ARTICLE IN PRESS

Psychoneuroendocrinology (2007) 32, 5664

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/psyneuen

Neuroendocrine and cardiovascular correlates of positive affect measured by ecological momentary assessment and by questionnaire
Andrew Steptoea,, E. Leigh Gibsonb, Mark Hamera, Jane Wardleb
Psychobiology Group, Department of Epidemiology and Public Health, University College London, 1-19 Torrington Place, London WC1E 6BT, UK b Health Behaviour Unit, Department of Epidemiology and Public Health, University College London, 1-19 Torrington Place, London WC1E 6BT, UK Received 31 July 2006; received in revised form 1 September 2006; accepted 4 October 2006
a

KEYWORDS
Positive affect; Happiness; Cortisol; Blood pressure; Stress

Summary The relationships between positive affect, salivary cortisol over the day, and cardiovascular responses to laboratory mental stress tests, were assessed in 72 healthy non-smoking men (mean age 33.678.8 years). Positive affect was measured by aggregating ecological momentary assessments (EMA) of happiness obtained at four times on each of 2 working days, and by questionnaire using the Positive and Negative Affect Schedule (PANAS). Saliva was sampled on 2 days, on waking, 30 and 60 min later, and four other times over the day. Blood pressure and heart rate responses to speech and mirror tracing tasks were measured over two sessions 4 weeks apart. Data were analysed using regression of positive affect on biology adjusting for age, body mass and negative affect, with additional adjustment for time of waking in cortisol analyses and for work stress in cardiovascular analyses. EMA positive affect was inversely associated with cortisol early in the day and with the cortisol increase after waking, controlling for age, body mass index, and negative affect (P 0:012). There was no relationship between PANAS positive affect and cortisol, or between EMA positive affect and cortisol later in the day. Diastolic pressure recovery poststress was more rapid among participants with high positive affect (P 0:022) and with lower systolic pressure throughout the stress sessions, after controlling for covariates including negative affect. PANAS positive affect was also inversely associated with systolic pressure, but not with diastolic stress or heart rate. We conclude that positive affect is related to biological responses in the laboratory and everyday life that may be health protective. Effects were substantially stronger when positive affect was assessed by aggregating EMA samples than with questionnaire measures. & 2006 Elsevier Ltd. All rights reserved.

Corresponding author. Tel.: +44 207 679 1804; fax: +44 207 916 8542.

E-mail address: a.steptoe@ucl.ac.uk (A. Steptoe). 0306-4530/$ - see front matter & 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.psyneuen.2006.10.001

ARTICLE IN PRESS
Biological correlates of positive affect 57 and Fredrickson (2004) found that undergraduates reporting higher levels of positive affect showed faster post-stress cardiovascular recovery following challenging tasks. In contrast, Maier et al. (2003) reported that positive affect was associated with heightened systolic and diastolic BP reactivity to mental arithmetic. One reason for the inconsistency in the results may be lack of representativeness of results from single laboratory stress sessions. It has been argued that psychophysiological stress proles should be aggregated across two or more sessions to identify robust individual differences (Kamarck et al., 1994, 2000). In the same vein, salivary cortisol measures over a single day may not accurately reect typical cortisol proles for the individual (Kirschbaum et al., 1990; Smyth et al., 1998). In the present study, psychophysiological stress testing and saliva sampling over the day were repeated at a 4-week interval, and we aggregated data across the two time points for EMA measures of positive affect, salivary cortisol and cardiovascular stress responsivity. We tested the hypothesis that positive affect would be associated with reduced BP and heart rate stress reactivity and accelerated post-stress recovery. The third issue addressed in this study was the method of assessing positive affect. The most common technique is to use positive affect scales from instruments such as the Positive and Negative Affect Schedule (PANAS, Watson et al., 1988), or the positively worded items from standard depression scales (e.g. Ostir et al., 2001). The respondent is typically average mood over a specied time period (such as 1 week). These methods have been criticised for being inuenced by recall bias, memory distortion, the dominant inuence of current mood state, and focusing illusions (Stone and Shiffman, 2002; Kahneman et al., 2006). The alternative is to assess affect with EMA techniques, in which people are prompted to rate how they feel at the moment on several occasions during a day or across days. Aggregation of these measures is thought to provide more valid estimates of typical affective states (Schwarz, 1999). However, EMA approaches involve a heavy participant burden, and of course the day or days over which ratings are obtained may not be typical for that individual. In the present study, we assessed positive affect both by aggregating EMA measures and by administering the PANAS at the time of each laboratory session. The third aim of this study was therefore to discover whether the biological correlates of EMA-derived and questionnaire-based measures of positive affect are the same.

