Effects of daily hassles and eating style on eating behavior

Health Psychology 2008, Vol. 27, No. 1(Suppl.), S20 –S31 Copyright 2008 by the American Psychological Association 0278-6133/08/$12.00 DOI: 10.1037/0278-6133.27.1(Suppl.).S20 Effects of Daily Hassles and Eating Style on Eating Behavior Daryl B. O’Connor, Fiona Jones, Mark Conner, and Brian McMillan Eamonn Ferguson University of Nottingham University of Leeds Objective: This study investigated the daily hassles-eating behavior relationship and its moderators in a naturalistic setting. Design: A multilevel diary design was used to examine day-to-day within-person effects of daily hassles on eating behavior (N ⫽ 422), together with the individual and simultaneous influence of potential moderating variables. Main Outcome Measures: Daily diary reports of betweenmeal snacking, fruit and vegetable consumption and perceived variations in daily food intake. Results: The results showed daily hassles were associated with increased consumption of high fat/sugar snacks and with a reduction in main meals and vegetable consumption. Ego-threatening, interpersonal and work-related hassles were associated with increased snacking, whereas, physical stressors were associated with decreased snacking. The overall hassles-snacking relationship was significantly stronger and more positive at high compared to low levels of restraint, emotional eating, disinhibition, external eating and in females and obese participants. Simultaneous consideration of these moderators indicated that emotional eating was the pre-eminent moderator of the hassles-snacking relationship. Conclusion: Daily hassles were associated with an increase in unhealthy eating behavior. These changes may indicate an important indirect pathway through which stress influences health risk. Keywords: stress, binge eating, obesity, health behaviors, emotional eating There is increasing evidence to indicate that stress affects health directly, through autonomic and neuroendocrine responses, but also indirectly, through changes to health behaviors (e.g., Jones & Bright, 2001; O’Connor, O’Connor, White, & Bundred, 2000). Stress may indirectly contribute to both cardiovascular disease (CVD) and cancer risk to the extent that it produces deleterious changes in diet and/or helps maintain unhealthy eating behaviors such as high fat intake, or low fiber or fruit/vegetable intake. Recent findings have suggested that high levels of stress can be associated with both increased (e.g. saturated fat consumption) and decreased (e.g. overall calories) food intake (Wardle, Steptoe, Oliver, & Lipsey, 2000). Other research has found stress to be associated with an increase in food consumed as snacks in adults (Conner, Fitter, & Fletcher, 1999; O’Connor & O’Connor, 2004) and adolescents (Cartwright et al., 2003). The present study set out to further elucidate the complex relationship between stress (assessed as hassles), eating and health risk in a sample of employed men and women in a naturalistic setting using a multilevel prospective diary design. Previous research into the effects of stress on eating behavior has been overly reliant on laboratory-based, cross-sectional methodologies that have used single indices (i.e., assessed stressful life events over the previous year or 5 years) or “snapshot” measurements of stress (i.e., assessed perceptions of stress over the past two weeks). Such approaches have ignored the growing body of evidence showing that fluctuations in within-person stressful daily hassles are important in understanding stress-outcome processes (e.g., Affleck, Tennen, Urrows, & Higgins, 1994; Dancey, Taghavi, & Fox, 1998; Delongis, Coyne, Dakof, Folkman, & Lazarus, 1982; DeLongis, Folkman, & Lazarus, 1988; Filfield et al., 2004; Kanner, Coyne, Schaefer, & Lazarus, 1981; Sher, 2004). For example, early work by Kanner et al (1981) suggested that indices of (life) stress provide no understanding of what actually happens in day-to-day life and it is “day-to-day events that ultimately have proximal significance for health outcomes and whose accumulative impact. . .should be assessed” (p. 3). The use of open-ended diaries allows respondents to record day-to-day minor life events or hassles that are part of everyday life and have the advantage of not constraining respondents to a limited number of events. Hassles are events, thoughts or situations which, when they occur produce negative feelings such as annoyance, irritation, worry or frustration, and/or make you aware that your goals and plans will be more difficult or impossible to achieve (see Conner et al., 1999; Delongis et al., 1982; Jones, O’Connor, Conner, McMillan, & Ferguson, 2007). Thus in the current study, measuring hassles allowed a large number of stress occurrences in each respondent to be identified, making it possible to look at whether snacking coincided with hassles on a number of occasions, enabling a realistic view of eating in response to stress. Within this context, the specific aims of the study are now described in turn. Daryl B. O’Connor, Fiona Jones, Mark Conner, and Brian McMillan, Institute of Psychological Sciences, University of Leeds, Leeds, England; Eamonn Ferguson, Department of Psychology, University of Nottingham, Nottingham, England. This research was funded by a grant from the UK Economic and Social Research Council to Daryl B. O’Connor, Mark Conner, and Fiona Jones (Award No. RES-000-23-0087). Correspondence concerning this article should be addressed to Daryl B. O’Connor, Institute of Psychological Sciences, University of Leeds, Leeds, England LS2 9JT. E-mail: d.b.o’connor@leeds.ac.uk Nature of Changes in Eating Behavior Our first aim was an exploration of the nature of changes in eating behavior associated with stress. The impact of stress may be S20 STRESS AND EATING on the types of food selected as well as on the amount of food consumed. For example, Grunberg and Straub (1992) demonstrated that when stressed, women were more likely to select foods high in calories (and fat) and Oliver, Wardle, and Gibson (2000) found changes in consumption of sweet high-fat foods and more energy dense foods. In addition, Steptoe, Lipsey, and Wardle (1998) demonstrated that ‘fast food’ was eaten more frequently when respondents reported experiencing greater number of hassles. Comparable results have been shown by Oliver and Wardle (1997) in a study of the perceived effects of stress on food choice. Taken together, these results suggest that individuals, when stressed, may shift their preference to more palatable and energy dense snack foods, which are less healthy and higher in fat, thus potentially increasing their risk of CVD and cancer. Therefore, in order to examine whether daily stressors can modify the intake of specific types of foods, we elected to utilize changes in betweenmeal snacks and fruit and vegetable consumption as our primary outcome measures. In addition, we also assessed perceptions of changes in daily food intake (e.g., main meals) through ratings designed to tap perceived increases and decreases in food intake. Types of Stress Affecting Changes in Eating The second aim of this study was to investigate the types of stress affecting changes in eating. Previous research has examined either the ‘general effects hypothesis’ that stress changes consumption of food generally or the ‘individual differences hypothesis’ that stress leads to changes in eating in particular groups (e.g. the obese, restrained eaters, women; Greeno & Wing, 1994). However, less is known about the importance of the type of stress experienced and whether some stressors produce differential effects on eating behavior (Conner & Armitage, 2002; Wallis & Hetherington, 2004). Several researchers have found stressors of an ego-threatening nature (e.g. where there is a fear of failure) to have distinct effects from those that elicit physical threat (e.g. fear of an electric shock). Heatherton and colleagues (1991; 1992) suggest that situations involving potential negative evaluation or task failure (ego-threats) will lead to disinhibition (over-eating) in restrained eaters (i.e., those attempting to control their food intake) or current dieters whereas physically threatening situations will not. In fact, Heatherton et al. (1991) found fear of an electric shock had the capacity to elicit a hypophagic response (reduced intake) in unrestrained participants. In addition, the impact of ‘interpersonal stress’ has received scientific attention (e.g. Oliver et al., 2001; Tanofsky-Kraff et al., 2000). Tanofsky-Kraff and colleagues (2000) compared the effects of interpersonal stress (i.e. stress associated with feeling a sense of social alienation and being interpersonally ill-equipped) with egoand physically threatening stress on eating and found that in the interpersonal stress group, individuals with high levels of restraint consumed the most food. More recently, laboratory work by Wallis