Eas

ESa

b

c

d

e

f

g

a

ARRA

KSBTBDS

i1

uNFtHA

SS

(

h1

Body Image 11 (2014) 488500

Contents lists available at ScienceDirect

Body Image

journa l homepage: www.e lsev ier .com/ locate /bodyimage

xamining an elaborated sociocultural model of disordered eatingmong college women: The roles of social comparison and bodyurveillance

llen E. Fitzsimmons-Crafta,, Anna M. Bardone-Coneb, Cynthia M. Bulikc,d,e,tephen A. Wonderlich f,g, Ross D. Crosbyf,g, Scott G. Engel f,g

Department of Psychiatry, Washington University School of Medicine, St. Louis, MO, United StatesDepartment of Psychology, University of North Carolina at Chapel Hill, Chapel Hill, NC, United StatesDepartment of Psychiatry, University of North Carolina at Chapel Hill, Chapel Hill, NC, United StatesDepartment of Nutrition, University of North Carolina at Chapel Hill, Chapel Hill, NC, United StatesDepartment of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, SwedenNeuropsychiatric Research Institute, Fargo, ND, United StatesDepartment of Clinical Neuroscience, University of North Dakota School of Medicine & Health Sciences, Grand Forks, ND, United States

r t i c l e i n f o

rticle history:eceived 6 March 2014eceived in revised form 23 July 2014ccepted 27 July 2014

eywords:ocial comparison

a b s t r a c t

Social comparison (i.e., body, eating, exercise) and body surveillance were tested as mediators of the thin-ideal internalization-body dissatisfaction relationship in the context of an elaborated sociocultural modelof disordered eating. Participants were 219 college women who completed two questionnaire sessions 3months apart. The cross-sectional elaborated sociocultural model (i.e., including social comparison andbody surveillance as mediators of the thin-ideal internalization-body dissatisfaction relation) provideda good fit to the data, and the total indirect effect from thin-ideal internalization to body dissatisfaction

ody surveillancehin-ideal internalizationody dissatisfactionisordered eatingociocultural model

through the mediators was significant. Social comparison emerged as a significant specific mediator whilebody surveillance did not. The mediation model did not hold prospectively; however, social comparisonaccounted for unique variance in body dissatisfaction and disordered eating 3 months later. Resultssuggest that thin-ideal internalization may not be automatically associated with body dissatisfactionand that it may be especially important to target comparison in prevention and intervention efforts.

2014 Elsevier Ltd. All rights reserved.

Introduction

There is support for sociocultural models of disordered eat-

ng among college women (e.g., Dual Pathway Model Stice,994; Stice, Nemeroff, & Shaw, 1996; Tripartite Influence ModelThompson, Heinberg, Altabe, & Tantleff-Dunn, 1999) a group

This research is based on the doctoral dissertation of Ellen Fitzsimmons-Craftnder the direction of Anna Bardone-Cone and was conducted at the University oforth Carolina at Chapel Hill. This research was supported by the following grants:31 MH093978 from the National Institute of Mental Health; T32 HL007456 fromhe National Heart, Lung, and Blood Institute; University of North Carolina at Chapelill, Department of Psychology, Earl and Barbara Baughman Dissertation Researchward. Corresponding author at: Department of Psychiatry, Washington Universitychool of Medicine, Campus Box 8134, 660 S. Euclid Ave., St. Louis, MO 63110, Unitedtates. Tel.: +1 314 286 2074.

E-mail address: fitzsimmonscrafte@psychiatry.wustl.eduE.E. Fitzsimmons-Craft).

ttp://dx.doi.org/10.1016/j.bodyim.2014.07.012740-1445/ 2014 Elsevier Ltd. All rights reserved.

with high rates of diagnosable eating disorders, disordered eat-ing, and body dissatisfaction in the USA and other Westerncountries (e.g., Berg, Frazier, & Sherr, 2009; Eisenberg, Nicklett,Roeder, & Kirz, 2011; Mikolajczyk, Maxwell, El Ansari, Stock,Petkeviciene, Guillen-Grima, 2010; Said, Kypri, & Bowman, 2013;White, Reynolds-Malear, & Cordero, 2011). These socioculturalmodels tend to have several elements in common. Accordingto them, disordered eating is partially a result of pressure forwomen to achieve the thin ideal (Striegel-Moore, Silberstein, &Rodin, 1986). In order for this pressure to have the most pro-nounced negative impact, it must be internalized. Indeed, if awoman internalizes this pressure/the thin ideal, it is likely thatthis thin-ideal internalization will have adverse effects (Thompson,van den Berg, Roehrig, Guarda, & Heinberg, 2004). It is of notethough that among samples of college women, pressure for thin-

ness accounts for unique variance in body dissatisfaction, evenabove and beyond the variance accounted for by thin-ideal inter-nalization (e.g., Stice, Nemeroff, & Shaw, 1996). That is, on theirown, repeated messages that one is not thin enough may increase

dx.doi.org/10.1016/j.bodyim.2014.07.012http://www.sciencedirect.com/science/journal/17401445http://www.elsevier.com/locate/bodyimagehttp://crossmark.crossref.org/dialog/?doi=10.1016/j.bodyim.2014.07.012&domain=pdfmailto:fitzsimmonscrafte@psychiatry.wustl.edudx.doi.org/10.1016/j.bodyim.2014.07.012

/ Body

dtrS

dbSi2(eswodwlwtridpa1ba

S

hlq2p(cwi2Itrtba

Fd

E.E. Fitzsimmons-Craft et al.

issatisfaction with the body (e.g., Stice, 2001). Thus, pressure forhinness may result in body dissatisfaction both directly and indi-ectly via its influence on thin-ideal internalization (e.g., Stice &haw, 2002).