1. Introduction
There is growing evidence that positive affect is associated with a longer healthy life expectancy and reduced risk of physical disease (Huppert et al., 2004; Pressman and Cohen, 2005). Studies with older adults have shown associations between positive affect and reduced risk of mortality, onset of disability, and stroke, independently of risk factors and negative affect (Ostir et al., 2000, 2001; Blazer and Hybels, 2004). Positive affect also predicted reduced risk of upper respiratory infectious illness following experimentally administered virus (Cohen et al., 2003), and greater antibody responses to hepatitis B vaccination (Marsland et al., 2006). These ndings have stimulated interest in the biological factors that might correlate with positive affect and contribute to health-protective effects. We carried out a study involving 216 men and women from the Whitehall II epidemiological cohort, assessing positive affect by aggregating ecological momentary assessments (EMA) of happiness throughout a working day and evening (Steptoe et al., 2005a). Positive affect was associated with lower salivary cortisol averaged from eight samples over the working day, with a similar pattern for the weekend day. These effects were independent of age, gender, socioeconomic status, smoking, and body mass index (BMI), and were replicated at 3-year follow up (Steptoe and Wardle, 2005a). Positive affect was also associated with lower ambulatory heart rate in men, and reduced brinogen responses to standardised mental stress tests. Crucially, the effects were independent of negative affect. Negative affective states have been associated with elevated cortisol (van Eck et al., 1996; Polk et al., 2005) and psychophysiological stress reactivity (Kibler and Ma, 2004), but these ndings indicated that associations between biological responses and positive affect were not merely due to low levels of negative affect. Associations between lower cortisol and positive affect have also been reported in a community sample in the USA (Smyth et al., 1998), and in a sample of Hong Kong Chinese (Lai et al., 2005). Polk et al. (2005) reported a more complex pattern in a large sample of healthy adults in which positive affect was either estimated as a trait characteristics by averaging seven sets of affect ratings obtained in the evenings over a 6-week period, or as a state characteristics from affect ratings taken on the day of cortisol assessment. Trait positive affect was associated with lower cortisol on waking in women and not men, but with lower cortisol later in the day in men. However, these data were collected while participants were living in a hotel before carrying out a virus challenge study, so may not reect cortisol output in normal life. In view of these varied results, the rst aim of this study was to discover whether we could replicate our nding associating EMA-derived positive affect with cortisol independently of negative affect in a young male adult sample. We paid particular attention to the cortisol awakening response (CAR) as a marker of neuroendocrine dysregulation (Clow et al., 2004). It is possible that in addition to biological correlates in everyday life, positive affect is related to attenuated physiological stress reactivity (Pressman and Cohen, 2005). In our previous study, we did not show any association between positive affect and blood pressure (BP) or heart rate reactions to behavioural stress tasks. However, Tugade

2. Methods
2.1. Participants
Participants were 73 non-smoking men who were all in fulltime employment. They were aged 33.678.8 years, were predominantly white European (81.7%) and well-educated, with 73.4% having a university degree. Participants reported being healthy, and were recruited for a study of the effect of tea consumption on health. All the data presented here were collected before any experimental manipulation of tea consumption. Results for the stability of laboratory stress

ARTICLE IN PRESS
58 reactivity over the two sessions have been described elsewhere (Hamer et al., 2006a). The study was approved by the Joint University College London/University College London Hospitals Research Ethics Committee, and all participants gave written consent. A. Steptoe et al. assessed by monitoring BP and heart rate for 510 min following tasks.

2.5. Psychological measures

Positive and negative affect at the time of the laboratory session was assessed using the PANAS (Watson et al., 1988). This consists of 20 adjectives, each of which is rated on a 5-point scale from 1 (very slightly or not at all) to 5 (extremely). Ten items contribute to the positive affect scale (e.g. enthusiastic, excited, active) and 10 to the negative affect scale (e.g. upset, afraid). Participants were asked to think about how they had been feeling over the past week when completing the scales. Ratings were averaged to generate positive and negative affect scores. The Cronbach as for the PANAS were 0.85 and 0.73 for session 1 and 0.88 and 0.78 for session two. Work stress was assessed using the effortreward imbalance scale (Kuper et al., 2002). Effort was assessed with ve items and reward with seven items, each of which was rated on a 4-point scale. Effortreward imbalance was calculated as effort divided by reward, where a score of 1 represents a perfect balance, and higher scores reect greater imbalance and therefore more work stress. Psychological distress was measured with the General Health Questionnaire (GHQ) 28-item version. This is a well established measure devised for screening for psychological distress in population studies (Goldberg, 1972). The total score (range 028) was used in this study, with higher values indicating greater distress.

2.2. Study design

Participants attended two laboratory psychophysiological stress sessions 4 weeks apart. After each session, they collected saliva samples over a working day for the analysis of cortisol and made happiness ratings at the same time. Analyses of the associations between positive affect and biological responses were adjusted for negative affect, and for factors known to be associated with the biological responses assessed. For salivary cortisol over the day, these included age, BMI, and time of waking in the morning. Laboratory cardiovascular analyses were adjusted for age, BMI, and for work stress assessed with the effortreward imbalance model, since we have previously shown that this inuences BP responses (Hamer et al., 2006b).

2.3. Saliva collection and cortisol analysis

Participants were instructed to collect saliva using salivettes (Sarstedt, Leicester, UK) on seven occasions on each of 2 working days: immediately on waking, 30 and 60 min after waking, and then as near as possible to 10:30, 12:30, 17:30 and 21:30 h. At the time of saliva sample, they completed a paper diary indicating the timing of saliva collection, their location and activities at the time, and also provided a rating of how happy they were on a 5-point scale from 1not at all to 5very much. Participants were instructed not to eat, drink or brush their teeth until after the 60 min post-waking sample, and not to eat or drink within 15 min of later samples, and explicitly to take the 12:30 h and 17:30 h samples before meals. Saliva samples were frozen after they were returned to the laboratory, and were later analysed using a time-resolved immunoassay at the Technical University, Dresden. The intra- and inter-assay coefcients of variation were o8%.