and Hetherington (2004) and Lattimore and Caswell (2004), has further underlined the importance of considering different types of stressors (e.g., cognitively demanding stressors, active and passive coping tasks) in conjunction with eating style variables (e.g., restraint, emotional eating). Other types of stressors, such as work-related stress, have received less attention and therefore require further investigation. For example, Wardle et al. (2000) found increased energy, fat and sugar intake in workers S21 during periods of high compared to low work stress (as indexed by hours worked), however, little or no research has examined whether work-related stress has a similar impact to work hours on eating behavior. Thus here, for the first time, we investigated simultaneously the impact of ego-threatening, interpersonal, physical and work-related stressors on eating behavior. Moderators of the Hassles–Snacking Relationship The third main aim of the study was to investigate the influence of individual differences variables on the hassles-eating relationship. Individual differences models of stress predict that groups reflecting different levels of vulnerability will vary in their eating behavior when stressed (e.g., the obese and non-obese; restrained and unrestrained; women and men). Although early research focused on obesity as the key variable moderating the stress-eating relationships, Greeno and Wing’s (1994) review found little evidence to support the predicted effects. Indeed studies such as Baucom and Aiken (1981) demonstrated that it was dieting rather than obesity that was the key moderator of stress-related eating in their study. For both obese and non-obese groups, stress only produced increases in eating in the dieting group. In contrast, the evidence supporting the moderating effect of restraint is stronger: stress generally produces greater increases in eating in restrained compared to unrestrained eaters (e.g., Greeno & Wing, 1994; Wallis & Hetherington, 2004; Wardle et al., 2000). Nevertheless it must be conceded that the vast majority of studies to date have only demonstrated this effect in college-aged women and often using laboratory-based or cross-sectional designs. The effect of restraint in men is an issue that is particularly worthy of further study and given the inconsistent evidence, it still remains an important issue whether being obese acts as a moderator of the stress-eating relationship. Similarly, ‘emotional’ and ‘external’ eating styles have also been identified as potentially important moderating variables within the stress-eating literature. The former refers to a tendency to eat more when anxious or emotionally aroused compared to non-emotional eaters who do not show such reactivity to emotion in their eating habits. In the laboratory, Oliver et al. (2000) found that emotional eaters (and not restrained eaters) consumed significantly more sweet high-fat foods and more energy dense foods in response to stress. Other studies have reported a limited (Schlundt et al. 1991) or lack of impact (e.g. Conner et al., 1999; O’Connor & O’Connor, 2004) of emotional eating on stress-eating relationships. External eating encapsulates the idea of eating being prompted by external (e.g. smelling appealing food) as opposed to internal (e.g. feeling hungry) stimuli. It has been suggested that stress might draw attention to external cues (e.g. Heatherton & Baumeister, 1991) and act as a stimulus to eating among those high in external eating. Conner et al. (1999) in a study examining multiple moderators of stress-eating relationships reported externality to be the key variable in moderating the relationship between daily hassles and number of snacks consumed (reducing other moderators to non-significance). Given the limited number of published studies with emotional and external eating, their impact on stress-eating relationships remain an important issue for further study. O’CONNOR, JONES, CONNER, MCMILLAN, AND FERGUSON S22 Two important additional potential moderators of the stresseating relationship are also examined in this research: disinhibition and gender. Disinhibition is one of the three-subscales measured by the Three Factor Eating Questionnaire (Stunkard & Messick, 1985) that assesses the tendency to overeat. Recent debate in the literature has highlighted the importance of considering disinhibition, as this may be a better predictor of food consumption than dietary restraint (see Ouwens, van Strien, & van der Staak, 2003). Moreover, comparatively little research has investigated the moderating role of disinhibition on the stress-eating relationship. Finally, gender remains an important potential moderator, with the majority of studies only examining or reporting effects for women (although see Stone & Brownell, 1994). A notable shortcoming of the literature on the individual differences approach to understanding stress-eating relationships is the general failure to examine potential moderating variables simultaneously. Most studies are restricted to exploring the influence of one or two individual differences variables, therefore, not allowing conclusions to be drawn relating to the relative importance of different variables. Moreover, multilevel modeling approaches, using diary methods, allow the modeling of day-to-day withinperson effects together with the impact of between-person factors. Therefore, in this study, we used a multilevel diary design to examine day-to-day within-person effects of daily hassles on eating behavior over 28 days, together with the influence of potential moderating variables considered individually and simultaneously (i.e., restraint, emotional eating, external eating, disinhibition, gender and obese status). In summary, this study had three main aims: 1) to explore the nature of changes in eating behavior associated with daily hassles, particularly relating to between-meal snacking; 2) to investigate the types of hassles associated with between-meal snacking, and 3) to examine the individual and simultaneous impact of moderating variables on relations between daily hassles and between-meal snacking. Method Participants Participants were recruited from a large local government organization. Adverts were placed across the organization, requesting participation and providing basic details and contact information. Six hundred and forty employees replied to this advert requesting additional information. A total of 466 individuals (215 male; 251 female) aged 18-65 years took part (73% uptake rate) in the first round of the study, which consisted of a one-week daily diary and completion of an initial demographics questionnaire (e.g., age, height, weight). A total of 449 of these individuals (207 male; 242 female) completed the second one-week daily diary, 443 (202 male; 241 female) completed the third one-week daily diary, 437 (200 male; 237 female) completed the fourth one-week daily diary and 428 (197 male; 231 female) completed a final questionnaire. Complete data across all the diaries and final questionnaire were available from 422 participants (193 male; 229 female; 65.9% of those recruited). There were no differences on the main study variables between the completers and non-completers. Therefore, all subsequent analyses focus on the sample of 422. The mean age of the final sample was 40.32 years (range ⫽ 18-65 years). Fifty-five participants (26 men; 29 women) had a BMI ⱖ 30 and were classified as obese. Participants were reimbursed approximately $75 for taking part in the study. Financial reimbursement of this nature may foster participant motivation in the study and it has been argued that increased motivation leads to increased compliance with the study protocol (Green, Rafaeli, Bolger, Shourt, & Reis, 2006). Design Participants completed an initial questionnaire, followed by four 7-day daily diaries and a final questionnaire at the end of the study. After initial recruitment and agreement to participate, participants were individually contacted by a researcher who explained the questionnaires and diary protocol and answered any questions or issues. The researcher made regular contacts with participants who were free to speak to the researcher (via phone or email) throughout the project. Diaries were returned at the end of each week and checked against postmarks in order to ensure the participants complied with the study protocol and completed their diary during the appropriate week. As a result no diaries were excluded. An interval-contingent method was employed, where the participants completed their diary at the end of each day for a period of four weeks, with the diary returned by post to the researchers at the end of each week. The use of weekly diaries to collect end of day data – especially over an extended period of time – is a well established protocol in the health, clinical and social psychology literature (e.g., Bolger DeLongis, Kessler, & Schilling, 