ARTICLE IN PRESS

Available at www.sciencedirect.com

Psychoneuroendocrinology (2007) 32, 125–132

0306-4530/$ - see frodoi:10.1016/j.psyne

�Corresponding auMedical School, TeviTel.: +44 131 651 394

E-mail address: e

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

Daily hassles and eating behaviour: The role ofcortisol reactivity status

Emily Newman�, Daryl B. O’Connor, Mark Conner

Institute of Psychological Sciences, University of Leeds, Leeds, LS2 9JT, UK

Received 1 August 2006; received in revised form 16 November 2006; accepted 19 November 2006

KEYWORDSStress;Cortisol;Hassles;Snacking;Obesity;Mood;Metabolic syndrome

nt matter & 2006uen.2006.11.006

thor. School of Hot Place, Universit5; fax: +44 131 651

mily.newman@ed

SummaryPrevious research has shown high cortisol reactors to consume a greater amount of snackfoods than low reactors following a laboratory stressor. The current study tested whetherhigh cortisol reactors also consume more snacks than low reactors in response to fieldstressors. Fifty pre-menopausal women completed a laboratory stressor, provided salivasamples to assess cortisol reactor status and then completed daily hassles and snack intakediaries over the next fourteen days. Hierarchical multivariate linear modelling showed asignificant association between daily hassles and snack intake within the overall sample,where an increased number of hassles was associated with increased snack intake. Thissignificant positive association between number of hassles and snack intake was onlyobserved within the high cortisol reactors and not within the low cortisol reactors. Thesefindings suggest that high cortisol reactivity to stress promotes food intake. Furthermore,the eating style variables of restraint, emotional eating, external eating and disinhibitionwere more strongly associated with snack intake in high reactors than in low reactors. Thissuggests that cortisol reactivity may in part account for the moderating role of eating styleon stress-induced eating. The results are discussed within the context of future health risk.& 2006 Elsevier Ltd. All rights reserved.

1. Introduction

It is becoming more apparent that stress and negative affectnot only have direct effects on health but also indirecteffects through behavioural changes, including changes in

Elsevier Ltd. All rights reserved.

ealth in Social Sciences, Oldy of Edinburgh, EH8 9AG, UK.3971.

.ac.uk (E. Newman).

the type and amount of food consumed (e.g., Macht andSimons, 2000; O’Connor et al., 2000; O’Connor and O’Con-nor, 2004). Laboratory and self-report studies demonstratethat individuals respond differently in their eating responseto stress with gender, bodyweight and the eating stylevariables of restraint, emotional eating, external eating anddisinhibition acting as significant moderators of the stres-s–eating relationship (McKenna, 1972; Herman and Polivy,1975; Grunberg and Straub, 1992; Greeno and Wing, 1994;Conner et al., 1999; Oliver et al., 2000; Van Strien et al.,2000; O’Connor et al., 2005). Although, while research has

ARTICLE IN PRESS

E. Newman et al.126

identified a number of important moderators of stress-induced eating, relatively little is known about the under-lying mechanisms.

One possible mechanism for stress-induced eating con-cerns the activity of the hypothalamic–pituitary–adrenalaxis during stress, particularly the release of glucocorticoidsfrom the adrenal cortex. Sapolsky (1998) proposed thatcorticotropic releasing hormone (CRH) and glucocorticoids(GC) have opposing effects on appetite, such that foodintake is inhibited by CRH and promoted by GC production.Direct manipulations of GC levels support their associationwith appetite and food intake. Adrenalectomised ratsunable to secrete GC have been shown to consume smalleramounts of carbohydrate relative to other macronutrients(Laugero, 2001), and GC appears to protect against thehypophagic effects of leptin (Zakrzewska et al., 1997). Inhumans, the administration of glucocorticoids in humans hasbeen shown to increase energy consumption, especiallycarbohydrates and proteins (Tataranni et al., 1996).

Furthermore, the release of GC during stress has beenassociated with increased snack intake. In a laboratoryinvestigation of snack intake in women after stressexposure, Epel et al. (2001) reported that during stressrecovery high cortisol reactors consumed more than lowreactors, especially of high fat, sweet foods. Thereforeindividual differences in the stress–eating response could bedependant on GC reactivity to stress, such that high cortisolreactors consume a greater amount when stressed than dolow reactors. As yet, this cortisol reactivity theory of stress-induced eating has only been tested in the laboratory andnot in the field. To test whether the effect is limited to thelaboratory it is essential to replicate and extend Epel et al.’s(2001) findings in a more natural setting.