Both cross-sectional and prospective research studies haveemonstrated that thin-ideal internalization is associated withody dissatisfaction (e.g., Keery, van den Berg, & Thompson, 2004;hroff & Thompson, 2006; Stice & Whitenton, 2002). Body dissat-sfaction can in turn lead to disordered eating (Halliwell & Harvey,006). Yet, as highlighted by Fitzsimmons-Craft, Harney, et al.2012), sociocultural models of disordered eating typically lackxplanations as to how thin-ideal internalization leads to body dis-atisfaction and subsequent disordered eating. In theory, womenho have internalized the thin ideal would be at risk for devel-

ping body dissatisfaction when the ideal is not met, but howoes a woman come to know that there is a discrepancy betweenhat she would ideally like to look like and what she currently

ooks like? A better understanding of the mechanisms throughhich thin-ideal internalization is associated with body dissatisfac-

ion would inform prevention and intervention efforts and provideesearchers and clinicians with a more comprehensive understand-ng of the sociocultural influences underlying body dissatisfactionevelopment. The current study focused on two prominent socialsychological theories, namely social comparison (Festinger, 1954)nd objectification (Fredrickson & Roberts, 1997; McKinley & Hyde,996) theories, as explanations of the thin-ideal internalization-ody dissatisfaction relation in the context of a sociocultural modelmong college women.

ocial Comparison

Social comparison theory (Festinger, 1954) holds that humansave a natural drive to assess their progress and standing in

ife. There is ample evidence that college women engage in fre-uent comparisons with peers (e.g., Leahey, Crowther, & Mickelson,007), and research and theory have suggested that social com-arisons with peers, which are usually in the upward directioni.e., an individual compares herself to someone whom she per-eives to be better off in some way), may be one pathway throughhich internalized pressures for thinness develop into body dissat-

sfaction and disordered eating (Dittmar, 2005; Dittmar & Howard,004; Fitzsimmons-Craft, Harney, et al., 2012; Leahey et al., 2007).

t may be that via social comparison, individuals come to knowhat they have not yet actualized their ideal. Indeed, appearance-

elated social comparison and body comparison have been foundo mediate the relation between thin-ideal internalization andody dissatisfaction in cross-sectional studies of preadolescentnd adolescent girls (Blowers, Loxton, Grady-Flesser, Occhipinti,

ig. 1. An elaborated sociocultural model of disordered eating. Social comparison and boissatisfaction link.

Image 11 (2014) 488500 489

& Dawe, 2003; Carey, Donaghue, & Broderick, 2014). However,Fitzsimmons-Craft, Harney, et al. (2012) found that neither gen-eral nor appearance-related social comparison tendencies uniquelymediated this relation (above and beyond the effects of bodysurveillance) in a cross-sectional study of college women. Theyhypothesized that the general measure of social comparison usedin this study may have been too general and that the appearance-related social comparison measure may have been too narrow. Forexample, other social comparison domains, such as those related toeating and exercise, may also stem from thin-ideal internalizationand be associated with body image disturbance. Thus, examiningthe roles of body, eating, and exercise comparisons may be impor-tant in terms of coming to a more comprehensive understanding ofthe ways in which social comparison behavior contributes to bodydissatisfaction and disordered eating.

Body Surveillance

Objectification theory holds that within dominant Americanculture, the female body has been constructed as an object tobe looked at (Fredrickson & Roberts, 1997; McKinley & Hyde,1996). As a result, girls and women learn to view themselves froman observers perspective and to treat themselves as objects tobe looked at. This self-objectification is thought to behaviorallymanifest itself in the act of body surveillance (Moradi & Huang,2008), which involves thinking about how ones body looks toan outside observer and thinking more about how ones bodylooks than how it feels (McKinley & Hyde, 1996). In other words,self-objectification describes a perspective of oneself, while bodysurveillance is the active, behavioral manifestation of this view-point. It is via this surveillance that many women realize there isa discrepancy between what they see and what they would ideallylike to look like, and thus, may experience negative consequences,such as body dissatisfaction (e.g., Knauss, Paxton, & Alsaker, 2008;McKinley & Hyde, 1996). Indeed, Myers and Crowther (2007) andFitzsimmons-Craft, Harney, et al. (2012) found that both the pro-cess of self-objectification and the behavior of body surveillancemediated the relation between internalization of the thin ideal andbody dissatisfaction in cross-sectional studies of college women.

The Current Study

Research on sociocultural models of disordered eating typi-cally lacks a comprehensive understanding as to how thin-ideal

internalization leads to body dissatisfaction and subsequent dis-ordered eating. Additionally, although aspects of socioculturalmodels of disordered eating have been tested longitudinally (e.g.,Stice, Shaw, & Nemeroff, 1998), much of the work in this area has

dy surveillance are conceived as mediators of the thin-ideal internalization-body

4 / Body

b1ibisda1(

Febbmpb(tttlifisfiwdtstsssWbpfttTT

tcclowrtbiifat

P

S

90 E.E. Fitzsimmons-Craft et al.

een cross-sectional (e.g., Stice, Schupak-Neuberg, Shaw, & Stein,994; Twamley & Davis, 1999), which has precluded understand-

ng causal mechanisms. Further, the longitudinal work that haseen done has typically not controlled for the temporal stabil-

ty of study constructs (Stice, 2001; Stice & Bearman, 2001). Thistudy addresses these limitations of prior research by collectingata from female undergraduates via questionnaires administeredt two time points: at the beginning of an academic semester (Time; T1) and about 3 months later at the end of the academic semesterTime 2; T2).