1989; Feldman, Downey, & Schaffer-Neitz, 1999; Ferguson, Cassaday, & Bibby, 2003; Green et al., 2006) that has been shown to produce reliable data (Green et al., 2006). End-of-day diaries, rather than event contingent diaries were utilized to increase motivation and compliance with the diary protocol. This procedure reduces participant burden when the time span of the study is long (lasting weeks or months) (see Tennen et al., 2006) and it has been argued that reduced burden increases participant compliance (Green et al., 2006). Measures Participants completed a daily diary form at the end of each day and a series of questionnaires at the beginning and the end of the study. Daily diary. In each 7-day daily diary, participants were requested – using free responses – to list each food eaten betweenmeals and to report each stressor or hassle experienced and then to rate its intensity on a scale extending from ‘not stressful’ (0) to ‘very stressful’ (4).1 Daily hassles were defined and examples provided in each diary booklet. These procedures were used by Conner et al. (1999) and are similar to procedures used by Ferguson et al. (2003). Perceived daily variations in diet were assessed by asking participants to indicate if they ate more / less or the same amount of main meals than they usually do on a scale extending ⫺2 (ate much less than usual), ⫺1 (ate less than usual), 0 (ate usual amount), 1 (ate more than usual), 2 (ate much more than 1 The results for ‘actual number’ and the ‘total intensity’ of hassles were substantively similar, therefore, only the results for the actual number of hassles are reported. STRESS AND EATING usual). The number of fruit and vegetables eaten each day were also recorded (using guidance from UK Department of Health’s Five-A-Day campaign). Each between-meal snack reported in the diaries was categorized as being high in fat and/or high in sugar based upon validated food composition tables (McCance & Widdowson, 1991). The coding was not mutually exclusive and each snack could potentially have been classified into more than one category. All coding was conducted by a panel of four health psychologists, trained to PhD level with a mean of 13 years research experience, resulting in good inter-rater reliability with all Kappas above 0.70 (Altman, 1991). Every snack reported was coded as whether it was high in fat, high in sugar or whether it was a fruit or vegetable. Similarly, every hassle reported was coded as whether or not it was ego-threatening (e.g., job interview, public talk, criticism), interpersonal (e.g., argument with partner, family problems, visiting relatives), work-related (e.g., difficult work task, late for meeting, deadline), or physical in nature (e.g., anxious/frightened, feeling ill, threat of attack by dog) based upon the existing research literature (e.g., Heatherton et al., 1991; 1992; Steptoe et al., 1998; Tanofsky et al., 2000). Again, the coding was not mutually exclusive and each hassle could potentially have been classified into more than one category. Disagreements were resolved through discussion. Initial questionnaire. Participants were asked to complete an initial demographics questionnaire (age, gender, height, weight and socioeconomic status). Socioeconomic status was measured using the National Statistics Socio-Economic Classification questionnaire (NS-SEC; National Statistics, 2002). This is an occupationally based classification used for all official statistics and surveys in the UK. This categorizes respondents into one of the following five classes (1 ⫽ managerial and professional classes, 2 ⫽ intermediate occupations, 3 ⫽ small employers and own account workers, 4 ⫽ lower supervisory and technical occupations, 5 ⫽ semi-routine and routine occupations). Final questionnaire. Eating styles (Restrained, Emotional and External eating) were measured using the Dutch Eating Behavior Questionnaire (DEBQ; Van Strien, Frijters, Bergers, & Defares, 1986) and Disinhibition was assessed using the scale from the Three Factor Eating Questionnaire (TFEQ; Stunkard & Messick, 1984). The DEBQ and TFEQ have been found to have good construct (and factorial) validity and internal reliability (van Strien et al., 1986; Stunkard & Messick, 1984; Wardle, 1987). All subscales exhibited excellent internal consistency in this sample (alphas in range 0.81 – 0.88). Data Analysis The daily data were analyzed using Hierarchical Multivariate Linear Modeling (HMLM) using HLM6 (Raudenbush, Bryk, & Congdon, 2004). HMLM allows alternative models of temporal variance-covariance structure to be assessed and is therefore a suitable modeling strategy for a fixed occasion design (Snijders & Boker, 1999; Raudenbush et al., 2004). For all the models tested here, the assumption of an unrestricted Level 1 variancecovariance structure fitted the data best.2 The data contained a two level hierarchical structure, Level 1 being the within-person variation (e.g., daily patterns in the number and type of snacks consumed, the number and type of hassles experienced), and Level 2 being the between-person variability (e.g., eating style, gender). The Level 1 predictor variables were S23 not centered as they were dichotomized due to skewness (range 1.78 – 7.21) and kurtosis (range 4.57–15.08) that were not correctable using transformations. Therefore, a score of zero represented not experiencing hassles on a given day, and ‘1’ represented experiencing hassles. The Level 2 variables were centered around the grand mean (cf. Nezleck, 2001; Raudenbush & Bryk, 2002). We provide estimates of standardized coefficients (calculated using the procedure outlined by Hox, 2002) along with unstandardized coefficients, as an index of effect sizes. Measures of R2 are not included as Kreft and De Leeuw (2006) strongly caution against calculating R2 values for models other than random intercept models, as (1) the Level-1 and Level-2 estimates are confounded and separate estimates of variance are difficult to define, (2) there is more than one way to estimate between variance, and (3) R2 can have negative values. In conclusion they argue ‘. . . not setting too much store by the calculation on an RB2 and RW2. Both concepts are ill defined and ambiguous, while their usefulness is limited to a random intercept model.’ (p. 119). Participants who had missed an entire week or more or did not return the final questionnaire were excluded from the analyses. Days within diaries that had missing data were also removed from the dataset. The 422 participants who provided usable data sets provided 11,444 days of data (372 days with missing data removed).3 Results Descriptive Statistics Table 1 shows descriptive statistics for main Level 1 and Level 2 study variables including daily measures of hassles and eating behavior. 2 The most appropriate multilevel model for data with a repeated measure (time series) nested within an individual is to use Hierarchical Multivariate Linear Models (HMLM) (Bryk & Raudenbush, 2005; Snidjers & Bosker, 1999). This allows for various variance-covariance structures to be examined. Following arguments presented by Raudenbush and Bryk (2002), the best fitting variance-covariance structure was based on the structure with the lowest Deviance and by comparing the pattern with the variance-covariance structures for a 1st order autoregressive and homogeneous Level-1 structure with those observed in the unrestricted model. A more parsimonious variance-covariance structure was chosen only if its Deviance was significantly smaller than that for the unrestricted model and the variance-covariance structure was similar to that observed in the unrestricted model. Based on the difference between the Deviance statistic, a model based on an unrestricted Level 1 variance/co-variance was the best fit to these data. A consequence of this is that the model does not explicitly specify random variation (error terms) at level-2 (in terms of individual error terms) as these are absorbed into the ⌬ matrix. The ⌬ matrix captures the variation and co-variation among the T measurement periods. As such error is modeled but not expressed in the individual equations. 