Field studies of stress-induced eating usually requireindividuals to complete diary records of workload, hasslesand food intake (e.g., Steptoe et al., 1998; Conner et al.,1999), allowing the researcher to measure natural eatingbehaviour in response to real-life stressors. Previously, diarystudies of stress-induced eating have compared overallsnack intake across high and low stress weeks (e.g., Steptoeet al., 1998). However, by averaging intake across days andweeks subtle daily variations in stress and intake may belost. Multivariate linear modelling enables researchers totest between-person associations and within-person dailyvariations in measures (Affleck et al., 1999), and wouldtherefore facilitate an investigation of the relationshipbetween daily stress and snack intake. Despite thisadvantage, only one previous study has employed thismethod of analysis to examine fluctuations in intake withdaily stress (O’Connor et al., 2005).

The present study aimed to test whether the relationshipbetween hassles and snacks outside the laboratory differsbetween high and low cortisol reactors as an extension ofEpel et al.’s (2001) study and a test of whether GC releasecould account for variations in stress-induced eating. Thestudy also aimed to test whether the relationship betweeneating style and snacking differed according to cortisolreactivity status. Following the procedure of Epel et al.(2001), pre-menopausal women were exposed to laboratorystressors to establish cortisol reactivity status and requiredto report daily hassles and snack intake in diaries over 2weeks. Because of reported gender differences in cortisol

reactivity to stress (e.g., Kudielka and Kirschbaum, 2005)and a greater prevalence of stress-induced food intake infemales (e.g., Grunberg and Straub, 1992), only femaleswere included in the current study. It was predicted thathigh cortisol reactors would show a stronger positiveassociation between daily hassles and snack intake thanlow reactors.

2. Methods

2.1. Participants

Fifty-five pre-menopausal women completed the laboratorytasks. Of these, four did not return both diaries and onewoman did not produce sufficient saliva for analysis.Therefore complete data were obtained and is reportedfrom 50 women. The participants had a mean age of 33.96years (SD ¼ 6.18) and mean body mass index (BMI) of 23.34(SD ¼ 3.62). Participants were recruited using a CountyCouncil Bulletin in an advert to take part in a ‘Women andEating Study’. Exclusion criteria were based on factorspreviously shown to affect baseline or reactive cortisollevels. Participants were not included in the study if theyhad been diagnosed with a neuroendocrine or metabolicdisorder, had a history of eating disorders or depression(Ellenbogen et al., 2002), were postmenopausal or takingoral contraceptives (Kirschbaum and Hellhammer, 1989;Pruessner et al., 1997). Menstrual cycle phase was leftrandom, although cortisol levels are reportedly higheraround the luteal phase (Kirschbaum et al., 1999). Partici-pants were paid 20 pounds (approximately $36) forcompleting all parts of the study.

2.2. Procedure

Participants’ cortisol reactivity was individually tested inthe laboratory in the afternoon, to control for the wakingresponse (Pruessner et al., 1997) and because cortisol stressreactivity is greater in the afternoon (Kirschbaum andHellhammer, 1989). Moreover, time of day has also beenfound to influence levels of anxiety during stressfulencounters (Willis, O’Connor and Smith, 2005). They wereasked not to smoke during the hour before testing due to theassociated rise in cortisol (Morgan et al., 2004), nor toexercise or drink alcohol on the test day. On arrival at thelaboratory, the participant was given an information sheetabout the study before providing written consent. She wasmeasured and weighed, and asked to provide a first salivasample (baseline measure one). The participant thenrelaxed for fifteen minutes with a selection of magazineswhile listening to a ‘Classical Chillout’ compact disc (CircaRecords Ltd, 2001), before providing a second baselinesaliva sample. The 15-min stress induction procedure wasthen conducted. The stress manipulation was based on theTrier Social Stress Test (Kirschbaum et al., 1993). For thefirst five minutes, each participant was asked to prepare a5-min presentation of their opinion on a controversial topicfrom a list (topics included abortion, sexual equality andcannabis legalisation), for later assessment by psychologistswho were experts in body language. The participant thenperformed the presentation for five minutes in front of the

ARTICLE IN PRESS

Cortisol and snacking 127

experimenter and a videocamera; although, the perfor-mance was not actually recorded or assessed. The partici-pant was prompted to continue if the presentation stoppedfor any length of time, until a total of 5min had passed. Forthe subsequent 5min, the participant was required to countbackwards serially in thirteens from the number 1022 asquickly and accurately as possible. If an incorrect responsewas made, the participant was required to restart thesubtraction from 1022. Such stress protocols involving bothan assessment and mathematical component have beenshown to be most effective in inducing cortisol reactivity(Dickerson and Kemeny, 2004).

After the stress procedure, a third saliva sample wastaken. The participant was asked to complete a question-naire battery for 40min. During this time, four more salivasamples were taken, at 10-min intervals. The participantwas then led into a different room, and asked to relax againfor 20min before an eighth saliva sample was taken. Theparticipant was then provided with the two weekly diaries infreepost envelopes. Diaries contained brief instructions onhow they should be completed; however the participant wasalso given an explanation of how to complete each dailyentry on receipt of the diaries. Participants were contactedfor debriefing after receipt of the second diary.