The current study should be considered an extension ofitzsimmons-Craft, Harney, et al. (2012), which simultaneouslyxamined social comparison (general and appearance-related) andody surveillance as mediators of the thin-ideal internalization-ody dissatisfaction relation using a cross-sectional design andanifest variables. In the present study, we examined social com-

arison (including body, eating, and exercise comparisons) andody surveillance, which are correlated but distinct constructse.g., Fitzsimmons-Craft, Harney, et al., 2012), as mediators of thehin-ideal internalization-body dissatisfaction relation in the con-ext of an elaborated sociocultural model of disordered eating (i.e.,hat includes social comparison and objectification theories) usingatent variables (see Fig. 1). All paths specified between constructsn the model were grounded in previously discussed empiricalndings and/or in theory. It was hypothesized that the elaboratedociocultural model of disordered eating would provide a goodt to the data and that social comparison and body surveillanceould significantly mediate the thin-ideal internalization-bodyissatisfaction relation within this model. Of note is the fact thathe Tripartite Influence Model (Thompson et al., 1999) examinesocial comparison and thin-ideal internalization as mediators ofhe relationship between sociocultural influence (similar to pres-ure for thinness) and body dissatisfaction. As such, the elaboratedociocultural model examined in the current study should be con-idered a revision and extension of the Tripartite Influence Model.

e also examined a model not including social comparison andody surveillance (i.e., a more traditional sociocultural model:ressure for thinness thin-ideal internalization body dissatis-action disordered eating, and including a path from pressure forhinness to body dissatisfaction) in order to ascertain what (if any-hing) is gained by incorporating these constructs in the model.hese models were examined cross-sectionally (i.e., using only the1 data).

In a more exploratory fashion, we examined the prospec-ive relations among the constructs involved in the mediationalomponent of this model (i.e., thin-ideal internalization, socialomparison, body surveillance, body dissatisfaction). In particu-ar, we examined if these constructs significantly predicted levelsf and change in one another over time, prior to examininghether social comparison and body surveillance mediated the

elation between thin-ideal internalization and body dissatisfac-ion prospectively. We also examined the prospective relationsetween social comparison/body surveillance and disordered eat-

ng (the endpoint of sociocultural models), as we were interestedn both whether social comparison and body surveillance predicteduture levels of disordered eating and whether social comparisonnd body surveillance predicted change in disordered eating overhe course of 3 months.

Method

articipants

Participants were 238 women attending a large, US publicoutheastern university; at the first study assessment, they ranged

Image 11 (2014) 488500

in age from 17 to 22 years, with a mean age of 18.71 years(SD = 1.00). Recruitment occurred through introductory psychologycourses. Most women (68.5%) identified themselves as White, 8.4%as African American or Black, 7.6% as Asian, 4.2% as Hispanic, 1.3% asAmerican Indian or Alaskan Native, 9.7% as multiracial/multiethnic,and .4% as other races/ethnicities. Highest parental education wasused as a proxy for socioeconomic status and ranged from 7 to21 years (M = 16.50 years, SD = 2.68). This samples mean score onthe Eating Attitudes Test-26 (EAT-26; Garner, Olmsted, Bohr, &Garfinkel, 1982) was 9.24 (SD = 7.30). The EAT-26 is a commonlyused measure of eating disorder attitudes and behaviors, and ascore of 20 or more indicates a probable eating disorder (King,1989, 1991). On average, this sample exhibited a level of disor-dered eating that was similar in magnitude to that observed in otherstudies of college women (e.g., Desai, Miller, Staples, & Bravender,2008; Fitzsimmons-Craft, Bardone-Cone, & Harney, 2012). Westonand Gore (2006) recommend a minimum sample size of 200 forany structural equation model (SEM); thus, the current sample sizeshould be sufficient for detecting significant effects if they exist.

Measures at Time 1 (T1) and Time 2 (T2)

Where possible, we assessed each study construct via at leastthree different measures so that latent variables with multipleindicators could be used in our SEM analyses (described below).However, there were three constructs for which we did not usethree separate measures. In the case of pressure for thinness, weused only two measures to assess this construct given that weidentified only two psychometrically supported existing measuresof this construct in the literature. In the case of social compar-ison, this construct was assessed via one measure (the Body,Eating, and Exercise Comparison Orientation Measure (BEECOM);Fitzsimmons-Craft, Bardone-Cone, & Harney, 2012), which com-prehensively assesses eating disorder-related social comparisonbehavior with three subscales. For SEM analyses, these three sub-scales were used as indicators of an eating disorder-related socialcomparison behavior latent variable. Given that, to our knowledge,only one measure of body surveillance exists, this was the onlymeasure of this construct that was administered, and the items thatcomprise this measure were used as indicators of a body surveil-lance latent variable in SEM analyses.

Demographics. Demographic data for age, parents highest lev-els of education, and race/ethnicity were collected at T1 via a setof questionnaires created for this study. Additionally, participantsreported on their current weight and height at T1, and we used thisinformation to compute body mass index (BMI), as we were inter-ested in the ways in which BMI may relate to the study constructs.There is evidence that individuals are generally accurate with theirself-reported weights (Shapiro & Anderson, 2003).