2.6. Data analysis

2.6.1. EMA positive affect assessment The number of happiness ratings of 4 or 5 on the 5-point scale obtained over the four daytime sampling points (10:30 h to 21:30 h) was expressed as a percentage of ratings obtained, as in our previous study (Steptoe et al., 2005a). These proportions were averaged over the 2 days of assessment to generate a mean that could range from 0% (never happy) to 100% (happy at all time points). Data were primarily analysed with this aggregate positive affect measure as a continuous variable using multiple linear regression. However, in order to illustrate the ndings, we also divided the sample into high and low positive affect groups, depending on whether they were in a positive state on fewer than half or more than half of samples. 2.6.2. Cortisol assessment Salivary cortisol results were screened and values over 70 nmol/l were excluded, since in previous studies we have found these high levels to indicate haemolysis. Cortisol samples at each time point were averaged across 2 days to produce aggregate values. The CAR was analysed by computing the area under the curve (CARauc) as described by Pruessner et al. (2003a), using the formula: ((waking cortisol+30 min cortisol)/2) 30+((30 min cortisol+60 min cortisol)/2) 30. Second, the increase in cortisol following waking (CARi) was calculated by subtracting the waking value extrapolated over 60 min from the CARauc as follows: CARauc(waking cortisol 60). Cortisol over the day

2.4. Psychophysiological testing procedure

Participants attended two identical psychophysiological testing sessions at the same time in the morning on each occasion. BP and heart rate were measured using a Finometer (TNO Biomedical Instrumentation, Amsterdam, NL), which provides beat-by-beat information from the nger using the vascular unloading technique. At the beginning of each session, weight and height were recorded for the calculation of BMI, and a blood sample was drawn that will not be discussed here. A 10 min resting period followed with BP and heart rate averaged over the last 5 min to represent baseline levels. Participants then carried out two behavioural tasks, a 3 min simulated public speaking task, and a 5-min mirror tracing task, as described previously (Hamer et al., 2006b). Cardiovascular monitoring continued throughout the tasks, and a rating of stress was obtained after each task on a 7-point scale (1not at all stressed to 7very stressed). Post-task recovery was

ARTICLE IN PRESS
Biological correlates of positive affect (CORTday) was analysed by averaging the four values collected from 10:30 to 21:30 h. 59

3.1. Positive affect and cortisol over the day

The timing of salivary samples averaged 06:56 h753 min for waking, 07:27755 min and 08:01755 min for the 30 and 60 min post-waking samples, and 10:30 h721 min, 12:40 h7 41 min, 15:44 h732 min, and 21:45 h731 min for the four samples over the day. Salivary cortisol averaged 16.47 5.8 nmol/l on waking, and increased to 22.577.1 nmol/l 30 min later. The CARauc averaged 1190.27335 nmol/l and the CARi averaged 204.67327 nmol/l. The CARauc on the 2 sampling days were positively correlated (r 0:31, P 0:008), as were the times of waking on the 2 days (r 0:65, Po0:001). The average CORTday was 5.587 2.6 nmol/l. Cortisol output over the rst hour after waking was inversely associated with EMA positive affect. Multiple regression on CARauc revealed a negative association (b 0:32, P 0:012) after controlling for age, BMI, time of waking and negative affect. The cortisol level on waking itself was not related to positive affect (b 0:19, P 0:12), but the CARi was (b 0:23, P 0:045). This indicates that positive affect was related to the rise in cortisol after waking rather than waking values. There was no association between EMA positive affect and CORTday (P 0:47). These ndings are illustrated in Fig. 1, where the cortisol proles of high and low EMA positive affect groups are summarised. Repeated measures analysis of variance conrmed the signicant group by time interaction (F 5:06, Po0:001). The group difference in the CARauc was signicant (F 7:55, P 0:001), as was the difference in CARi (F 5:25, P 0:025) after controlling for age, BMI, negative affect, and time of waking. When analyses were based on the PANAS positive affect measure, a different pattern of results emerged. There was a near signicant association with cortisol level on waking, with lower cortisol in people reporting more positive affect (b 0:23, P 0:073). But PANAS positive affect was unrelated to CARauc (P 0:55), CARi (P 0:22) or CORTday (P 0:32). Thus the association between positive affect and cortisol identied using aggregate EMA samples was not replicated with the summary questionnaire measure of positive affect. In addition to these analyses, we also tested associations between positive affect and the slope of the diurnal decline

2.7. Statistical analysis

The characteristics of high and low EMA positive affect groups were compared using analysis of variance for continuous variables and w2 tests for categorical variables. The relationship between positive affect (assessed either with EMA or PANAS measures) and cortisol was assessed using linear regression on waking cortisol, CARauc, CARi, and CORTday. Negative affect from the PANAS, age, BMI, and time of waking in the morning were included in the models as covariates. Results were expressed as standardised (b) regression coefcients with signicance levels. Additionally, repeated measures analysis of variance was conducted with positive affect group (high, low) as the between-subject factor, and cortisol sample as the within-subject factor. The GreenhouseGeisser correction for degrees of freedom was applied where appropriate. BP and heart rate data from the two laboratory sessions were averaged. The pattern of cardiovascular and subjective stress response was analysed using repeated measures analysis of variance, with trial (baseline, speech task, mirror tracing task, recovery) as the withinsubject factor. Associations with positive affect were analysed with linear regression on the different trials, with negative affect, age, BMI, and job stress included as covariates.