2.3. Measures

State anxiety was measured before and after the stressmanipulation using the Shortened State Anxiety Inventory(STAI; Marteau and Bekker, 1992). The participants alsorecorded how stressful they had found the manipulation,using a seven-point anchored scale from ‘not at all’ to‘extremely’. The questionnaire battery consisted of theDutch Eating Behavior Questionnaire (DEBQ; Van Strien etal., 1986) and disinhibition scale of the Three Factor EatingQuestionnaire (TFEQ; Stunkard and Messick, 1985). TheDEBQ contains 33 items designed to measure dietaryrestraint (10 items, e.g., ‘Do you deliberately eat foodsthat are slimming?’), emotional eating (13 items, e.g., ‘Doyou have a desire to eat when you are upset?’) and externaleating (10 items, e.g., ‘If you see others eating, do you alsohave a desire to eat?’) and has previously been validated in aBritish sample by Wardle (1987). The disinhibition scale ofTFEQ contains 16 items (e.g., ‘Sometimes when I starteating, I just can’t seem to stop’) designed to measuretendency to disinhibit food intake. The reliability of theTFEQ has been previously validated by Stunkard and Messick(1985). In the current sample, each of the scales exhibitedvery good internal reliability (range a ¼ 84–.92). In thefield, participants completed two weekly diaries over twoconsecutive weeks. Daily mood was recorded using theshortened Positive and Negative Affect Scale (PANAS;MacKinnon et al., 1999). Participants reported their dailyhassles, rating the intensity of each on four-point anchoredscale from ‘not at all’ to ‘very much’ (scored 0–4). Previousresearch suggests that snack intake is more susceptible tochange with daily hassles than is meal intake (Conner et al.,1999; Crowther et al., 2001; O’Connor and O’Connor, 2004)and so participants were only required to record snackintake. Snacks were defined within the diaries as foodsconsumed between meals and thus discriminated from

meals. The participants were asked to complete their dailyentries at the end of each day, and to return each weeklydiary by post as soon as it was completed, using the pre-addressed freepost envelopes.

2.4. Cortisol measurements

The participants provided salivary samples of cortisol, usingsalivette tubes (Sarstedt, UK). Salivary samples provide anon-invasive measure of free, bio-available cortisol levels(Kirschbaum and Hellhammer, 1989; De Weerth et al., 2003).The salivette contains a cotton dental roll inside a plastictube, which the participant is required to place in the mouthfor 30 to 45 s before replacing in the tube. Salivettes werefrozen at –20 1C on the same day, to preserve samplestability as well as possible (Groschl et al., 2001). Afterdefrosting and spinning, the saliva samples were testedusing a fluorescence immunoassay with an autodelfia kit.Assays were performed over three days with samples fromeach participant tested on the same immunoassay plate tominimise inter-assay variation. Boxplots revealed that threewomen had outlying high cortisol reactivity levels; however,given that reactivity was not treated as a continuousvariable as participants were divided into high and lowreactor groups, this did not impact on the statistical analysis(cf. Epel et al., 2001).

2.5. Data analysis

Cortisol reactivity was determined using the differencebetween mean baseline cortisol level and maximum levelafter the stressor. Those individuals who increased incortisol levels were classified as high reactors, while thosewho showed no change or decreased in levels from baselinewere classified as low reactors. Hierarchical multivariatelinear modelling was conducted to test the relationshipbetween daily hassles and snack intake and the relationshipbetween eating style and snack intake in the field, using theprogram HLM6 (Raudenbush et al., 2004). The datacontained a two-level hierarchical structure, Level 1 beingthe within-person variation (e.g., daily patterns in thenumber snacks consumed, the number of hassles experi-enced), and Level 2 being the between-person variability(e.g., eating style). The 50 participants with usable datasetsprovided a total of 700 days of data.

3. Results

3.1. Stress ratings

The stressfulness ratings of the stress procedure rangedfrom 1 (not at all stressful) to 7 (extremely stressful) on the7 point scale, with a mean of 4.78 (SD ¼ 1.43), indicatingthat the stressor was moderately, but not extremely,stressful. The mean state anxiety level was 13.94(SD ¼ 3.32) before the stress manipulation, and 16.76(3.68) after the manipulation. A repeated measures t-testindicated that this difference was significant (t(49) ¼ 4.96,po0.001), indicating that the stress manipulation wassuccessful in increasing anxiety.