Pressure for thinness. Pressure for thinness was measuredvia two questionnaires. The Perceived Sociocultural Pressure Scale(PSPS; Stice & Agras, 1998; Stice & Bearman, 2001; Stice, Ziemba,Margolis, & Flick, 1996) assesses perceived pressure from family,friends, dating partners, and the media to be thin. This measureconsists of eight items that are rated on a 1 (none) to 5 (a lot) scale,and items are averaged to create a total score. Past research with asimilar measure has found that child reports of parental pressureto lose weight correspond well with parental self-reports of suchpressure (Thelen & Cormier, 1995), and research has indicated ade-quate internal consistency among samples of adolescent femalesand individuals with bulimia nervosa (Cronbachs alphas from .83to .88; Stice & Agras, 1998; Stice, Ziemba, et al., 1996). In the current

study, alpha was .79 at T1.

We also used the Pressures subscale of the Sociocultural Atti-tudes Toward Appearance Questionnaire-4 (SATAQ-4; Schaeferet al., in press) to assess pressure for thinness. This subscale of the

/ Body

Smsts(cve.

nISal1avd(qobe

I(taisgceaS

tADumaawwaaP.

baHosptotwt(Cc

E.E. Fitzsimmons-Craft et al.

ATAQ-4 assesses perceived pressure from family, peers, and theedia to be thin/to strive for cultural ideals of beauty. The sub-

cale consists of 12 items that are rated on a 1 (definitely disagree)o 5 (definitely agree) scale, and items are summed to create a totalcore. Schaefer et al. (in press) reported high internal consistencyCronbachs alphas = .85.96) in a large sample of college womenollected from four sites and demonstrated excellent constructalidity (e.g., the Pressures subscale correlated with a measure ofating disorder psychopathology). In the current study, alpha was

92 at T1.Thin-ideal internalization. The construct of thin-ideal inter-

alization was assessed using three measures. First, we used thedeal-Body Stereotype Scale-Revised (IBSS-R; Stice & Agras, 1998;tice, Ziemba et al., 1996), which assesses participants level ofgreement with statements concerning what attractive womenook like. This measure consists of six items that are rated on a

(strongly disagree) to 5 (strongly agree) scale, and items are aver-ged to create a total score. Regarding convergent and discriminantalidity, the IBSS-R evidenced a stronger correlation with a bodyissatisfaction measure than with a measure of negative affectStice, Ziemba et al., 1996) in a sample of adolescent females. Ade-uate internal consistency has been demonstrated among samplesf adolescent females and individuals with bulimia nervosa (Cron-achs alphas from .89 to .91; Stice & Agras, 1998; Stice, Ziembat al., 1996). In the current study, alpha was .72 at T1.

Thin-ideal internalization was also measured via thenternalization-Thin/Low Body Fat subscale of the SATAQ-4Schaefer et al., in press), which assesses endorsement and accep-ance of messages that espouse unrealistic ideals for female beautynd the striving toward such ideals. This subscale consists of fivetems that are rated on a 1 (definitely disagree) to 5 (definitely agree)cale, and items are summed to create a total score. Evidence ofood construct validity has been demonstrated (e.g., relatively highorrelation with a measure of eating disorder pathology; Schaefert al., in press) and high internal consistency has been reported inlarge sample of college women from four sites (alphas of .87.92;chaefer et al., in press). In the current study, alpha was .83 at T1.

Finally, thin-ideal internalization was measured via the Impor-ance of Being Attractive and Thin subscale of the Beliefs Aboutttractiveness Scale-Revised (BAAR; Petrie, Rogers, Johnson, &iehl, 1996), which assesses agreement with Western societal val-es regarding the importance of being thin and attractive. Thiseasure consists of 10 items that are rated on a 1 (strongly dis-

gree) to 7 (strongly agree) scale, and items are averaged to createtotal score. Evidence of construct validity in a sample of collegeomen is demonstrated by the subscales significant associationsith bulimic symptomatology, concern with body size and shape,

nd depression (Petrie et al., 1996). In samples of college women,lphas of .84.86 have been reported for this subscale (Bradford &etrie, 2008; Wood & Petrie, 2010). In the current study, alpha was

89 at T1.Social comparison. Social comparison behavior, including

ody, eating, and exercise social comparison tendencies, wasssessed using the BEECOM (Fitzsimmons-Craft, Bardone-Cone, &arney, 2012). This measure consists of 18 items that are ratedn 1 (never) to 7 (always) scale. Items are summed to createubscale scores (i.e., Body Comparison Orientation, Eating Com-arison Orientation, and Exercise Comparison Orientation) andhe total score (i.e., eating disorder-related social comparisonrientation). Evidence of construct validity is demonstrated byhe subscales and total scores significant positive correlationsith general social comparison orientation, eating disorder symp-

omatology, and body dissatisfaction in a female college sampleFitzsimmons-Craft, Bardone-Cone, & Harney, 2012). Fitzsimmons-raft, Bardone-Cone, and Harney (2012) found estimates of internalonsistency for the subscale and total scores that ranged from .93 to

Image 11 (2014) 488500 491

.97. In the current study, alphas for the BEECOM total and subscalescores were all .91 or greater at both T1 and T2.