3 While calculating statistical power for HMLM is complex, a number of heuristics – based on simulation studies – have been suggested (see Hoffman, 1997). These rules of thumb generally focus on the trade off between the number of Level 1 and Level 2 responses. To detect cross-level interactions the general rule is that as the number of Level 2 units increases the number of Level 1 assessments required reduces. It has been suggested that with 150 Level 2 units then 5 Level 1 assessments are required to achieve a power of. 90 to detect a cross-level interaction (Hoffman, 1997). Based on this heuristic the data analyzed in this study should be considered to have adequate power (Level 2 ⫽ 422 units, Level 1 ⫽ 11,444 assessments, with each participant providing on average 27 days of data each). O’CONNOR, JONES, CONNER, MCMILLAN, AND FERGUSON S24 Table 1 Descriptive Statistics for the Daily (Level 1) and BetweenPerson (Level 2) Measures Across 28 Days Level and variable Level 1 Snacks per day Hassles per day High-fat snacks per day High-sugar snacks per day Ego-threatening hassles per day Interpersonal hassles per day Work-related hassles per day Physical hassles per day Ate less/usual/more main meals per day Ate less/usual/more snacks per day Average daily fruit & vegetable consumption Level 2 Age Body mass index Socioeconomic classification Restrained eating Emotional eating External eating Disinhibition M SD 3.79 0.97 1.12 1.50 0.07 0.26 0.19 0.11 0.02 0.00 3.29 1.81 0.80 1.19 1.45 0.25 0.44 0.39 0.31 0.17 0.31 1.65 40.32 25.61 1.68 2.60 2.41 2.97 6.07 10.73 3.99 1.17 0.88 0.91 0.62 4.05 Initial Level 1 Models Effects of daily hassles on eating behavior. Initial Level 1 models were examined to investigate the effect of experiencing one or more hassles on each of the eating behavior outcome measures (i.e., perceptions of amount eaten during main meals, actual reported number of between-meal snacks, high fat snacks, high sugar snacks, fruit and vegetables consumed) over the 28 days study period (Table 2). The general form of each daily hassles-eating behavior model is expressed by the following equations: Level 1: yij ⫽ ␤0j ⫹ ␤1 (daily hassles) ⫹ rij Level 2: ␤0 ⫽ ␥00 ␤ 1 ⫽ ␥10 where yij ⫽ the within-person variation in daily measure of eating behavior across I days for person j, ␤0j ⫽ the intercept term, ␤1 ⫽ the slope estimates for daily hassles, and rij represents the error associated with each daily measure; ␥ ⫽ the structural coefficients expressing the Level 1 variation. The results for each model are presented in Table 2. The findings for the initial models showed that each of the measures of eating behavior were significantly different from zero (␥00). The results also showed significant positive associations between daily hassles experienced and many of the eating behaviors measures (␥10). That is, those who experienced one or more hassles reported consuming significantly more between-meal snacks, high fat snacks and high sugar snacks. Conversely, there was a significant negative association between experiencing one or more daily hassles and perceived level of main meals consumed and number of vegetables eaten indicating that those who experienced one or more hassles reported consuming less main meals and portions of vegetables (see Table 2). These results were generally unchanged when repeated for men and women separately. However, we found that the relationship between daily hassles and high fat (coefficient ⫽.136, p ⬍ .001) and high sugar (coefficient ⫽.159, p ⬍ .001) snack consumption was only significant in the female sample (with total snacks being significant for both men and women). Effects of types of hassles on between-meal snacking behavior and perceptions of main meal consumption. Based upon the existing research literature (cf., Heatherton et al., 1991; 1992 Steptoe et al., 1998; Tanofsky et al., 2000), daily hassles were categorized into ego-threatening, interpersonal, work-related and physical stressors (as outlined in the Method section) in order to allow us to investigate whether particular types of stressors were associated with the consumption of between-meal snacks and main meals. The types of daily stressor were modeled separately because every observation potentially was not independent (e.g. a work-related stressor may also have been classified as egothreatening etc.). The results shown in Table 3 indicated that each of the types of stressors were significantly positively associated with between-meal snacks, except physical stressors, which were negatively associated with snacking (Table 3). These results are important as they indicate that ego-threatening, interpersonal and work-related hassles were significantly associated with increased snacking (a hyperphagic response), whereas, physical stressors were significantly associated with decreased consumption of snacks (a hypophagic response) in naturalistic settings. In addition, these results identified work-related hassles as the type of daily hassles that exerted the strongest effects on the between-meal snacking response. For perceptions of main meals consumed, experiencing one or more daily interpersonal, work-related or physical hassle was significantly associated with eating less than usual at main meals. Ego-threatening hassles were not significantly associated with perception of main meal consumption. Moderators of the hassles-between meals snacking relationship. In order to examine moderators of the stress-snacking relationship, we modeled the day-to-day within-person effects of daily hassles on between-meal snacking (Level 1 variables), together with the impact of between person variations in restraint, emotional eating, external eating, disinhibition, gender and obesity status (BMI ⬎ 30) (Level 2 variables). The first step of the analysis modeled the impact of each moderating variable independently. The second step of the analysis modeled the simultaneous impact of the moderating variables on the hassles-snacking relationship – a central aim of this research. Only the statistically significant moderators from step one were entered into the second step of the analysis. The general form for the equations for the (conditional) step 1 models are as follows: Level 1: yij ⫽ ␤0j ⫹ ␤1 (daily hassles) ⫹ rij Level 2: ␤0 ⫽ ␥00 ⫹␥01 (restraint) ␤ 1 ⫽ ␥10 ⫹␥11 (restraint) This model includes cross-level terms. In this case for restraint, ␥00 indicates (as in the initial model) the mean level of between-meal snacks, and ␥01 (restraint) indicates, the extent to which this average is influenced by level of dietary restraint. Similarly, ␥10 indicates the average size of the relationship between daily hassles STRESS AND EATING S25 Table 2 Within-Person Associations of Daily Hassles Experienced, Daily Perceptions of Main Meals Consumed, and Actual Reported Daily Between-Meals Snacks, Fruit, and Vegetable Intake MRCM effect ␥ Intercept: perception of main meals Level 1 slope: daily hassles—main meals Intercept: actual reported between-meal snacks Level 1 slope: daily hassles—snacks Intercept: high-fat snacks Level 1 slope: daily hassles—high-fat snacks Intercept: high-sugar snacks Level 1 slope: daily hassles—high-sugar snacks Intercept: portions of fruit Level 1 slope: daily hassles—fruit Intercept: portions of vegetables Level 1 slope: daily hassles—vegetables ␥00 ␥10 ␥00 ␥10 ␥00 ␥10 ␥00 ␥10 ␥00 ␥10 ␥00 ␥10 B 0.037 ⫺0.041 3.788 0.201 1.045 0.101 1.407 0.098 1.435 0.023 1.830 ⫺0.093 SE 0.009 0.007 0.084 0.028 0.037 0.020 0.048 0.022 0.048 0.020 0.046 0.022 ␤ ⫺0.050 0.043 0.042 0.033 0.008 ⫺0.032 p ⬍.001 ⬍.001 ⬍.001 ⬍.001 ⬍.001 ⬍.001 ⬍.001 ⬍.001 ⬍.001 .237 ⬍.001 ⬍.001 Note. Level 1 n ⫽ 11,444. MRCM ⫽ multilevel random coefficient model; ␥ ⫽ hierarchical multivariate linear modeling symbol; B ⫽ unstandardized coefficients; SE ⫽ standard error; ␤ ⫽ standardized coefficients. and between-meal snacking, and ␥11 (restraint), the extent to which that relationship is moderated by (or conditional on) level of dietary restraint. The models for emotional eating, external eating, disinhibition, gender and obesity status (BMI ⬎ 30) were identical to that for restraint. The results shown in Table 4 indicate that the significant positive associations between daily hassles and between-meal snacking remained in each case (␥10). They also showed that none of the individual differences variables directly influenced the average level of between-meal snacks consumed across the 28 days (␥01). Moreover, consistent with our predictions, each of the individual differences variables showed significant cross level interactions with the slope coefficients for the daily hassles – snacking relationships (␥11). The findings indicate that individuals who were high on restraint, emotional eating, external eating, disinhibition or who were obese or female showed significantly stronger positive associations between daily hassles and snacking. That is, individuals who have higher levels of vulnerability were significantly more likely to consume more snacks in response to daily stressors (and not because they were in a vulnerable group per se). For example, in the case of restrained eaters, the experience of daily hassles led to disinhibition (ie., breaking of their restraint). Next we modeled the impact of multiple moderators on the hassles-snacking relationship, simultaneously (step two). The equation for this (conditional) model is as follows: Level 1: yij ⫽ ␤0j ⫹ ␤1 (daily hassles) ⫹ rij Level 2: ␤0 ⫽ ␥00 ⫹␥01 (gender) ⫹ ␥02 (obesity) ⫹ ␥ 03 共restraint兲 ⫹ ␥ 04 共emotional 兲 ⫹ ␥ 05 共external 兲 ⫹ ␥ 06 共disinhibition兲 Table 3 Within-Person Associations of Different Types of Hassles Experienced, Actual Reported Daily Between-Meal Snack Intake, and Perceptions of Main Meals Consumed