ARTICLE IN PRESS

E. Newman et al.128

3.2. Cortisol reactivity

Cortisol reactivity was measured by taking the differencebetween the average of the two baseline samples and thepeak response (between 10 and 40min following the start ofthe stress protocol). There was an average cortisol increaseof 1.36 nmol/l (SD ¼ 3.77). A previous study has shown anaverage cortisol increase of �7.00 nmol/l in women after40min (Kirschbaum et al., 1992), therefore average reac-tivity was lower in the current study. Reactivity ranged from–3.39 to +13.43 nmol/l, indicating that some participantsshowed a decline in cortisol levels following baseline.Twenty-six women showed an increase in cortisol levels,and were classified as high reactors (mean reactivi-ty ¼ 3.69 nmol/l). Twenty-three women who showed adecrease in levels and one participant who showed nochange were classified as low reactors (mean reactivi-ty ¼ �1.18 nmol/l). Fig. 1 shows the cortisol reactivityprofiles for high and low reactors. High and low reactorssignificantly differed in reactivity from baseline(t(48) ¼ �6.18, po0.001), but not in average baselinevalues (t(48) ¼ 0.08, n.s.). There was no difference in stressratings of the manipulation between the reactor groups(t(48) ¼ �0.93, n.s.), but there was a significant differencein state anxiety post-manipulation (t(48) ¼ �3.11,po0.01), indicating that high reactors had a greater anxietyrating following the stress manipulation than did lowreactors.

3.3. Relationship between daily hassles and snackintake

The number of reported daily hassles ranged from zero tofive, with a mode of one. The number of daily hassles waspositively skewed so this variable was dichotomised: Nohassles or one hassle experienced per day was coded as low,and 2 or more coded as high, to provide the most evendivision of low and high numbers of hassles (65.7% dayscoded as low hassle days, and 34.3% coded as high hassledays). The rated intensities of each hassle were summed foreach day to give a total daily hassle intensity score. Thesehassle intensity scores ranged from 0 to 16 and were alsopositively skewed. The variable was dichotomised into highand low hassle intensity days, with intensity scores of 0 to 2

0

2

4

6

8

10

12

14

0 15 25 35 45 55

Minutes

Cor

tiso

l (nm

ol/l)

Low reactorsHigh reactors

-15

Figure 1 Cortisol reactivity profiles of high and low reactorsduring the laboratory session.

coded as low, and 3 to 16 as high (47.1% days were coded aslow intensity and 52.9% as high intensity). Because thenumber and intensity of hassles were dichotomised, boththese variables were entered as uncentred variables in themultivariate models. The number of daily snacks consumedranged from 0 to 6, with a mean of 1.84 snacks per day.

The effect of the number and intensity of daily hassles onoverall snack intake was tested using a level one model,with the number and intensity of hassles as predictors. Thismodel is expressed as

Yi ¼ b0 þ b1 þ �i,

where Yi is the outcome variable of the number of dailysnacks, b0 the intercept, b1 the slope for the level onepredictor variable and ei the random error term.

Each level one predictor variable (number of hassles,intensity of hassles, negative affect and positive affect) wasentered individually into the equation, rather than all beingentered simultaneously. Table 1 shows the associationbetween each predictor variable and snack intake.

Table 1 shows that across the sample snack intake wassignificantly associated with the number of hassles, intensityof hassles and negative affect. An increase in hassleintensity, number and negative affect was associated withincreased snack intake. Daily positive affect score wasunrelated to snack intake.

3.4. Daily hassles and intake in high and lowcortisol reactors

To test whether the relationship between hassles and snackintake differed between the low and high peak cortisolreactors, the same hierarchical modelling was also con-ducted separately for the two groups (shown in Table 1).Within the low reactors, snack intake was not significantlyassociated with the number of daily hassles (b ¼ 0.01,t ¼ 0.14, n.s.) or the intensity of hassles (b ¼ �0.14,t ¼ �1.84, n.s.). In high peak reactors, there weresignificant positive associations between hassle numberand snack intake (b ¼ 0.39, t ¼ 3.96, po0.01), and hassleintensity and snack number (b ¼ 0.51, t ¼ 6.30, po0.001),indicating that snack intake increased with a greaternumber and intensity of hassles in the high reactor group.Beta coefficients were significantly different between thehigh and low reactor groups for number of hassles(t(46) ¼ 2.83, po0.01) and numbers of hassles (t(46) ¼5.81, po0.01). With regards to daily mood, snack intake wassignificantly positively associated with negative affect(b ¼ 0.04, t ¼ 2.75, po0.05) and negatively associatedwith positive affect (b ¼ �0.01, t ¼ �2.61, po0.05) inthe low reactors only indicating that snack intake increasedwith negative mood and decreased with positive mood inlow and not in high peak reactors.

3.5. Effect of eating style on snack consumption

To test whether snack consumption was predicted by theeating style variables and if this relationship differedaccording to cortisol reactivity status, each eating style

ARTICLE IN PRESS

Table 1 Associations between daily stress and snack intake in the overall sample and within high and low reactor groupsseparately.

Predictor Overall sample High reactors Low reactors

b SE t b SE t b SE t

Number of hassles 0.324 0.086 3.76�� 0.387 0.098 3.96�� 0.012 0.089 0.14Intensity of hassles 0.240 0.079 3.06�� 0.506 0.080 6.30�� �0.143 0.078 �1.84Negative affect 0.043 0.014 3.15�� 0.027 0.014 1.95 0.043 0.016 2.75�

Positive affect 0.001 0.010 0.12 0.016 0.011 1.52 �0.028 0.011 �2.61�

Note: b coefficients are unstandardised.�po0.05.��po0.01.