3. Results
The characteristics of individuals in the high and low EMA positive affect groups are summarised in Table 1. The higher positive affect groups were signicantly younger (F 4:95, P 0:029), but there were no differences in educational attainment, marital status, ethnicity, job stress, or in negative affect or psychological distress assessed with the GHQ. PANAS positive affect was associated with higher EMA positive affect as expected (F 16:5, Po0:001), and the correlation between the two positive affect measures was 0.46 (Po0:001).

Table 1

Comparison of high and low positive affect groups. High positive affect (EMA) N 30 Low positive affect (EMA) N 43 35.479.2 73.8% 61.9% 85.7% 37.5% 1.1170.32 2.4872.9 7.8473.7 20.575.6 0703753 P

Age (years) College/university education Married White European Former smokers Effortreward imbalance GHQ 28 Negative affect (PANAS) Positive affect (PANAS) Time of waking (h)

30.877.5 72.4% 58.6% 75.9% 34.8% 1.1070.33 2.5072.9 7.3374.9 25.574.3 0646750

0.029 0.89 0.78 0.29 0.84 0.87 0.97 0.62 0.001 0.17

ARTICLE IN PRESS
60
28 24 20 nmol/l 16 12 8 4 0 Wake Wake+30 Wake+60 1030 1230 1730 2130

A. Steptoe et al.

Figure. 1 Mean cortisol sampled on waking, 30 and 60 min later, and at four later times of the day in participants reporting high (solid line) and low (dashed line) EMA positive affect. Values are adjusted for age, BMI, time of waking in the morning, and negative affect. Error bars are standard error of the mean (s.e.m.).

in cortisol over the day, but effects were not signicant. The statistical signicance of associations between positive affect and cortisol was similar when negative affect was omitted as a covariate in the regression models.
mm Hg

170 160 150

3.2. Positive affect and stress reactivity

The tasks elicited substantial increases in systolic and diastolic BP, averaging 39.3 and 21.5 mmHg for the speech and 33.4 and 19.9 mmHg for the mirror tracing tasks (Fig. 2). Post-stress recovery was incomplete 510 min following tasks. Linear regression on systolic BP, adjusting for age, BMI, effortreward imbalance and negative affect, showed signicant negative associations with EMA positive affect during the baseline (b 0:26, P 0:029), speech (b 0:29, P 0:018), mirror tracing (b 0:24, P 0:044), and recovery trials (b 0:29, P 0:015). In all cases, participants with higher positive affect had lower systolic BP. There were, however, no differences in systolic BP stress reactivity or recovery expressed as changes from baseline. These effects indicate that absolute levels of systolic BP recorded during tasks were greater in people experiencing lower positive affect, but their stress reactivity was not elevated. Diastolic BP showed signicant negative associations with EMA positive affect in regression analyses on the baseline (b 0:25, P 0:020) and recovery trials (b 0:35, P 0:001), but not the speech (P 0:069) or mirror tracing tasks (P 0:13). Additionally, delayed recovery (calculated as change between the recovery and baseline trials) was associated with positive affect (b 0:23, P 0:022, adjusted for covariates). Thus, not only were absolute levels of diastolic BP lower in individuals with greater positive affect, they also showed more prompt post-stress recovery. There were no associations between EMA positive affect and heart rate responses. Associations between cardiovascular responses and PANAS positive affect were also assessed. Regression analyses showed signicant associations between PANAS positive

140 130 120 110 Base 100 95 90 Speech MT Recovery

mm Hg

85 80 75 70 Base Speech MT Recovery

Figure. 2 Mean systolic BP (upper panel) and diastolic BP (lower panel) during the baseline, speech, mirror tracing (MT) and recovery trials in participants reporting high (solid line) and low (dashed line) EMA positive affect. Values are adjusted for age, BMI, effortreward imbalance and negative affect. Error bars are s.e.m.

affect and systolic BP during all trials (b 0:32 to 0.43, all Po0:01) after adjusting for age, BMI, effortreward imbalance, and negative affect. There were no associations with stress reactivity or recovery change scores. Neither diastolic BP nor heart rate was signicantly associated with PANAS positive affect at any point. The associations between positive affect measured either with