Body surveillance. Body surveillance was assessed via the BodySurveillance subscale of the Objectified Body Consciousness Scale(OBCS; McKinley & Hyde, 1996). This subscale consists of eightitems that are rated on a 1 (strongly disagree) to 7 (strongly agree)scale, and items are averaged to create a subscale score. This sub-scale contains one comparison-related item (i.e., I rarely comparehow I look with how other people look); in order to minimizeissues related to construct overlap, this item was not included whencomputing the subscale score, so that all analyses were run usingthe 7-item version of the OBCS Body Surveillance score. Also ofnote, if more than two items are missing on an OBCS subscale(with a not applicable response option being counted as miss-ing), then the score for that subscale is not computed (McKinley &Hyde, 1996). However, no study participants had more than twoitems missing on the Body Surveillance subscale at either T1 or T2.Construct validity in a sample of college women is demonstratedby high correlations with public self-consciousness and nonsignif-icant relations with private self-consciousness (McKinley & Hyde,1996). McKinley and Hyde (1996) reported a coefficient alpha of.89 in a sample of student and nonstudent women. In the currentstudy, alpha was .79 at T1 and .86 at T2 for the 7-item version ofthe OBCS Body Surveillance subscale.

Body dissatisfaction. The construct of body dissatisfaction wasassessed in a number of ways. First, we measured body dissatis-faction using the Body Shape Questionnaire (BSQ; Cooper, Taylor,Cooper, & Fairburn, 1987), which assesses concerns about bodyshape, in particular, participants experience of feeling fat overthe past four weeks. This measure consists of 34 items that are ratedon a 1 (never) to 6 (always) scale, and items are summed to createa total score. The BSQ has demonstrated validity and reliability insamples of body image therapy patients, obese individuals seekingweight reduction, and nonclinical samples of college students andadults (e.g., Rosen, Jones, Ramirez, & Waxman, 1996). For example,research has indicated that scores on the BSQ are correlated withother types of negative body image symptoms, including concernsabout non-weight-related appearance features (Rosen et al., 1996).In the current study, alpha was .97 at T1 and .98 at T2.

Next, we measured body dissatisfaction via the Body Dissatis-faction subscale of the Eating Disorder Inventory (EDI-BD; Garner,Olmsted, & Polivy, 1983), which assesses the belief that specificparts of the body are too large (e.g., hips, thighs, buttocks). This mea-sure consists of nine items that are rated on a six-point scale rangingfrom never to always. Garner et al. (1983) originally recommendedthat item responses never, rarely, and sometimes receive a score of0, and the responses often, usually, and always receive scores of 1, 2,and 3 respectively; however, because this reduces the variability inresponses in non-clinical samples, we coded these responses usingthe continuous six-point scale (see Tylka & Subich, 2004). Itemsare summed to create a total subscale score. Construct validity insamples of college women is demonstrated by the measures highcorrelations with body preoccupation (Tylka & Subich, 2004) andeating disordered behavior (Spillane, Boerner, Anderson, & Smith,2004). Reliability coefficients for college women range from .83 to.93 (Garner et al., 1983). In the current study, alpha was .91 at T1and .92 at T2.

Finally, body dissatisfaction experienced over the past 28 dayswas assessed via the Weight Concern and Shape Concern sub-scales of the Eating Disorder Examination-Questionnaire (EDE-Q;Fairburn & Beglin, 2008), which is one of the most commonly usedmeasures of disordered eating attitudes and behaviors in clini-

cal and community populations (Anderson & Williamson, 2002).These two subscales focus on weight and shape dissatisfactionand the degree to which ones self-worth and acceptance of one-self are defined by weight or shape, and were combined since

4 / Body

poi(tihL1wa

itSsscacrosWs

2w(ticag

wEessmanb1b

P

tedIo(trifnsI

2i

92 E.E. Fitzsimmons-Craft et al.

revious work has indicated that these two subscales load ontone underlying factor (Peterson et al., 2007). In particular, the 12tems that comprise these two subscales were rated on a 06 scalewith items either rated on a no days to everyday scale or a not at allo markedly scale). Items are averaged to create a total body dissat-sfaction score. The Weight Concern and Shape Concern subscalesave demonstrated good internal consistency (alphas of .89.93;uce & Crowther, 1999) and convergent validity (Fairburn & Beglin,994; Grilo, Masheb, & Wilson, 2001) among samples of collegeomen and community and patient groups. In the current study,

lpha for this combined subscale was .94 at both T1 and T2.Disordered eating. We used several measures of disordered eat-

ng in order to capture various facets of this construct. First, we usedhe Bulimia Test-Revised (BULIT-R; Thelen, Farmer, Wonderlich, &mith, 1991) to assess bulimic attitudes and behaviors. This mea-ure consists of 36 items (with 28 items contributing to the BULIT-Rcore) that have a five-option multiple choice format. Constructoverage is broad, with items on binge eating, purging, and neg-tive attitudes related to weight and shape. Items are summed toreate a total score. The BULIT-R has well-established psychomet-ic properties and has been successfully used to aid in the diagnosisf bulimia nervosa and in the measurement of bulimic symptomeverity in clinical and nonclinical populations (Thelen et al., 1991;

illiamson, Anderson, Jackman, & Jackson, 1995). In the currenttudy, alpha was .93 at both T1 and T2.

We used the Restraint subscale of the EDE-Q (Fairburn & Beglin,008) to assess attempts to restrict food intake over the past foureeks. This subscale consists of five items that are rated on a 0

no days) to 6 (everyday) scale, and items are averaged to create aotal subscale score. The Restraint subscale has demonstrated goodnternal consistency (alphas of .84.85; Luce & Crowther, 1999) andonvergent validity (Fairburn & Beglin, 1994; Grilo et al., 2001)mong samples of college women and community and patientroups. In the current study, alpha was .81 at T1 and .83 at T2.