Table 2 Main effects of eating style on snack intake in the overall sample and within high and low reactor groups separately.

Eating style predictor Overall sample High reactors Low reactors

b SE t b SE t b SE T

Restraint 0.075 0.011 6.83�� 0.089 0.011 7.85�� 0.030 0.010 3.12��

Emotional eating 0.041 0.010 4.16�� 0.055 0.012 4.62�� 0.025 0.007 3.61��

External eating 0.041 0.017 2.47� 0.046 0.021 2.19� 0.026 0.012 2.21�

Disinhibition 0.101 0.025 4.08�� 0.127 0.030 4.23�� 0.041 0.019 2.19�

Note: b coefficients are unstandardised.�po0.05.��po0.01.

Cortisol and snacking 129

variable was added individually to

Yi ¼ b0 þ b1 þ �ij,

where Yi is the outcome variable of the number of dailysnacks, b0 the intercept, b1 the slope for the level twoeating style predictor variable and eij the random errorterm.

Table 2 shows the relationship between each of the eatingstyle measures and daily snack intake in the overall sampleand within the high and low cortisol reactor groupsseparately.

Snack intake was significantly associated with dietaryrestraint (b ¼ 0.08, t ¼ 6.83, po0.01), emotional eating(b ¼ 0.04, t ¼ 4.16, po0.01), disinhibition (b ¼ 0.10,t ¼ 4.08, po0.01) and external eating (b ¼ 0.04, t ¼ 2.47,po0.01) in the overall sample. In each case, an increase inthe eating style variable was associated with an increase inthe number of snacks consumed. Table 2 also shows that allthe eating style variables were significantly associated withsnack intake in both the high and low cortisol reactors, butthat the associations were stronger within the high reactors.A comparison of the beta coefficients between the groupsshowed that associations were significantly stronger withinthe high reactors for restraint (t(46) ¼ 3.97, po0.01),emotional eating (t(46) ¼ 2.16, po0.01) and disinhibition(t(46) ¼ 2.42, po0.01), but not external eating style(t(46) ¼ 0.83, n.s.).

4. Discussion

The aim of the current study was to test whether therelationship between stress and food intake would differbetween high and low cortisol reactors when tested in thefield. The results of a previous study indicated that eatingresponse to stress in the laboratory differed between highand low cortisol reactor groups, such that high cortisolreactors consumed a greater amount of food, especially highfat, sweet foods (Epel et al., 2001). The results of thecurrent study support Epel et al.’s findings. A greaternumber and intensity of daily hassles was associated withincreased snack intake across the sample. However, thispositive association only remained significant among thehigh cortisol reactor group when participants were testedseparately according to reactivity status. The findingstherefore suggest that a difference in eating response tostress between high and low reactors can also be observedoutside of the laboratory.

The results from the current study provide further supportfor the role of GC activity in appetite (e.g., Tataranni et al.,1996). Furthermore, the findings suggest that not only doesthe administration of GC increase food intake, but that therelease of GC during the physiological stress response alsopromotes increased intake. Sapolsky (1998) proposed thatCRH and GC have opposing effects on appetite, so that therelease of CRH at the start of the HPA axis response has ananorectic effect, but the later release of GC promotes food

ARTICLE IN PRESS

E. Newman et al.130

intake during the recovery stage for the replenishment ofenergy required during the ‘fight or flight’ response. Whilethis mechanism may have been adaptive in our evolutionarypast, the effects of GC may be less beneficial in response tothe modern day psychological stressors. In fact, continuousexposure to stressors could contribute to obesity. Therelease of GC appears to promote the release of insulin(Dallman et al., 1994). This combination of insulin releaseand food consumption increases the likelihood that con-sumed energy will be stored as fat, and particularly aroundthe abdominal region where GC receptors are abundant(Strack et al., 1995; Laugero, 2001). Therefore, the appetitestimulating effects of GC may have adverse consequencesfor obesity and health.

Moreover, it has been suggested that the effects of stresson food intake may also contribute to increased risk ofmetabolic syndrome (Epel et al., 2004). The characteristicsof which include obesity, insulin resistance (or type IIdiabetes), high blood pressure, high blood triglyceride levelsand low HDL cholesterol levels. Recently, Epel et al. (2004)in a longitudinal study examined the effects of self-reportedstress-eating tendencies (more-eaters vs. less-eaters) onchanges in cortisol, insulin, adiposity, lipid levels and foodintake from baseline to exam-stress periods in medicalstudents. These authors found that increases in weight,cortisol, insulin, and lipid profile in response to stress wereonly observed in stress ‘more-eaters’ and not in the ‘less-eaters’. Taken together with the current findings, theseresults indicate that stress-related changes in food intake, ifmaintained overtime, may be a risk factor for the develop-ment of metabolic syndrome. Larger and more comprehen-sive longitudinal investigations are required in order todetermine the causal pathways.