ARTICLE IN PRESS
Biological correlates of positive affect aggregated EMA ratings or the PANAS and subjective stress responses were not signicant. The signicance of associations between positive affect and BP was similar when negative affect was omitted as a covariate in the regression models. 61 were no differences by 10:30 h onwards. These inconsistent ndings may be related to variations in the methods of measurement of positive affect or in the timing and circumstances of cortisol sampling. Previously, we aggregated eight EMA samples of happiness over a single day to measure positive affect (Steptoe et al., 2005a), while in this study we obtained four samples on each of 2 days. Other studies, unlike ours, may have included sampling within 1 h of meals, which could produce associations with affect later in the day, since meal-induced cortisol secretion has been found to be associated with psychological distress (Gibson et al., 1999). Additionally, our own previous work involved men and women aged 4759 years, whereas this study was carried out with younger men aged 33 on average. It could be that dysfunction in cortisol regulation among people with low positive affect is limited to the particularly sensitive early morning period in young adults, but generalises to other times of day as people age. Our study did not include any checks on the reliability of timing of cortisol samples over the day. Inaccurate timing can be especially problematic in assessing the CAR, since delays between waking and taking the waking sample are associated with a reduced or absent CAR (Wright and Steptoe, 2005). Three methods of increasing the reliability of morning cortisol sampling have been advocated: the use of electronically time-tagged cotton roll containers (Kudielka et al., 2003), excluding individuals who report a delay between waking and taking their rst cortisol sample (Kunz-Ebrecht et al., 2004), and excluding individuals who show variability in the CAR across days (Clow et al., 2004). It is theoretically possible that happier individuals in this study delayed their waking cortisol sample compared with less happy participants, so that their CARs were reduced. However, this is unlikely for several reasons. First, the cortisol level on waking was not related to positive affect; when the waking sample is delayed, it is typically higher than in non-delayed groups, since the post-waking rise in cortisol has already begun. Second, there was no difference in reported time of waking between people high and low in positive affect. Third, the consistency of the CAR over days did not vary with positive affect. A related issue is the accuracy of timing of the EMA affects measures. We used pencil-and-paper measures for recording positive affect over the day, and no prompts were in place to ensure accurate timing. The use of diary measures has been investigated in detail over recent years, with variation between studies of how reliably people complete entries when required (Stone et al., 2003; Green et al, 2006; Broderick and Stone, 2006). The mean reported times of completion of affect samples over the day were within 15 min of those requested, but this does not guarantee that they were necessarily measured at these times. However, our analysis involved aggregating samples rather than analysing variation over the day, so violations of the timing are less critical than in analyses of withinsubjective covariation between affect and biology. The results of psychophysiological stress testing are consistent with Tugade and Fredricksons (2004) hypothesis that positive affect is associated with more rapid post-stress recovery of cardiovascular activity. We found that EMAderived positive affect was related to more prompt diastolic BP recovery independently of negative affect, age, BMI and

4. Discussion
The main ndings of this study were that low positive affect assessed with EMA-derived ratings of happiness were associated with higher cortisol early in the day and a heightened CAR, independently of negative affect, age, BMI and time of waking. EMA positive affect was also related to lower systolic BP at rest, during and after stress tasks, and to more rapid diastolic BP recovery following stress. Positive affect assessed more conventionally with the PANAS was not related to the CAR, to cortisol over the day, or to diastolic BP recovery. It was, however, associated with lower systolic BP at rest, during and after stress tasks. These ndings were obtained using data averaged over two cortisol sampling days and two psychophysiological testing sessions 1 month apart. The results were unchanged when EMA ratings of stress were substituted for the negative affectivity scale of the PANAS as the measure of negative affect (data not shown). The ndings broadly support the hypothesis that positive affect has protective biological correlates that may reduce risk of future physical morbidity. These associations do not merely reect the well-established relationship between negative affect and biology, but represent distinctive correlates of positive affect. There is not complete consensus about the signicance of elevated cortisol early in the day and a heightened CAR. A greater CAR has previously been observed on working rather than weekend days (Kunz-Ebrecht et al., 2004), and has been correlated with chronic stress, nancial strain, and work stress (Schulz et al., 1998; Wust et al., 2000; Steptoe et al., 2004, 2005b). Its association with depression has been inconsistent, but most studies have found that depressed individuals have larger CARs (Bhagwagar et al., 2003, 2005; Pruessner et al., 2003b; Stetler and Miller, 2005). High cortisol early in the day has also been found to predict future depressive illness in high risk adolescents and women (Goodyer et al., 2000; Harris et al., 2000). By contrast, a blunted CAR has been observed in some but not all studies of posttraumatic stress disorder (Luecken et al., 2004; Young et al., 2004; Yehuda et al., 2005; Wessa et al., 2006). Heightened cortisol plays an important role in cardiovascular pathology and in many other conditions (McEwen et al., 1997; Girod and Brotman, 2004). Our nding of a larger CAR in less happy individuals is therefore consistent with a favourable effect of positive affect on neuroendocrine pathways to illness. The results do not, however, replicate the observation of lower cortisol over the day and evening in people with higher positive affect that we and others have previously reported (Lai et al., 2005; Steptoe et al., 2005a). Polk et al. (2005) showed that state positive affect (measured on a single occasion on the day of cortisol sampling) was associated with differences in cortisol early in the day that converged as the day progressed, but in that study differences were still present at midday. In our study, there