Finally, the Eating Attitudes Test-26 (EAT-26; Garner et al., 1982)as used to assess eating disorder symptoms more generally. The

AT-26 is one of the most widely used standardized measures ofating disorder attitudes and behaviors (Garner, 1993). This mea-ure consists of 26 items that are rated on a 1 (never) to 6 (always)cale. Items endorsed as 1, 2, or 3 are scored as 0, while itemsarked as 4, 5, or 6, are scored as 1, 2, or 3, respectively. Items

re then summed to create a total score. Additionally, good inter-al consistency ( = .83.90) and testretest reliability (r = .84) haveeen demonstrated in samples of young women (Carter & Moss,984; Garner et al., 1982). In the current study, alpha was .81 atoth T1 and T2.

rocedure

As part of a larger study, a questionnaire battery was adminis-ered at baseline (beginning of semester, T1) and then again at thend of the semester (T2; about 3 months later), so as to completeata collection within a single semester and maximize retention.

nformed consent was obtained at T1. Questionnaire completionccurred online in private locations of the participants choosinge.g., their homes), and each battery of questionnaires (T1 and T2)ook about one hour to complete. Participants were provided withesearch credit in their introductory psychology courses for partic-pating in this study. Participants were also entered into a drawingor one of six $100 prizes if they adequately completed all compo-ents of the larger study (including completing both questionnaireessions). This study was reviewed and approved by the universitys

nstitutional Review Board.

Regarding attrition, three study participants (out of a total N of38) only completed T1 (i.e., did not complete T2). Two of these

ndividuals dropped their introductory psychology course after

Image 11 (2014) 488500

completing T1 and thus no longer needed credit for this course;the third individual dropped the study for personal reasons. Thus,in total, 235 individuals completed both T1 and T2; however, thedata from the three individuals who completed only T1 were usedfor analyses involving only T1 data. Additionally, in an attempt tocontrol for random responding and inattentiveness, a validity checkitem was included in both the T1 and T2 questionnaire batteries(about three-quarters of the way through the survey). These itemsasked participants to choose a specific response choice (i.e., theitems stated: Please choose Disagree Strongly); not respondingappropriately to these items suggests possible random or inatten-tive responding. Of the 238 participants who completed T1, 12failed the T1 validity check (5.1%), and of the 235 participantswho completed T2, eight failed the T2 validity check (3.4%). Ofthe 235 participants who completed both T1 and T2, 16 failedeither the T1 or T2 validity check or both validity checks (6.8%).For analyses involving T1 but not T2 data, we excluded all whofailed the T1 validity check (resulting in n = 226), and for analysesinvolving T2 but not T1 data (e.g., obtaining alphas for measures atT2), we excluded all who failed the T2 validity check (resulting inn = 227). For analyses involving both T1 and T2 data, we excluded allwho failed either validity check or both validity checks (resultingin n = 219).

Analytic Strategy

SEM using data from T1 and Mplus Version 6.1 (Muthn &Muthn, 2010) was used to examine the more traditional and elabo-rated sociocultural models of disordered eating. Because these twomodels are not nested, it was not possible to test the difference in fitbetween them using a nested chi-square difference test (Widaman& Thompson, 2003). Instead, we focused on determining whetherthe total and specific indirect effects of thin-ideal internalizationon body dissatisfaction via social comparison and body surveil-lance were significant and on the size of the residual direct effectfrom thin-ideal internalization to body dissatisfaction once thesemediators were included in the model.

We used maximum likelihood estimation to examine themore traditional and elaborated versions of the socioculturalmodel of disordered eating cross-sectionally using latent variables.Goodness-of-fit was evaluated using the root mean square errorof approximation (RMSEA), the standardized root-mean-squareresidual (SRMR), the comparative fit index (CFI), and the Tucker-Lewis Index (TLI). Good model fit was defined by the followingcriteria: RMSEA values of about .08 or below (Browne & Cudeck,1993), SRMR values less than about .08 (Hu & Bentler, 1999), CFIvalues of about .95 or above (Bentler, 1990; Hu & Bentler, 1999),and TLI values above about .90 (Hu & Bentler, 1999). Multiple fitindices were used together because they provide a more conserva-tive and reliable approach to the evaluation of model fit than theexamination of a single index of fit.

Next, we examined if thin-ideal internalization prospectivelypredicted social comparison and body surveillance and if socialcomparison and body surveillance prospectively predicted bodydissatisfaction both without and with controlling for baseline lev-els of the outcome variable. We used latent variables and SEMfor investigating these possibilities and used a similar strategy forinvestigating the prospective relations between social compari-son/body surveillance and disordered eating as well.

If we determine that these constructs (i.e., thin-ideal internal-ization, social comparison, body surveillance, body dissatisfaction)significantly predict change in one another over time, our inten-

tion was to proceed with examining whether social comparisonand body surveillance mediated the relation between thin-idealinternalization and body dissatisfaction prospectively. Since ourdata came from two time points, half-longitudinal tests were to

E.E. Fitzsimmons-Craft et al. / Body Image 11 (2014) 488500 493

Table 1Means and standard deviations of the measured variables at T1 (n = 226).