It is as yet unclear exactly how GC affects food intake.One possibility is that they initiate the release of neuropep-tide Y, a known appetite stimulant (Sainsbury et al., 1997).It is also possible that GC protects against the hypophagiceffects of leptin. Early tests have revealed that leptin is onlyeffective in promoting weight loss in adrenalectomisedanimals (Zakrzewska et al., 1997), suggesting that GCcounteract its effects. It is also unclear why certainindividuals are more reactive to stress in terms of cortisoloutput. It is likely that psychological factors are involved,since the results of the present study reported that self-report anxiety after the stressor was greater in highreactors. Moreover, these results also point to diathesis-stress mechanisms that suggest that psychological vulner-abilities, when activated by stress, may result in negativeoutcomes. For example, coping styles, the behavioural andcognitive responses individuals use when they encounterstress, in particular, have been shown to have well-established moderating effects on an individual’s responseto a stressful encounter (cf., O’Connor and O’Connor, 2003).It would therefore be beneficial for future research toaddress the psychological factors (such as coping) associatedwith the cortisol response to stress and whether this differsaccording to stressor type (e.g., ego-threatening, inter-personal, work-related).

Previous research has highlighted that eating stylevariables are significant moderators of stress-induced eat-ing, such that restrained, emotional and external eaters andthose with a high tendency to disinhibit food intake are

more likely to increase food intake during stress (e.g.,Conner et al., 1999; Oliver et al., 2000; Van Strien et al.,2000; O’Connor et al., 2005). The results of the presentstudy indicated that the positive associations between snackintake and restrained, emotional and disinhibited eatingstyle variables were greater among the high cortisol reactorsthan within the low reactors. One possible explanation forthis finding is that individuals high in these eating stylevariables produce high levels of cortisol, which then has anappetitive effect. Previous research has reported that highdietary restraint individuals have greater salivary andurinary cortisol levels than low restraint individuals (e.g.,McLean et al., 2001; Anderson et al., 2002), which may bedue to the stress caused by diet monitoring. However, therelationship between cortisol output and other eating stylevariables has not been investigated. Therefore individualdifferences in cortisol reactivity to stress may contribute toan understanding of the moderating effect of these eatingstyle measures on eating response to stress, though furtherinvestigations of the relationship between eating style andcortisol output are needed to further elucidate this finding.

We also found fluctuations in daily mood only impacted onsnack intake in the low cortisol reactors: negative affect wasassociated with a hyperphagic snacking response, whereas,positive affect was associated with a hypophagic snackingresponse. These findings are important and indicate that themechanisms associated with between-meal snacking aredifferent in the high and low cortisol reactors with moodbeing an important driver in the low reactors and cortisol inthe high reactors. To the best of our knowledge, this is alsothe first study to show the differential effects of positiveand negative mood on food intake within a naturalisticsetting. The role of mood in understanding food intakewithin the context of cortisol reactivity status requiresfurther examination.

The present study used a modified version of the TrierSocial Stress Test (Kirschbaum et al., 1999), where partici-pants were asked to present their opinion about acontroversial topic for 5min to the experimenter. This firstpart of the test differed from the original, which requiresparticipants to convince a three-person panel of male andfemale confederates that they are the right person for aspecified job. Although the modified version combined anassessment task with a mathematical task to maximisecortisol reactivity (Dickerson and Kemeny, 2004), it is likelythat the changes to the presentation task account for thelower cortisol reactivity levels compared with previousresearch (Kirschbaum et al., 1992). Therefore it would beuseful to replicate the study using the original version of theTrier Social Stress Test. As a result of the lower cortisolreactivity in the current study, the reactor groups may bemore accurately described as reactors and non-reactors,rather than high and low reactors. However, the cortisolprofiles of the reactor groups are similar to those reportedby Epel et al. (2001), therefore the terms high and lowreactors have been used to maintain consistency with thisprevious study. It should also be noted that menstrual phasewas not identified in the present study, although evidencehas suggested that cortisol reactivity can vary over the cycle(Kirschbaum et al., 1999). It would therefore be useful forfuture research to include some measure of menstrual cyclephase. In addition, this study only examined the effects of

ARTICLE IN PRESS

Cortisol and snacking 131

cortisol reactivity in women; consequently, it remainsunknown whether these effects will generalise to a malesample. Recent research, using a diary methodology, hasfound similar effects of daily hassles on between-mealsnacking in men and women, although, this study did notassess cortisol reactivity to stress (O’Connor et al., 2005).Therefore, it would be useful if further research examinedthe effects of stress and cortisol on eating behaviour in asample of men and considered utilising alternatives to theself-report diary methodology (e.g., electronic techniques)assessing these variables at random time points throughoutday.