MRCM effect ␥ B SE Intercept: total between-meal snacks Level 1 slope: ego-threatening hassles—snacks Intercept: Total between-meal snacks Level 1 slope: interpersonal hassles—snacks Intercept: total between-meal snacks Level 1 slope: work-related hassles—snacks Intercept: total between-meal snacks Level 1 slope: physical threat hassles—snacks Intercept: perception of main meals Level 1 slope: ego-threatening hassles—meals Intercept: perception of main meals Level 1 slope: interpersonal hassles—meals Intercept: perception of main meals Level 1 slope: work-related hassles—meals Intercept: perception of main meals Level 1 slope: physical threat hassles—meals ␥00 ␥10 ␥00 ␥10 ␥00 ␥10 ␥00 ␥10 ␥00 ␥10 ␥00 ␥10 ␥00 ␥10 ␥00 ␥10 3.886 0.190 3.871 0.166 3.792 0.341 3.921 ⫺0.199 0.013 ⫺0.016 0.017 ⫺0.017 0.022 ⫺0.031 0.017 ⫺0.111 0.082 0.041 0.082 0.031 0.082 0.029 0.082 0.066 0.008 0.011 0.008 0.008 0.008 0.007 0.007 0.017 Note. MRCM ⫽ multilevel random coefficient model; ␥ ⫽ hierarchical multivariate linear modeling symbol. ␤ 0.026 0.031 0.068 ⫺0.017 ⫺0.012 ⫺0.018 ⫺0.034 ⫺0.056 p ⬍.001 ⬍.001 ⬍.001 ⬍.001 ⬍.001 ⬍.001 ⬍.001 ⬍.01 .066 .130 .034 .030 .006 ⬍.001 .030 ⬍.001 O’CONNOR, JONES, CONNER, MCMILLAN, AND FERGUSON S26 Table 4 Individual Moderators of the Within-Person Effects of Daily Hassles on Actual Reported Between-Meal Snacking MRCM effect Intercept: between-meal snacks Level 1 slope: daily hassles—snacks Cross-level interaction with restraint Restraint—snacks Restraint ⫻ Daily Hassles—snacks Intercept: Between-meal snacks Level 1 slope: Daily hassles—snacks Cross-level interaction with emotional eating Emotional eating—snacks Emotional Eating Daily Hassles—snacks Intercept: Between-meal snacks Level 1 slope: daily hassles—snacks Cross-level interaction with external eating External eating—snacks External Eating ⫻ Daily Hassles—snacks Intercept: between-meal snacks Level 1 slope: daily hassles—snacks Cross-level interaction with disinhibition Disinhibition—snacks Disinhibition ⫻ Daily Hassles—snacks Intercept: between-meal snacks Level 1 slope: daily hassles—snacks Cross-level interaction with gender Gender—snacks Gender ⫻ Daily Hassles—snacks Intercept: between-meal snacks Level 1 slope: daily hassles—snacks Cross-level interaction with obesity status Obesity status—snacks Obesity Status ⫻ Daily Hassles—snacks ␥ B SE ␤ p ␥00 ␥10 3.785 0.202 .084 .028 ␥01 ␥11 ␥00 ␥10 ⫺0.019 0.082 3.782 0.203 .085 .028 .084 .028 ⫺0.008 0.021 ␥01 ␥11 ␥00 ␥10 0.075 0.148 3.789 0.199 .084 .029 .084 .028 0.032 0.038 ␥01 ␥11 ␥00 ␥10 0.045 0.081 3.790 0.204 .084 .029 .084 .028 0.019 0.021 ␥01 ␥11 ␥00 ␥10 ⫺0.059 0.135 3.788 0.200 .086 .029 .083 .028 ⫺0.025 0.035 ␥01 ␥11 ␥00 ␥10 0.261 0.139 3.791 0.202 .167 .057 .083 .028 0.057 0.071 0.043 .118 ⬍.05 ⬍.001 ⬍.001 ␥01 ␥11 ⫺0.167 0.257 .247 .085 0.030 0.200 .499 ⬍.01 0.043 0.043 0.043 0.043 0.043 ⬍.001 ⬍.001 .820 ⬍.01 ⬍.001 ⬍.001 .373 ⬍.001 ⬍.001 ⬍.001 .592 ⬍.01 ⬍.001 ⬍.001 .491 ⬍.001 ⬍.001 ⬍.001 Note. MRCM ⫽ multilevel random coefficient model; ␥ ⫽ hierarchical multivariate linear modeling symbol; B ⫽ unstandardized coefficients; SE ⫽ standard error; ␤ ⫽ standardized coefficients. ␤ 1 ⫽ ␥ 10 ⫹ ␥ 11 共 gender兲 ⫹ ␥ 12 共obesity兲 ⫹ ␥ 13 共restraint兲 ⫹ ␥ 14 共emotional 兲 ⫹ ␥ 15 共external 兲 ⫹ ␥ 16 共disinhibition兲 In this model, ␥00 indicates (as in the initial model) the mean level of between-meal snacks, and ␥01 - ␥ 06 (individual differences variables) indicate, the extent to which this average is impacted by each individual differences variable. Similarly, ␥10 indicates the average size of the relationship between daily hassles and between-meal snacking, and ␥11 - ␥16 (individual differences variables), the extent to which the relationships are moderated by (or conditional) each individual differences variable. The resultant model (Table 5) again showed that none of the individual differences variables, when entered simultaneously, directly influenced the average level of between-meal snacks consumed (␥01- ␥06). However, when modeled together, the cross level interactive effect of emotional eating (␥14) on the daily hassles – snacking relationship (␥11- ␥16) was the only effect to remain statistically significant (Table 5). This is an important and noteworthy finding because it indicates, for the first time, that emotional eating is the main moderator of the relationship between hassles and eating. In order to explore this finding further, we conducted the same analyses for high fat snacks, high sugar snacks and fruit and vegetable portions. For snacks high in fat and sugar, the result was replicated; such that emotional eating was the only individual differences variable to emerge to significantly moderate the daily hassles-high fat/sugar snacking relationship when all variables were considered simultaneously (Table 6). No moderating effects were found for fruit and vegetable intake in response to stress. Moderators of the hassles-perceptions of main meals consumed relationship. The moderation analysis was repeated for the hassles-perceptions of main meals consumed relationship. The resultant models showed that restraint, emotional eating, external eating, disinhibition, gender and obesity status each failed to moderate the hassles-perception of main meals consumed relationship (data not shown). Given the absence of effects for individual moderator variables, no further analysis for perceptions of main meals consumed was conducted. Effects of individual and multiple moderators on the impact of different types of hassles. Finally, we examined the effects of individual and multiple moderators on the impact of different types of daily hassles on between-meal snacking. Firstly, we modeled the cross-level effects of the separate individual moderators (e.g., restraint, emotional eating etc.) on each type of hassle–snacking STRESS AND EATING S27 Table 5 Effects of Moderators Entered Simultaneously on the Within-Person Effects of Daily Hassles on Actual Reported Between-Meal Snacking MRCM effect ␥ SE ␤ Intercept: between-meal snacks Level 1 slope: daily hassles—snacks Level 2 effects Gender—snacks Obesity—snacks Restraint—snacks Emotional eating—snacks External eating—snacks Disinhibition—snacks Cross-level interactions Gender ⫻ Daily Hassles—snacks Obesity ⫻ Daily Hassles—snacks Restraint ⫻ Daily Hassles—snacks Emotional Eating ⫻ Daily Hassles—snacks External Eating ⫻ Daily Hassles—snacks Disinhibition ⫻ Daily Hassles—snacks ␥00 ␥10 3.785 0.206 .083 .028 0.044 ⬍.001 ⬍.001 ␥01 ␥02 ␥03 ␥04 ␥05 ␥06 0.284 ⫺0.125 ⫺0.061 0.142 0.057 ⫺0.191 .187 .255 .096 .124 .109 .121 0.062 ⫺0.019 0.026 0.061 0.025 ⫺0.082 .130 .623 .525 .254 .604 .114 ␥11 ␥12 ␥13 ␥14 ␥15 ␥16 0.028 0.153 0.015 0.099 ⫺0.026 0.067 .064 .089 .033 .044 .038 .041 0.118 0.123 0.004 0.039 ⫺0.009 0.024 .659 .084 .648 ⬍.05 .496 .106 B p Note. MRCM ⫽ multilevel random coefficient model; ␥ ⫽ hierarchical multivariate linear modeling symbol; B ⫽ unstandardized coefficients; SE ⫽ standard error; ␤ ⫽ standardized coefficients. relationship (physical-snacking, ego-threatening-snacking relationship etc.). Secondly, we modeled the simultaneous impact of the moderating variables on the hassles-snacking relationship, in order to identify the key moderator(s). Only the statistically significant moderators were entered into the second step of the analysis. The findings are presented in Table 7 and for the sake of brevity only significance levels ( p values) are reported for each cross level model indicating whether the Level 1 hassles-snacking relationship is significantly influenced by each moderating variable. The results showed that emotional eating significantly moderated all types of hassle-snacking relations and was the only variable to influence the physical hassles – snacking relationship. When the moderating variables were considered simultaneously (final column), emotional eating (again) emerged as the most important vulnerability variable in three out of four models (i.e., not for the interpersonal hassles-snacking relationship). Taken together, these findings suggest that emotional eating style is the pre-eminent individual differences variable in understanding the impact of stress on eating behavior. Discussion The current research aimed to explore the nature of changes in eating behavior associated with daily hassles, particularly relating to between-meal snacking and to investigate the impact of different types of hassles on eating behavior outcomes. In addition, the study aimed to examine the individual and simultaneous impact of moderating variables on daily hassles-between-meal snacking relations. These are discussed in turn. We found strong evidence that daily hassles were associated with increased consumption of high fat and high sugar betweenmeal snack foods and also with a perceived reduction in main meals and vegetable consumption. These results are concerning as an overwhelming body of evidence has shown the importance of maintaining a balanced diet, including