Construct Measure M SD Possible range

Pressure for thinness 1. Perceived Sociocultural Pressure Scale (PSPS) 2.48 0.75 152. Sociocultural Attitudes Toward Appearance Questionnaire-4 (SATAQ-4), Pressure 34.42 11.08 1260

Thin-ideal internalization 3. Ideal-Body Stereotype Scale-Revised (IBSS-R) 3.78 0.53 154. SATAQ-4, Internalization 17.48 4.75 5255. Beliefs About Attractiveness Scale-Revised (BAAR), Importance of Being Attractive and Thin 3.04 1.15 17

Social comparison 6. Body, Eating, and Exercise Comparison Orientation Scale (BEECOM), Body 31.26 7.20 6427. BEECOM, Eating 28.26 7.43 6428. BEECOM, Exercise 22.97 8.03 6429. BEECOM, Total 82.51 20.48 18126

Body surveillance 10. Objectified Body Consciousness Scale (OBCS), Surveillance 5.04 0.93 17

Body dissatisfaction 11. Body Shape Questionnaire (BSQ) 92.17 34.48 3420412. Eating Disorder Inventory-Body Dissatisfaction (EDI-BD) 31.94 10.22 95413. Eating Disorder Examination-Questionnaire-6 (EDE-Q-6), Weight Concern/Shape Concern 2.67 1.53 06

Disordered eating 14. Bulimia Test-Revised (BULIT-R) 49.57 16.39 2814015. EDE-Q-6, Restraint 1.50 1.36 0616. Eating Attitudes Test-26 (EAT-26) 9.24 7.30 078

btoutofsiilttAtpbbpis

D

itT(ttterB(tribi

Body mass index 17. Body mass index (BMI)

e employed in this study (Cole & Maxwell, 2003). We intendedo first estimate the paths in the regression of the T2 mediatorsnto T1 thin-ideal internalization controlling for T1 mediator val-es (i.e., the a1 and a2 regression coefficients). Then we intendedo estimate the paths in the regression of T2 body dissatisfactionnto the T1 mediators controlling for T1 levels of body dissatis-action (i.e., the b1 and b2 regression coefficients). Estimates of thepecific indirect effects can be calculated by means of multiply-ng together the a1 and b1 terms and the a2 and b2 terms. The totalndirect effect associated with the two mediators can then be calcu-ated using the formula a1b1 + a2b2, where the two terms representhe indirect effects of thin-ideal internalization on body dissatisfac-ion through social comparison and body surveillance, respectively.ssuming that the conditions for stationarity (i.e., stable casual rela-

ionship between two variables over time; Kenny, 1979) are met,aths between the T1 mediators and T2 body dissatisfaction woulde equal to the paths between T2 mediators and a hypothetical T3ody dissatisfaction. Under this assumption, the aibi product termsrovide estimates of the mediational effect of thin-ideal internal-

zation on body dissatisfaction through social comparison and bodyurveillance.

Results

escriptive and Preliminary Analyses

We evaluated skewness and kurtosis for each measure (includ-ng the individual items of the OBCS Body Surveillance subscale)hat was to be used in the SEM analyses at either T1 or both1 and T2. We determined that no substantial violations existedper Tabachnick & Fidell, 2012), and thus, no measures/items wereransformed. Table 1 contains means and standard deviations forhe study variables at T1, and Table 2 contains correlations amonghe study variables at T1. Correlations were in the directionsxpected based on the literature; that is, we found positive cor-elations between all measured variables, with the exception thatMI exhibited a significant negative correlation with the IBSS-Rr = .18, p = .006). In SEM analyses, it is recommended that indica-ors of separate latent variables not be very highly correlated (i.e.,

s should be less than .90; Tabachnick & Fidell, 2012). As can be seenn Table 2, many indicators of separate latent variables are relatedut the rs do not reach .90. This was also the case when we exam-

ned the correlations between the OBCS Body Surveillance subscale

22.69 3.42 Actual range:17.4341.60

items, which were used as indicators of the body surveillance latentvariable, and the other study constructs at T1, and when we exam-ined correlations for the study variables/items at T2 that were usedin SEM analyses.

Structural Equation Modeling (SEM) Analyses

More traditional sociocultural model of disordered eating.Following the recommendations of Tabachnick and Fidell (2012),we evaluated the adequacy of the more traditional socioculturalmodel of disordered eating measurement model before simultane-ously evaluating both the measurement and structural componentsof the model. The SRMR (.042), CFI (.959), and TLI (.941) allapproximated good fit for the more traditional sociocultural modelmeasurement model according to the aforementioned criteria.However, the RMSEA value we obtained (.088; 90% confidenceinterval: .068.108) indicated mediocre model fit (MacCallum,Browne, & Sugawara, 1996). Chen, Curran, Bollen, Kirby, and Paxton(2008) recommend that RMSEA values be evaluated in the contextof other fit indices, rather than solely on strict cutoff values. Onthis basis, and considering that the other fit indices indicated goodmodel fit, we concluded that the more traditional socioculturalmodel measurement model had an acceptable fit. Additionally, allmeasures loaded significantly onto their respective latent factors.This information suggests that these latent factors were adequatelyoperationalized. Thus, this measurement model was used to testthe more traditional sociocultural model of disordered eating struc-tural model. Correlations between the latent variables were allpositive and significant (rs of .54.86, all ps < .001), and factor load-ings are included in Fig. 2.