In conclusion, the results of this study suggest that theassociation between stress and snack intake is greater inhigh cortisol reactors than low reactors, when investigatedin the field. This suggests that the release of glucocorticoidsduring the HPA axis stress response promotes the intake offood. However, this may have adverse consequences for thedevelopment of central obesity and metabolic syndrome.Future research ought to explore further the psychologicalfactors that predict cortisol reactivity to stress.

Acknowledgements

We would like to thank Duncan Talbot at Unilever inColworth, UK for conducting the cortisol analysis.

References

Affleck, G., Zautra, A., Tennen, H., Armeli, S., 1999. Multileveldaily process designs for consulting and clinical psychology: apreface for the perplexed. J. Consult. Clin. Psych. 67, 746–754.

Anderson, D.A., Shapiro, J.R., Lundgren, J.D., Spataro, L.E., Frye,C.A., 2002. Self-reported dietary restraint is associated withelevated levels of salivary cortisol. Appetite 38, 13–17.

Conner, M., Fitter, M., Fletcher, W., 1999. Stress and snacking: adiary study of daily hassles and between-meal snacking. Psychol.Health. 14, 51–63.

Crowther, J.H., Snaftner, J., Bonifazi, D.Z., Shepherd, K.L., 2001.The role of daily hassles in binge eating. Int. J. Eat. Disorder. 29,449–454.

Dallman, M.F., Akana, S.F., Bradbury, M.J., Strack, A.M., Hanson,E.S., Scribner, K.A., 1994. Regulation of the hypothalamo–pitui-tary–adrenal axis during stress: feedback, facilitation andfeeding. Semin. Neurosci. 6, 205–213.

De Weerth, C., Graat, G., Buitelaar, J.K., Thijssen, J.H.H., 2003.Measurement of cortisol in small quantities of saliva. Clin. Chem.49, 658–660.

Dickerson, S.S., Kemeny, M.E., 2004. Acute stressors and cortisolresponses: a theoretical integration and synthesis of laboratoryresearch. Psychol. Bull. 130, 355–391.

Ellenbogen, M.A., Schwartzman, A.E., Stewart, J., Walker, C., 2002.Stress and selective attention: the interplay of mood, cortisollevels and emotional information processing. Psychophysiology39, 723–732.

Epel, E., Jimenez, S., Brownell, K., Stroud, L., Stoney, C., Niaura,R., 2004. Are stress eaters at risk for the metabolic syndrome?Annals of the New York Academy of Sciences 1032, 208–210.

Epel, E., Lapidus, R., McEwen, B., Brownell, K., 2001. Stress mayadd bite to appetite in women: a laboratory study of stress-induced cortisol and eating behavior. Psychoneuroendocrinology26, 37–49.

Greeno, C.G., Wing, R.R., 1994. Stress-induced eating. Psychol.Bull. 115, 444–464.

Groschl, M., Wagner, R., Rauh, M., Dorr, H.G., 2001. Stability ofsalivary steroids: the influences of storage, food and dental care.Steroids 66, 737–741.

Grunberg, N.E., Straub, R.O., 1992. The role of gender and taste classin the effects of stress on eating. Health Psychol. 11, 97–100.

Herman, C.P., Polivy, J., 1975. Anxiety, restraint, and eatingbehavior. J. Abnorm. Psychol. 84, 666–672.

Kirschbaum, C., Hellhammer, D.H., 1989. Salivary cortisol inpsychobiological research: an overview. Neuropsychobiology22, 150–169.

Kirschbaum, C., Bartussek, D., Strasburger, C.J., 1992. Cortisolresponses to psychological stress and correlations with person-ality traits. Pers. Individ. Dif. 13, 1353–1357.

Kirschbaum, C., Kudielka, B.M., Gaab, J., Schommer, N.C.,Hellhammer, D.H., 1999. Impact of gender, menstrual cyclephase, and oral contraceptives on the activity of the hypotha-lamus–pituitary–adrenal axis. Psychosom. Med. 61, 154–162.

Kirschbaum, C., Pirke, K., Hellhammer, D.H., 1993. The ‘Trier SocialStress Test’—A tool for investigating psychobiological stressresponses in a laboratory setting. Neuropsychobiology 28,76–81.

Kudielka, B.M., Kirschbaum, C., 2005. Sex differences in HPA axisresponse to stress: a review. Biol. Psychol. 69, 113–132.

Laugero, K.D., 2001. A new perspective on glucocorticoid feedback:relation to stress, carbohydrate feeding and feeling better. J.Neuroendocrinol. 13, 827–835.

Macht, M., Simons, G., 2000. Emotions and eating in everyday life.Appetite 35, 65–71.

Mackinnon, A., Jorm, A.F., Christensen, H., Korten, A.E., Jacomb,P.A., Rodgers, B., 1999. A short form of the positive and negativeaffect schedule: evaluation of factorial validity and invarianceacross demographic variables in a community sample. Pers.Individ. Dif. 27, 405–416.