eating a low fat diet and five portions of fruit and vegetables a day, in terms of reducing risk of cardiovascular disease and cancer risk (e.g., Heimendinger, Van Ryn, Chapelsky, Forester, & Stables, 1996; Kumanyika et al., 2000; Van Horn & Kavey, 1997; Wong & Lam, 1999). Therefore, the changes observed in the current study may indicate a serious indirect pathway through which stress influences health risk. Moreover, these findings represent an important step towards improving our understanding of stress-induced changes in eating behavior in a number of ways. They demonstrate, for the first time using a naturalistic daily diary design, that hassles have the capacity to disrupt the normal patterning of food intake by causing hyperphagic and hypophagic eating responses in men and women. Previous work in real-world studies of stress and eating have tended to concentrate on one aspect of eating behavior (e.g., snacking; Conner et al., 1999; O’Connor & O’Connor, 2004) or has been restricted to laboratory investigations of food consumption in artificial environments (e.g., ice-cream; Heatherton et al., 1991). More generally, many of the studies have also reported inconsistent and conflicting findings (e.g., Bellisle et al., 1992; Grunberg & Straub, 1992; Wallis & Hetherington, 2004; Wardle et al., 2000). Our results highlight the importance of considering different aspects of eating behavior and suggest that the disparate findings in the existing literature may be accounted for by the variations in the outcome measures employed. The current research has also advanced our understanding of the moderating effects of different types of daily hassles on the stresssnacking relationship. Ego-threatening, interpersonal and workrelated hassles were found to elicit a hyperphagic response, whereas, physical hassles were found to elicit a hypophagic response. The latter findings confirm the work by Heatherton et al. (1991) and demonstrate that the differential effects of stressors observed in the laboratory are generalizable to naturalistic settings. In addition, contrary to previous investigations, this research identified work-related hassles, and not ego-threatening or interpersonal, as the type of hassles that exerted the strongest effects on O’CONNOR, JONES, CONNER, MCMILLAN, AND FERGUSON S28 Table 6 Effects of Moderators Entered Simultaneously on the Within-Person Effects of Daily Hassles on Actual Reported Between-Meal Snacking and Fruit and Vegetable Consumption ␥ Moderator ␤ B SE 0.104 0.029 ⫺0.022 0.105 ⫺0.041 0.017 0.026 .019 .044 .022 .029 .025 .028 .044 0.009 ⫺0.008 0.053 ⫺0.018 0.008 0.020 ⬍.001 .503 .326 ⬍.001 .119 .605 .503 0.101 0.053 0.009 0.111 0.047 0.001 0.025 .021 .049 .025 .033 .029 .031 .067 0.036 0.003 0.052 ⫺0.019 0.0004 0.024 ⬍.001 .277 .697 ⬍.001 .101 .961 .708 0.024 ⫺0.043 0.016 0.016 ⫺0.040 0.024 0.021 .020 .045 .023 .030 .027 .029 .062 ⫺0.028 0.005 0.007 ⫺0.016 0.010 0.019 .238 .333 .489 .602 .126 .395 .736 .021 .048 .025 .032 .028 .031 .067 ⫺0.010 0.003 0.003 0.004 ⫺0.008 ⫺0.051 ⬍.001 .751 .708 .836 .725 .572 .439 p High-fat snacks Cross-level interaction Gender ⫻ Daily Hassles—snacks Restraint ⫻ Daily Hassles—snacks Emotional Eating ⫻ Daily Hassles—snacks External Eating ⫻ Daily Hassles—snacks Disinhibition ⫻ Daily Hassles—snacks Obesity Status ⫻ Daily Hassles—snacks ␥11 ␥12 ␥13 ␥14 ␥15 ␥16 High-sugar snacks Cross-level interaction Gender ⫻ Daily Hassles—snacks Restraint ⫻ Daily Hassles—snacks Emotional Eating ⫻ Daily Hassles—snacks External Eating ⫻ Daily Hassles—snacks Disinhibition ⫻ Daily Hassles—snacks Obesity Status ⫻ Daily Hassles—snacks ␥11 ␥12 ␥13 ␥14 ␥15 ␥16 Portions of fruit Cross-level interaction Gender ⫻ Daily Hassles—fruit Restraint ⫻ Daily Hassles—fruit Emotional Eating ⫻ Daily Hassles—fruit External eating ⫻ daily Hassles—fruit Disinhibition ⫻ Daily Hassles—fruit Obesity Status ⫻ Daily Hassles—fruit ␥11 ␥12 ␥13 ␥14 ␥15 ␥16 Portions of vegetables Cross-level interaction Gender ⫻ Daily Hassles—vegetables Restraint ⫻ Daily Hassles—vegetables Emotional Eating ⫻ Daily Hassles—vegetables External Eating ⫻ Daily Hassles—vegetables Disinhibition ⫻ Daily Hassles—vegetables Obesity Status ⫻ Daily Hassles—vegetables ⫺0.095 ⫺0.015 0.009 0.006 0.010 ⫺0.017 ⫺0.052 ␥11 ␥12 ␥13 ␥14 ␥15 ␥16 Note. ␥ ⫽ hierarchical multivariate linear modeling symbol; B ⫽ unstandardized coefficients; SE ⫽ standard error; ␤ ⫽ standardized coefficients. between-meal snacking (cf., Heatherton et al., 1991; 1992; Oliver et al., 2001; Tanofsky-Kraff et al., 2000). It is also particularly noteworthy in this current study, that these effects were not restricted to vulnerable individuals who were inhibiting their food intake per se (e.g., restrained eaters) as they have been in other studies (Heatherton et al., 1991; Wardle et al., 2000). Instead, these findings clearly demonstrate that daily hassles can directly influence snack food intake. Table 7 Individual and Combined Effects of Moderators on Different Types of Hassles Moderators Types of hassles Restraint Emotional eating External eating Disinhibition Gender Obese status Combined Physical Ego threatening Work related Interpersonal .145 .983 .435 .025 .031 .003 .003 .012 .252 .012 .192 .316 .141 .016 .028 .001 .092 .573 .199 .454 .823 .407 .048 .393 Emotion Emotion Emotion Disinhibition Note. Significance levels are shown for each Level 2 model indicating whether the Level 1 hassles-snacking relationship is significantly influenced by each moderating variable. STRESS AND EATING As predicted, each of the individual differences variables when considered separately influenced the overall hassles-snacking relationship. In other words, individuals who were high on restraint, emotional eating, external eating, disinhibition, or who were obese or female showed a significantly stronger positive association between daily hassles and between-meal snacking. These results provide strong support for the individual differences approach to stress-eating research and indicate that individuals do differ in their level of vulnerability to stress-induced eating (cf., Conner et al., 1999; Greeno & Wing, 1994; O’Connor & O’Connor, 2004). However future research ought to examine the impact of other important diathesis-stress mechanisms that might be relevant to eating behavior such as the influence of coping style and trait perfectionism (cf., O’Connor & O’Connor, 2003; O’Connor, O’Connor, & Marshall, 2007). These findings also represent the first study to replicate the moderating role of external eating and disinhibition identified in prior investigations (Conner et al., 1999). More importantly, the current research has also made a novel contribution by improving our understanding of the relative weight of different moderating variables. For the first time, when the impact of multiple moderators was considered simultaneously using multilevel random coefficient modeling, emotional eating reliably emerged as the pre-eminent moderating variable of the stress-between meal snacking relationship. Previous research has consistently identified dietary restraint or gender as the key variables in understanding stress-induced eating (e.g., Cools et al., 1992; Grunberg & Straub, 1992; Heatherton et al., 1991; Lattimore & Caswell, 2004; Tanofsky et al., 2000; Wallis & Hetherington, 2004; Wardle et al., 2000). However, in the main, these studies have not measured emotional eating (e.g., Heatherton et al., 1991), have studied females only (e.g., Wallis & Hetherington, 2004), and/or have not examined the relative importance of different individual differences variables (e.g., Stone & Brownell, 1994). Two recent laboratory studies that have considered the role of emotional eating alongside dietary restraint have found support for its central position in understanding stress-induced eating. Oliver et al. (2000) showed that stressed emotional eaters (and not restrained eaters) ate more sweet high fat foods and energy dense meals than unstressed and non-emotional eaters. Wallis and Hetherington (2004) found both variables were important in understanding hyperphagic responses to stress, although, these authors did not directly compare their relative importance. Several plausible psychological and biological mechanisms have been proposed that may explain why stress promotes increased between-meal snacking in those susceptible to emotional eating. Psychosomatic theory contends that the normal response to stress/arousal is loss of appetite (Bruch, 1973; Kaplan & Kaplan, 1957). However, some individuals have an inability to differentiate between hunger and other unpleasant sensations/feelings and as a result respond to stress by overeating. The theory also suggests that this response probably originates through early learning experiences. Neurohormonal routes have also been investigated. Markus et al. (1998) found that carbohydrate-rich, protein-poor food intake can prevent stress related deterioration of mood and