ARTICLE IN PRESS
62 work stress. Tugade and Fredrickson used a multivariate index of cardiovascular response involving BP, heart rate, pulse amplitude and other measures, so separate associations with diastolic BP were not tested. Additionally, they measured very short-term recovery (within 60 s) rather than the more extended periods tested in this experiment. As well as the diastolic BP recovery effect, we found that systolic BP was lower in all phases of the sessions in happier individuals (Fig. 2). Steptoe and Wardle (2005a) reported that ambulatory systolic BP over the day and evening was inversely associated with positive affect, but our previous experimental investigation with middle-aged adults showed no association with BP stress reactivity or recovery. There may be several reasons for this difference in results. In our previous experiment, we tested recovery 4045 min following tasks, whereas in this study, recovery was measured after 510 min. Effects may be present at this earlier stage that do not persist over the longer period. The tasks used in this study were more emotionally provocative than those we used before. In particular, the socially evaluative speech task generated very substantial responses averaging 39.3/ 33.4 mmHg. Additionally, as noted earlier, the participants in this study were younger and healthier than those in our earlier work. The pathways through which these cardiovascular effects are mediated is not clear. There were no differences in subjective stress appraisal, so happier people did not show smaller BP effects because they felt less stressed. Differences may either exist at the level of subcortical regulation, or relate to more subtle aspects of coping that were not captured by subjective stress ratings. The elevated BP recorded during and after stress in participants reporting lower positive affect may put them at increased risk for the development of coronary heart disease and other cardiovascular disorders (Treiber et al., 2003). Prolonged BP responses or delayed post-stress recovery has been found to predict heightened future clinic BP (Steptoe and Marmot, 2005), increased abdominal adiposity in men (Steptoe and Wardle, 2005b), and progression of carotid atherosclerosis (Jennings et al., 2004). Delayed cardiovascular recovery is also associated with prolonged haemostatic stress responses and with heightened inammatory cytokine responses, suggesting that it may be a marker of more extensive psychobiological dysregulation (Steptoe and Marmot, 2006). These ndings were all generated with positive affect assessed using EMA techniques. In contrast, the associations between biological responses and PANAS positive affect were very limited. The two measures of positive affect were only moderately correlated (r 0:46), suggesting they may measure somewhat different aspects of experience. Kahneman et al. (2006) have argued that measures based on EMA or retrospective momentary assessments such as the Day Reconstruction Method may provide more reliable estimates of affect because they take account of experienced happiness, and how much time people spend in activities that vary in affective valence. Questionnaire measures such as the PANAS may be strongly inuenced by current mood and recent experience, even though the individual is asked to rate their feelings over the last week. Our ndings suggest that biological correlates of positive affect may be identied more reliably using the admittedly laborious EMA A. Steptoe et al. measures, compared with simpler questionnaire assessments of emotional style. This study was limited to healthy young non-smoking men, so results may not generalise to other groups. EMA-derived positive affect was assessed with a single happiness rating on each occasion, rather than with more subtle measures. Cardiovascular assessments were conned to BP and heart rate, and did not include more detailed measures of haemodynamics. Nevertheless, the results add to the growing body of evidence indicating that independent of negative affect, positive affect has health-protective biological correlates. The extent to which these processes contribute to differences in health outcomes in happy and less happy individuals will become more clear when prospective evaluations are completed.

Acknowledgements
This research was funded by the Biotechnology and Biological Sciences Research Council and Unilever Research, with additional support from the British Heart Foundation. Jane Wardle is supported by Cancer Research UK. Leigh Gibson is now at the University of Roehampton. We are grateful to Raisa Vuononvirta and Emily Williams for their assistance in data collection, and to Clemens Kirschbaum from the Technical University Dresden for carrying out cortisol assays.

References
Bhagwagar, Z., Hazi, S., Cowen, P.J., 2003. Increase in concentration of waking salivary cortisol in recovered patients with depression. Am. J. Psychiatry 160, 18901891. Bhagwagar, Z., Hazi, S., Cowen, P.J., 2005. Increased salivary cortisol after waking in depression. Psychopharmacology (Berl) 182, 5457. Blazer, D.G., Hybels, C.F., 2004. What symptoms of depression predict mortality in community-dwelling elders? J. Am. Geriatr. Soc. 52, 20522056. Broderick, J.E., Stone, A.A., 2006. Paper and electronic diaries: too early for conclusions on compliance rates and their effects comment on Green, Rafaeli, Bolger, Shrout, and Reis (2006). Psychol. Methods 11, 106111. Clow, A., Thorn, L., Evans, P., Hucklebridge, F., 2004. The awakening cortisol response: methodological issues and signicance. Stress 7, 2937. Cohen, S., Doyle, W.J., Turner, R.B., Alper, C.M., Skoner, D.P., 2003. Emotional style and susceptibility to the common cold. Psychosom. Med. 65, 652657. Gibson, E.L., Checkley, S., Papadopoulos, A., Poon, L., Daley, S., Wardle, J., 1999. Increased salivary cortisol reliably induced by a protein-rich midday meal. Psychosom. Med. 61, 214224. Girod, J.P., Brotman, D.J., 2004. Does altered glucocorticoid homeostasis increase cardiovascular risk? Cardiovasc. Res. 64, 217226. Goldberg, D., 1972. Manual of the General Health Questionnaire. NFER-Nelson, Windsor. Goodyer, I.M., Herbert, J., Tamplin, A., Altham, P.M., 2000. Recent life events, cortisol, dehydroepiandrosterone and the onset of major depression in high-risk adolescents. Br. J. Psychiatry 177, 499504. Green, A.S., Rafaeli, E., Bolger, N., Shrout, P.E., Reis, H.T., 2006. Paper or plastic? Data equivalence in paper and electronic diaries. Psychol. Methods 11, 87105.