Marteau, T.M., Bekker, H., 1992. The development of a six-itemshort-form of the state scale of the Spielberger State-TraitAnxiety. Brit. J. Clin. Psychol. 31, 301–306.

McKenna, R.J., 1972. Some effects of anxiety level and food cues onthe eating behavior of obese and normal subjects: a comparisonof the schachterian and psychosomatic conceptions. J. Pers. Soc.Psychol. 22, 311–319.

McLean, J.A., Barr, S.I., Prior, J.C., 2001. Cognitive dietary restraintis associated with higher urinary cortisol excretion in healthypremenopausal women. Am. J. Clin. Nutr. 73, 7–12.

Morgan, T.M., Crawford, L., Stoller, A., Toth, D., Yeo, K.J., Baron,J.A., 2004. Acute effects of nicotine on serum glucose, insulin,growth hormone, and cortisol in healthy smokers. Metabolism53, 578–582.

O’Connor, D.B., O’Connor, R.C., 2004. Perceived changes in foodintake in response to stress: the role of conscientiousness. StressHealth 20, 279–291.

O’Connor, D.B., O’Connor, R.C., White, B.L., Bundred, P.E., 2000. Jobstrain and ambulatory blood pressure in British general practi-tioners: a preliminary study. Psychol. Health Med. 5, 241–250.

O’Connor, D.B., Conner, M.T. Jones, F., 2005. Effects of stress oneating behaviour: an integrated approach. End of Award Reportto Economic and Social Research Council.

O’Connor, R.C., O’Connor, D.B., 2003. Predicting hopelessness andpsychological distress: the role of perfectionism and coping. J.Counseling Psychol. 50, 362–372.

Oliver, G., Wardle, J., Gibson, L., 2000. Stress and food choice: alaboratory study. Psychosom. Med. 62, 853–865.

Pruessner, J.C., Gaab, J., Hellhammer, D.H., Lintz, D., Schommer,N., Kirschbaum, C., 1997. Increasing correlations betweenpersonality traits and cortisol stress responses obtained by dataaggregation. Psychoneuroendocrinology 22, 615–625.

Raudenbush, S.W., Bryk, A.S., Cheong, Y.F., Congdon Jr., R.T., 2004.HLM 6: hierarchical linear and nonlinear modeling. ScientificSoftware International, Lincolnwood, IL.

ARTICLE IN PRESS

E. Newman et al.132

Sainsbury, A., Cusin, I., Rohner-Jeanrenaud, F., Jeanrenaud, B.,1997. Adrenalectomy prevents the obesity syndrome producedby chronic central neuropeptide Y infusion in normal rats.Diabetes 46, 209–214.

Sapolsky, R.M., 1998. Why zebras don’t get ulcers: an updated guideto stress, stress-related diseases, and coping. W.H. Freeman &Co., New York.

Steptoe, A., Lipsey, Z., Wardle, J., 1998. Stress, hasslesand variations in alcohol consumption, food choice andphysical exercise: a diary study. Br. J. Health Psychol. 3,51–63.

Strack, A.M., Sebastian, R.J., Schwartz, M.W., Dallman, M.F., 1995.Glucocorticoids and insulin: reciprocal signals for energybalance. Am. J. Physiol. 268, R142–R149.

Stunkard, A.J., Messick, S., 1985. The Three-Factor Eating Ques-tionnaire to measure dietary restraint disinhibition and hunger.J. Psychosom. Res. 29, 71–83.

Tataranni, P.A., Larson, D.E., Snitker, S., Young, J.B., Flatt, J.P.,Ravussin, E., 1996. Effects of glucocorticoids on energy

metabolism and food intake in humans. Endocrinol. Metab.271, E317–E325.

Van Strien, T., Cleven, A., Schippers, G., 2000. Restraint, tendencytoward overeating and ice cream consumption. Int. J. Eat.Disorder. 28, 333–338.

Van Strien, T., Frijters, J.E.R., Bergers, G.P.A., Defares, P.B., 1986.The Dutch Eating Behavior Questionnaire (DEBQ) for assessmentof restrained, emotional, and external eating behavior. Int. J.Behav. Med. 5, 295–315.

Wardle, J., 1987. Eating style: a validation study of the Dutch EatingBehavior Questionnaire in normal subjects and women witheating disorders. J. Psychosom. Res. 31, 161–169.

Willis, T.A., O’Connor, D.B., Smith, L., 2005. The influence ofmorningness–eveningness on anxiety and cardiovascular re-sponses to stress. Physiol. Behav. 85, 125–133.

Zakrzewska, K.E., Cusin, I., Sainsbury, A., Rohner-Jeanrenaud, F.,Jeanrenaud, B., 1997. Glucocorticoids as counterregulatoryhormones of leptin: toward an understanding of leptin resis-tance. Diabetes 46, 717–719.