cortisol elevations in stress-prone individuals via serotonergic mechanisms. These authors argued that carbohydrate-rich, protein-poor diets allow greater uptake in the brain of the precursor amino-acid tryptophan, thus facilitating reduced negative mood and better coping. Moreover, as suggested by Oliver et al. (2000), emotional S29 eaters may learn to “self-medicate” by shifting their preferences to low protein foods such as sweet and high fat snacks in order to experience reduced general dysphoria. Another relevant psychological theory suggests that overeating may be part of a motivated attempt to escape from self-awareness or negative self-referent appraisals (cf., Heatherton & Baumeister, 1991). Central to this strategy is the notion that particular individuals narrow their level of attention to the current and immediate stimulus environment (i.e., accessible food such as snacks) in order to shift attention away from negative self-appraisals and/or from feeling distressed. It is argued that such low levels of attention reduce self-awareness to a level where meaningful thought, evaluations of self and the implications of one’s activities are avoided. Therefore, in the context of the current study, emotional eaters may experience greater distress and negative self-appraisals when they encounter daily hassles and shift their attention towards high fat/sugar, energy dense snack foods in order to escape from this negative emotional state. Moreover, as outlined earlier, this shift towards snack-type foods may also be driven by the learned outcomes from dietary-induced changes in neurohormonal mechanisms. Future research ought to investigate further the effects of high fat and high sugar snack foods on the stress response within the context of the psychosomatic and escape from self theories. Cortisol reactivity to stress has recently been identified as a potential mechanism underpinning the stress-eating relationship and individual differences in emotional eating (Epel et al., 2001; Newman, O’Connor, & Conner, 2007). In the laboratory, Epel et al. (2001) found that participants who exhibited a high cortisol response to stress consumed more calories afterwards compared to those who exhibited a low cortisol response. In addition, using a naturalistic diary design, Newman et al. (2007) found a significant positive relationship between daily hassles and between-meal snacking within a sample of high cortisol reactors but not within a sample of low cortisol reactors. They also found that the relationships between the eating style variables (including emotional eating) and snack intake were significantly stronger in the high stress-induced cortisol reactivity group compared to the low reactivity group. These findings indicate that it is the stress-induced release of cortisol that may underpin the effects of stress on eating behavior. Finally, the present study is noteworthy because it is the first to examine the effects of daily stressors on eating behavior using multilevel modeling techniques and a daily diary design. Although the effects observed are relatively small, when interpreted within the context of well-established effect sizes reported in behavioral medicine, public health and general medicine, these results are comparable to those previously reported. Moreover, Rutledge and Loh (2004) argue that we should abandon ‘. . . simplistic effect size rules in favor of more thorough data interpretation efforts and closer consideration of real-world implications of . . . outcome variables.’ (pp. 143-144). In addition, these results suggest that the magnitude of the true effects of stress on eating behavior in naturalistic settings, when viewed using traditional statistical metrics (e.g., standardized coefficients), is lower than that observed in the laboratory. Therefore, future research ought to examine additional aspects of the stress process using multilevel approaches and naturalistic designs in order to establish whether the magnitude of these effects is generalizable to other health outcomes. A number of limitations of the current research require comment. We acknowledge that the current design did not allow us to S30 O’CONNOR, JONES, CONNER, MCMILLAN, AND FERGUSON precisely determine whether the experience of daily hassles occurred before or after the consumption of snacks. However, we believe that sufficient evidence exists to indicate for the correlations reported here, that the causal direction is likely to have been from experience of hassles to the consumption of snacks and not vice versa. First, there is a vast literature that has shown the causal ordering between stress and eating in animal and human studies, clearly demonstrating the effects of stress on eating behavior (e.g., see Greeno & Wing, 1994 for a review), Second, as outlined earlier, empirical data in the laboratory have clearly demonstrated that increases in snack consumption have been observed following exposure to psychosocial stress and not following exposure to neutral, non-stressful stimuli (Epel et al., 2001), Third, using a similar diary methodology, we have found previously that daily hassles-snacking relationships are only observed in individuals who exhibit significant increases in cortisol in response to stress indicating a clear causal pathway such that the experience of stress precedes the eating behavior (Newman et al., 2007). Finally, we are not aware of any research that suggests a plausible psychological or biological mechanism that indicates that snack consumption causes changes in the experience of daily hassles. Therefore, taken together, we feel that the only credible temporal sequence relates to hassles preceding changes in snack consumption. We also acknowledge that our measures of eating behavior are self-reported and that we did not utilize a detailed daily dietary assessment method such as ‘24 hour recall’ to examine changes in food intake. Given it was important to measure changes in daily hassles and eating behavior over a meaningful period of time (e.g., 28 days), it was decided that using such methods would be too burdensome for the participants and also that detailed daily assessment may influence participants’ normal eating behavior. Moreover, similar daily diary methods assessing discrete aspects of eating behavior (e.g., between-meal snacking) have been found to be reliable and valid measures of food intake (see Conner et al., 1999). Nonetheless, it would be fruitful if future research attempted to incorporate more objective measures of eating behavior and/or included careful assessment of portion size in order to allow reported food intake to be converted into caloric count. This would provide a better understanding of exactly how many more calories are consumed in response to each hassle. To conclude, the results of this research showed daily hassles were associated with a shift in preference towards high fat and high sugar between-meal snack foods and with a reduction in main meals and vegetable consumption. Ego-threatening, interpersonal and work-related hassles were significantly associated with increased snacking, whereas, physical stressors were significantly associated with decreased snacking. Finally, simultaneous consideration of the eating style variables showed emotional eating to be the pre-eminent moderating variable of the hassles-snacking relationship such that individuals who had higher levels of emotional eating were more likely to consume more between-meal snacks in response to daily hassles. References Affleck, G., Tennen, H., Urrows, S., & Higgins, P. (1994). Person and contextual features of daily stress reactivity: Individual differences in relations of undesirable daily events with mood disturbance and chronic pain intensity. Journal of Personality and Social Psychology, 66, 329 –340. Affleck, G., Zautra, A., Tennen, H., & Armeli, S. (1999). Multi-level daily process designs for consulting and clinical psychology: A preface for the perplexed. Journal of Consulting and Clinical Psychology, 67, 746 –754. Altman, D. G. (1991). Practical statistics for medical research. London: Chapman and Hall. Baumcom, D. H., & Aiken, P. A. (1981). Effects of depressed mood on eating among obese and nonobese dieting and nondieting persons. Journal of Personality and Social Psychology, 41, 577–585. Bellisle, F., Louis-Sylvestre, J., Linet, N., Rocaboy, B., Dalle, B., Cheneau, F., L’Hinoret, D., & Guyot, L. (1990). Anxiety and food intake in men. Psychosomatic Medicine, 52, 452– 457. Bolger, N., DeLongis, A., Kessler, R. C., & Schilling, E. A. (1989). Effects of daily stress on negative mood. Journal of Personality and Social Psychology, 57, 808 – 818. Bruch, H. (1973). Eating disorders: Obesity, anorexia nervosa, and the person within. In H. Bruch (Ed.), Hunger awareness and individuation eating disorders (pp. 44 – 65). New York: Basic Books. Cartwright, M., Wardle, J., Steggles, N., Simon, A. E., Croker, H., & Jarvis, M. J. (2003). Stress