ARTICLE IN PRESS
Biological correlates of positive affect
Hamer, M., Gibson, E.L., Vuononvirta, R., Williams, E., Steptoe, A., 2006a. Inammatory and hemostatic responses to repeated mental stress: individual stability and habituation over time. Brain Behav. Immun. 20, 456459. Hamer, M., Williams, E., Vuonovirta, R., Giacobazzi, P., Gibson, E.L., Steptoe, A., 2006b. The effects of effortreward imbalance on inammatory and cardiovascular responses to mental stress. Psychosom. Med. 68, 408413. Harris, T.O., Borsanyi, S., Messari, S., Stanford, K., Cleary, S.E., Shiers, H.M., Brown, G.W., Herbert, J., 2000. Morning cortisol as a risk factor for subsequent major depressive disorder in adult women. Br. J. Psychiatry 177, 505510. Huppert, F.A., Baylis, N., Kerverne, B. (Eds.), 2004. The science of well-being: integrating neurobiology, psychology and social science. Philos. Trans. R. Soc. London B: Biol. Sci. 359, n 1449. Jennings, J.R., Kamarck, T.W., Everson-Rose, S.A., Kaplan, G.A., Manuck, S.B., Salonen, J.T., 2004. Exaggerated blood pressure responses during mental stress are prospectively related to enhanced carotid atherosclerosis in middle-aged Finnish men. Circulation 110, 21982203. Kahneman, D., Krueger, A.B., Schkade, D., Schwarz, N., Stone, A.A., 2006. Would you be happier if you were richer? A focusing illusion. Science 312, 19081910. Kamarck, T.W., Jennings, J.R., Pogue-Geile, M., Manuck, S.B., 1994. A multidimensional measurement model for cardiovascular reactivity: stability and cross-validation in two adult samples. Health Psychol. 13, 471478. Kamarck, T.W., Debski, T.T., Manuck, S.B., 2000. Enhancing the laboratory-to-life generalizability of cardiovascular reactivity using multiple occasions of measurement. Psychophysiology 37, 533542. Kibler, J.L., Ma, M., 2004. Depressive symptoms and cardiovascular reactivity to laboratory behavioral stress. Int. J. Behav. Med. 11, 8187. Kirschbaum, C., Steyer, R., Eid, M., Patalla, U., Schwenkmezger, P., Hellhammer, D.H., 1990. Cortisol and behavior: 2. Application of a latent state-trait model to salivary cortisol. Psychoneuroendocrinology 15, 297307. Kudielka, B.M., Broderick, J.E., Kirschbaum, C., 2003. Compliance with saliva sampling protocols: electronic monitoring reveals invalid cortisol daytime proles in noncompliant subjects. Psychosom. Med. 65, 313319. Kunz-Ebrecht, S.R., Kirschbaum, C., Marmot, M., Steptoe, A., 2004. Differences in cortisol awakening response on work days and weekends in women and men from the Whitehall II cohort. Psychoneuroendocrinology 29, 516528. Kuper, H., Singh-Manoux, A., Siegrist, J., Marmot, M., 2002. When reciprocity fails: effortreward imbalance in relation to coronary heart disease and health functioning within the Whitehall II study. Occup. Environ. Med. 59, 777784. Lai, J.C., Evans, P.D., Ng, S.H., Chong, A.M., Siu, O.T., Chan, C.L., Ho, S.M., Ho, R.T., Chan, P., Chan, C.C., 2005. Optimism, positive affectivity, and salivary cortisol. Br. J. Health Psychol. 10, 467484. Luecken, L.J., Dausch, B., Gulla, V., Hong, R., Compas, B.E., 2004. Alterations in morning cortisol associated with PTSD in women with breast cancer. J. Psychosom. Res. 56, 1315. Maier, K.J., Waldstein, S.R., Synowski, S.J., 2003. Relation of cognitive appraisal to cardiovascular reactivity, affect, and task engagement. Ann. Behav. Med. 26, 3241. Marsland, A.L., Cohen, S., Rabin, B.S., Manuck, S.B., 2006. Trait positive affect and antibody response to hepatitis B vaccination. Brain Behav. Immun. 20, 261269. McEwen, B.S., Biron, C.A., Brunson, K.W., Bulloch, K., Chambers, W.H., Dhabhar, F.S., Goldfarb, R.H., Kitson, R.P., Miller, A.H., Spencer, R.L., Weiss, J.M., 1997. The role of adrenocorticoids as modulators of immune function in health and disease: neural, endocrine and immune interactions. Brain Res. Rev. 23, 79133.

