volunteers were phone interviewed seven times over a 3-week period. For each interview, participants rated how accurately each of nine positive and nine negative adjectives described how they felt over the last day. Examples of PA items included lively, energetic, happy, cheerful, at ease, and calm. Examples of negative-affect (NA) items included sad, depressed, nervous, and hostile. Daily mood scores were calculated and averaged across the 7 days to create summary measures of trait PA and NA. Subsequently, subjects were exposed to one of two viruses that cause a common cold and were monitored for 5 days for the development of clinical illness. Colds were defined by objective markers of illness, including infection, mucus production (assessed by weighing tissues), and congestion (assessed by the amount of time it took for a dye put into the nostrils to reach the back of the throat). Those with high levels of PA were less likely to develop a cold when exposed to a virus (see Fig. 1). This relationship remained after controlling for age, sex, immunity (baseline antibody to the experimental virus), education, and NA. In other morbidity studies, trait PA has been associated with lower rates of stroke among noninstitutionalized elderly (Ostir, Markides, Peek, & Goodwin, 2001), lower rates of rehospitalization for coronary problems (Middleton & Byrd, 1996), fewer injuries (e.g., Koivumaa-Honkanen et al., 2000) and improved pregnancy outcomes among women undergoing assisted fertilization (Klonoff-Cohen, Chu, Natarajan, & Sieber, 2001). These studies are often limited by a lack of control for factors such as NA, optimism, and personal control that may influence both PA and disease susceptibility, and many do not rule out the possibility that PA itself (e.g., endorsing of items such as energetic, full-of-pep, and vigorous) is merely a marker of subclinical disease processes. Survival A popular hypothesis is that trait PA increases longevity of persons suffering from life-threatening disease. However, comparatively few studies have examined whether PA predicts survival among people with chronic diseases, and available findings are at best mixed. A pattern of results does however suggest a hypothesis. Individuals with diseases that have decent prospects for long-term survival, such as early-stage breast cancer, coronary heart disease, and AIDS, may benefit from PA. However, high levels of trait PA may be detrimental to the health of individuals who have advanced diseases with poor and shortterm prognoses—e.g., patients with melanoma, metastatic breast cancer, and end-stage renal disease—possibly as a consequence of underreporting of symptoms resulting in inadequate care, or of a lack of adherence to treatment (Pressman & Cohen, 2005). Symptoms and Pain There is considerable evidence linking PA to reports of fewer symptoms, less pain, and better health. These outcomes have practical importance, but there is reason to think that this association may be driven primarily by PA influences on how people perceive their bodies rather than by affect-elicited changes in physiological processes (e.g., Pennebaker, 1983). For example, a study from our own lab suggests that trait PA is associated with less symptom reporting when objective disease is held constant (Cohen et al., 2003). As described earlier, PA and NA were assessed by averaging responses across seven nightly interviews. Volunteers were then exposed to a virus that causes the common cold and monitored for objective signs of illness. To test whether trait affect could influence symptom reporting, we predicted self-reported cold symptoms (collected for 5 days following viral exposure) from trait affect, controlling for the objective markers of disease mentioned earlier. When objective signs of illness were held constant, those higher in trait PA reported less severe symptoms, and those higher in trait NA reported more severe ones. Figure 2 presents the residual scores derived from the PA analysis. These scores represent the extent to which one reports more (1 scores) or fewer ( scores) symptoms than would be predicted from the objective markers of disease. Interestingly, when both PA and NA were entered in the same regression equation, only PA continued to predict symptom reporting, suggesting that low PA (not high NA) may be the driving force in the reporting of unfound symptoms. Low Middle High 0 5 10 15 20 25 30 35