and dietary practices in adolescents. Health Psychology, 22, 362–9. Conner, M., & Armitage, C. J. (2002). The Social psychology of food. Buckingham: Open University Press. Conner, M., Fitter, M., & Fletcher, W. (1999). Stress and snacking: A diary study of daily hassles and between meal snacking. Psychology and Health, 14, 51– 63. Dancey, C. P., Taghavi, M., & Fox, R. J. (1998). The relationship between daily stress and symptoms of irritable bowel: A time-series approach. Journal of Psychosomatic Research, 44, 537– 45. Delongis, A., Coyne, J. C., Dakof, G., Folkman, S., & Lazarus, R. S. (1982). Relationship of daily hassles, uplifts and minor life events to health status. Health Psychology, 1, 119 –136. Delongis, A., Folkman, S., & Lazarus, R. S. (1988). The impact of daily stress on health and mood: Psychological resources as mediators. Journal of Personality and Social Psychology, 134, 609 – 619. Department of Health (n.d.). Five-A-Day. Retrieved January 14th, 2007, from http://www.dh.gov.uk/PolicyAndGuidance/HealthAndSocialCareTopics/FiveADay/fs/en Epel, E., Lapidus, R., McEwen, B., & Brownell, K. (2001). Stress may add bite to appetite in women: A laboratory study of stress-induced cortisol and eating behaviour. Psychoneuroendocrinology, 26, 37– 49. Feldman, S. I., Downey, G., & Schaffer-Neitz, R. (1999). Pain, negative mood, and perceived social support in chronic pain patients: A daily diary study of people with reflex sympathetic dystrophy syndrome. Journal of Consulting and Clinical Psychology, 67, 776 –785. Ferguson, E., Cassaday, H. J., & Bibby, P. (2004). Odors and sounds as triggers for medically unexplained symptoms: A fixed occasion diary study of in Gulf War veterans. Annals of Behavioral Medicine, 27, 205–214. Fifield, J., McQuillan, J., Armeli, S., Tennen, H., Reisine, S., & Affleck, G. (2004). Chronic strain, daily work stress and pain among workers with rheumatoid arthritis: Does job stress make a bad day worse? Work & Stress, 18, 275–291. Green, A. S., Rafaeli, E., Bolger, N., Shrout, P. E., & Reis, H. T. (2006). Paper or plastic? Data equivalence in paper and electronic diaries. Psychological Methods, 11, 87–105. Greeno, C. G., & Wing, R. R. (1994). Stress-induced eating. Psychological Bulletin, 115, 444 – 464. Grunberg, N. E., & Straub, R. O. (1992). The role of gender and taste class in the effects of stress on eating. Health Psychology, 11, 97–100. Heatherton, T. F., Herman, C. P., & Polivy, J. (1991). Effects of physical threat and ego threat on eating behavior. Journal of Personality and Social Psychology, 60, 138 –143. Heatherton, T. F., Herman, C. P., & Polivy, J. (1992). Effects of distress on eating: The importance of ego-involvement. Journal of Personality and Social Psychology, 62, 801– 803. STRESS AND EATING Heimendinger, J., Van Ryn, M. A., Chapelsky, D., Forester, S., & Stables, G. (1996). The national 5 A Day for Better Health Program: A largescale nutrition intervention. Journal of Public Health Management and Practice, 2, 27–35. Herman, C. P., Polivy, J., Pliner, P., Threlkeld, J., & Munic, D. (1978). Distractability in dieters and nondieters: An alternative view of externality. Journal of Personality and Social Psychology, 36, 536 –548. Hoffman, D. D. (1997). On overview of the logic and rationale for hierarchical linear Models. Journal of Management, 23, 723–744. Hox, J. (2002). Multilevel analysis: Techniques and applications. Lawrence Erlbaum Associates, London. Jones, F., & Bright, J. (2001). Stress: Myth, theory and research. London: Prentice Hall. Jones, F., O’Connor, D. B., Conner, M., McMillan, B. & Ferguson, E. (in press). Impact of Daily Mood, Work Hours, and Iso-Strain Variables on Self-reported Health Behaviors. Journal of Applied Psychology. In press. Kanner, A. D., Coyne, J. C., Schaefer, C., & Lazarus, R. S. (1981). Comparison of two modes of stress measurement: Daily hassles and uplifts versus major life events. Journal of Behavioral Medicine, 4, 1–39. Kaplan, H. I., & Kaplan, H. S. (1957). The psychosomatic concept of obesity. Journal of Nervous and Mental Diseases, 125, 181–201. Kreft, I., & De Leeuw, J. (2006). Introducing multilevel modeling. Sage, London. Kumanyika, S. K., Van Horn, L., Bowen, D., Perri, M. G., Rolls, B. J., Czajkowski, S. M., & Schron, E. (2000). Maintenance of dietary behavior change. Health Psychology, 19, 42–56. Lattimore, P., & Caswell, N. (2004). Differential effects of active and passive stress on food intake in restrained and unrestrained eaters. Appetite, 42, 167–73. Markus, C. R., Panhuysen, G., Tuiten, A., Koppeschaar, H., Fekkes, D., & Peters, M. L. (1998). Does carbohydrate-rich, protein-poor food prevent a deterioration of mood and cognitive performance of stress-prone subjects when subjected to a stressful task? Appetite, 31, 49 – 65. McCance, R. A., & Widdowson, E. M. (1991). The composition of foods. Letchworth, UK: The Royal Society of Chemistry. Newman, E., O’Connor, D. B., & Conner, M. (2007). Daily hassles and eating behaviour: The role of cortisol reactivity. Psychoneuroendocrinology, 32, 125–132. Nezlek, J. B. (2001). Multilevel random coefficient analyses of event- and interval contingent data in social and personality psychology research. Personality and Social Psychology Bulletin, 27, 771–785. O’Connor, D. B., & O’Connor, R. C. (2003). Predicting hopelessness and psychological distress: The role of perfectionism and coping. Journal of Counseling Psychology, 50, 362–372. O’Connor, D. B., & O’Connor, R. C. (2004). Perceived changes in food intake in response to stress: The role of conscientiousness. Stress & Health, 20, 279 –291. O’Connor, D. B., O’Connor, R. C., & Marshall, R. (2007). Perfectionism and psychological distress: Evidence of the mediating effects of rumination. European Journal of Personality, 21, 429 – 452. O’Connor, D. B., O’Connor, R. C., White, B. L., & Bundred, P. E. (2000). S31 Job strain and ambulatory blood pressure in British general practitioners: A preliminary study. Psychology, Health & Medicine, 5, 241–250. Oliver, G., Wardle, J., & Gibson, E. L. (2000). Stress and food choice: A laboratory study. Psychosomatic Medicine, 62, 853– 865. Oliver, K. G., Huon, G. F., Zadro, L., & Williams, K. D. (2001). The role of interpersonal stress in overeating among high and low disinhibitors. Eating Behaviors, 2, 19 –26. Ouwens, M. A., van Strien, T., & van der Staak, C. P. (2003). Tendency towards overeating and restraint as predictors of food consumption. Appetite, 40, 291–298. Raudenbush, S. W., Bryk, A., Cheong, Y. F., & Congdon, R. (2000). HLM5. Chicago, IL: Scientific Software. Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods. London: Sage. Rutledge, T., & Loh, C. (2004). Effects sizes and statistical testing in the determination of clinical significance on behavioral medicine research, Annals of Behavioral Medicine, 27, 138 –145. Sher, L. (2004). Daily hassles, cortisol, and the pathogenesis of depression. Medical Hypotheses, 62, 198 –202. Snijders, T., & Bosker, R. (1999). Multilevel analysis. London: Sage. Steptoe, A., Lipsey, Z., & Wardle, J. (1998). Stress, hassles and variations in alcohol consumption, food choice and physical exercise: A diary study. British Journal of Health Psychology, 3, 51– 63. Stone, A. A., & Brownell, K. D. (1994). The stress-eating paradox: Multiple daily measurements in adult males and females. Psychology & Health, 9, 425– 436. Stunkard, A. J., & Messick, S. M. (1985). The three-factor eating questionnaire to measure dietary restraint, disinhibition and hunger. Journal of Psychosomatic Research, 29, 71– 83. Tanofsky-Kraff, M., Wilfley, D. E., & Spurrell, E. (2000). Impact of interpersonal and ego-related stress on restrained eaters. International Journal of Eating Disorders, 27, 411– 418. Tennen, H., Affleck, G., Coyne, J. C., Larsen, R. J., & DeLongis, A. (2006). Paper or plastic in daily diary research: Comment on Green, Rafaeli, Bolger, Shrout, and Reis (2006). Psychological Methods, 11, 112–118. Van Horn, L., & Kavey, R. E. (1997). Diet and cardiovascular disease prevention: What works? Annals of Behavioral Medicine, 19, 197–212. Van Strien, T., Frijters, J. E., Bergers, G. P., & Defares, P. B. (1986). The Dutch Eating Behavior Questionnaire for assessment of restrained, emotional and external eating behavior. International Journal of Eating Behavior, 5, 295–315. Wallis, D. J., & Hetherington, M. M. (2004). Stress and eating: The effects of ego-threat and cognitive demand on food intake in restrained and emotional eaters. Appetite, 43, 39 – 46. Wardle, J. (1987). Eating style: A validation study of the Dutch eating behavior questionnaire in normal subjects and women with eating disorders. Journal Psychosomatic Research, 31, 161–169. Wardle, J., Steptoe, A., Oliver, G., & Lipsey, Z. (2000). Stress, dietary restraint and food intake. Journal of Psychosomatic Research, 48, 195–202. Wong, B. C., & Lam, S. K. (1999). Diet and cancer. GI Cancer, 3, 1–10.