63
Ostir, G.V., Markides, K.S., Black, S.A., Goodwin, J.S., 2000. Emotional well-being predicts subsequent functional independence and survival. J. Am. Geriatr. Soc. 48, 473478. 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. Psychosom. Med. 63, 210215. Polk, D.E., Cohen, S., Doyle, W.J., Skoner, D.P., Kirschbaum, C., 2005. State and trait affect as predictors of salivary cortisol in healthy adults. Psychoneuroendocrinology 30, 261272. Pressman, S.D., Cohen, S., 2005. Does positive affect inuence health? Psychol. Bull. 131, 925971. Pruessner, J.C., Kirschbaum, C., Meinlschmidt, G., Hellhammer, D., 2003a. Two formulas for computation of the area under the curve represent measures of total hormone concentration versus time-dependent change. Psychoneuroendocrinology 28, 916931. Pruessner, M., Hellhammer, D.H., Pruessner, J.C., Lupien, S.J., 2003b. Self-reported depressive symptoms and stress levels in healthy young men: associations with the cortisol response to awakening. Psychosom. Med. 65, 9299. Schulz, P., Kirschbaum, C., Pruessner, J., Hellhammer, D., 1998. Increased free cortisol secretion after awakening in chronically stressed individuals due to work overload. Stress Med. 14, 9197. Schwarz, N., 1999. Self-reports: how the questions shape the answers. Am. Psychol. 54, 93105. Smyth, J., Ockenfels, M.C., Porter, L., Kirschbaum, C., Hellhammer, D.H., Stone, A.A., 1998. Stressors and mood measured on a momentary basis are associated with salivary cortisol secretion. Psychoneuroendocrinology 23, 353370. Steptoe, A., Marmot, M., 2005. Impaired cardiovascular recovery following stress predicts 3-year increases in blood pressure. J. Hypertens. 23, 529536. Steptoe, A., Marmot, M., 2006. Psychosocial, hemostatic and inammatory correlates of delayed post-stress blood pressure recovery. Psychosom. Med. 68, 531537. Steptoe, A., Wardle, J., 2005a. Positive affect and biological function in everyday life. Neurobiol. Aging 26 (Suppl 1), 108112. Steptoe, A., Wardle, J., 2005b. Cardiovascular stress responsivity, body mass and abdominal adiposity. Int. J. Obes. 29, 13291337. Steptoe, A., Siegrist, J., Kirschbaum, C., Marmot, M., 2004. Effortreward imbalance, overcommitment, and measures of cortisol and blood pressure over the working day. Psychosom. Med. 66, 323329. Steptoe, A., Wardle, J., Marmot, M., 2005a. Positive affect and health-related neuroendocrine, cardiovascular, and inammatory processes. Proc. Natl. Acad. Sci. USA 102, 65086512. Steptoe, A., Brydon, L., Kunz-Ebrecht, S., 2005b. Changes in nancial strain over three years, ambulatory blood pressure, and cortisol responses to awakening. Psychosom. Med. 67, 281287. Stetler, C., Miller, G.E., 2005. Blunted cortisol response to awakening in mild to moderate depression: regulatory inuences of sleep patterns and social contacts. J. Abnorm. Psychol. 114, 697705. Stone, A.A., Shiffman, S., 2002. Capturing momentary, self-report data: a proposal for reporting guidelines. Ann. Behav. Med. 24, 236243. Stone, A.A., Shiffman, S., Schwartz, J.E., Broderick, J.E., Hufford, M.R., 2003. Patient compliance with paper and electronic diaries. Control Clin. Trials 24, 182199. Treiber, F.A., Kamarck, T., Schneiderman, N., Shefeld, D., Kapuku, G., Taylor, T., 2003. Cardiovascular reactivity and development of preclinical and clinical disease states. Psychosom. Med. 65, 4662. Tugade, M.M., Fredrickson, B.L., 2004. Resilient individuals use positive emotions to bounce back from negative emotional experiences. J. Pers. Soc. Psychol. 86, 320333.

ARTICLE IN PRESS
64
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. Psychosom. Med. 58, 447458. Watson, D., Clark, L.A., Tellegen, A., 1988. Development and validation of brief measures of positive and negative affect: the PANAS scales. J. Pers. Soc. Psychol. 54, 10631070. Wessa, M., Rohleder, N., Kirschbaum, C., Flor, H., 2006. Altered cortisol awakening response in posttraumatic stress disorder. Psychoneuroendocrinology 31, 209215. Wright, C.E., Steptoe, A., 2005. Subjective socioeconomic position, gender and cortisol responses to waking in an elderly population. Psychoneuroendocrinology 30, 582590.

A. Steptoe et al.
Wust, S., Federenko, I., Hellhammer, D.H., Kirschbaum, C., 2000. Genetic factors, perceived chronic stress, and the free cortisol response to awakening. Psychoneuroendocrinology 25, 707720. Yehuda, R., Golier, J.A., Kaufman, S., 2005. Circadian rhythm of salivary cortisol in Holocaust survivors with and without PTSD. Am. J. Psychiatry 162, 9981000. Young, E.A., Tolman, R., Witkowski, K., Kaplan, G., 2004. Salivary cortisol and posttraumatic stress disorder in a low-income community sample of women. Biol. Psychiatry 55, 621626.