Next, we evaluated the more traditional sociocultural modelstructural model. As with the measurement model, the structuralmodel provided an acceptable fit to the data. The SRMR (.048),CFI (.951), and TLI (.933) all approximated good fit. However, theRMSEA (.093; 90% confidence interval: .074.113) again indicatedmediocre model fit (MacCallum et al., 1996). Considering thatthe other fit indices indicated good model fit, we concluded thatthe more traditional sociocultural model structural model had an

acceptable fit. All model paths were positive and significant andare presented in Fig. 2. Results indicated that pressure for thinnessaccounted for 37.7% of the variance in thin-ideal internalization.Pressure for thinness and thin-ideal internalization accounted for

494 E.E. Fitzsimmons-Craft et al. / Body Image 11 (2014) 488500

Fig. 2. Standardized path coefficients and factor loadings for a more traditional sociocultural model of disordered eating structural model. PSPS = Perceived SocioculturalP ire-4.S hin suD Bulim

6f

etaWt.mafqumai

savfi.rgtasitpspio5tiibia

ressure Scale. SATAQ-4 = Sociocultural Attitudes Toward Appearance Questionnacale-Revised; we note that we are using the Importance of Being Attractive and Tissatisfaction. EDE-Q-6 = Eating Disorder Examination-Questionnaire-6. BULIT-R =

5.7% of the variance in body dissatisfaction. Lastly, body dissatis-action accounted for 75.2% of the variance in disordered eating.

Elaborated sociocultural model of disordered eating. We nextxamined an elaborated sociocultural model of disordered eatinghat incorporated social comparison and body surveillance as medi-tors of the thin-ideal internalization-body dissatisfaction relation.e again first tested the measurement model before analyzing

he structural model. The RMSEA (.068; 90% confidence interval:058.078), SRMR (.054), CFI (.935), and TLI (.921) all approxi-

ated good fit according to the aforementioned criteria. Further,ll measures/items loaded significantly onto their respective latentactors. This information suggests that the latent factors were ade-uately operationalized, and thus, this measurement model wassed to examine the elaborated sociocultural model structuralodel. Correlations between the latent variables were all positive

nd significant (rs of .38.86, all ps < .001), and factor loadings arencluded in Fig. 3.

We then evaluated the structural model for the elaboratedociocultural model of disordered eating, which provided ancceptable fit to the data. The RMSEA (.070; 90% confidence inter-al: .060.079), SRMR (.057), and TLI (.917) all approximated goodt. The CFI (.928) was slightly below the aforementioned criterion of

95; however, some work has indicated that CFI values greater thanoughly .90 may indicate adequate fit (Kline, 2005). Additionally,iven that other fit indices indicated good model fit, we concludedhat the elaborated sociocultural model structural model had ancceptable fit. All model paths except for two were positive andignificant; the non-significant paths were: the path from thin-deal internalization to body dissatisfaction ( = .07, p = .703) andhe path from body surveillance to body dissatisfaction ( = .11,= .196). See Fig. 3 for the full structural model for the elaborated

ociocultural model of disordered eating. Results indicated thatressure for thinness accounted for 40.4% of the variance in thin-

deal internalization. Thin-ideal internalization accounted for 77.3%f the variance in eating disorder-related social comparison and for1.5% of the variance in body surveillance. Pressure for thinness,hin-ideal internalization, eating disorder-related social compar-son, and body surveillance accounted for 72.0% of the variance

n body dissatisfaction. Thus, by including social comparison andody surveillance in the model, an additional 6.3% of the variance

n body dissatisfaction was explained. Finally, body dissatisfactionccounted for 75.2% of the variance in disordered eating.

IBSS-R = Ideal-Body Stereotype Scale-Revised. BAAR = Beliefs About Attractivenessbscale. BSQ = Body Shape Questionnaire. EDI-BD = Eating Disorder Inventory-Bodyia Test-Revised. EAT-26 = Eating Attitudes Test-26. ***p < .001.

Given our interest in examining whether social comparison andbody surveillance would mediate the thin-ideal internalization-body dissatisfaction relation in the context of the socioculturalmodel of disordered eating, it is notable that the path fromthin-ideal internalization to body dissatisfaction was no longer sig-nificant once these constructs (i.e., social comparison, body surveil-lance) were included in the model. Indeed, results indicated that thetotal indirect effect of thin-ideal internalization on body dissatis-faction through social comparison and body surveillance (as a set)was significant, with a standardized point estimate of .47 (p < .001).Thus, social comparison and body surveillance significantly medi-ated the relation between thin-ideal internalization and bodydissatisfaction in the context of this model. Given that the directeffect of thin-ideal internalization on body dissatisfaction in thismodel was not significant ( = .07, p = .703), this suggests indirect-only mediation (Zhao, Lynch, & Chen, 2010), which is also known asfull mediation (Baron & Kenny, 1986). The specific indirect effectsof each mediator showed that social comparison was a uniqueand significant mediator, with a standardized point estimate of .39(p = .003). However, body surveillance was not a significant specificmediator of the thin-ideal internalization-body dissatisfaction rela-tion, with a standardized point estimate of .08 (p = .198). A contrastconfirmed that the indirect effect of social comparison in the thin-ideal internalization-body dissatisfaction relation was significantlystronger (p = .001) than the indirect effect of body surveillance.

Prospective examination of social comparison and bodysurveillance as mediators of the thin-ideal internalization-bodydissatisfaction link. Given that the elaborated sociocultural modelof disordered eating provided a good fit to the data and the factthat social comparison and body surveillance (as a set) were foundto significantly mediate the thin-ideal internalization-body dissat-isfaction relation in the context of this cross-sectional model, wewere interested in investigating whether this mediation modelwould hold when investigating it using half-longitudinal tech-niques (i.e., using the paths in the regression of the T2 mediatorsonto T1 thin-ideal internalization controlling for T1 mediator val-ues and the paths in the regression of T2 body dissatisfaction ontothe T1 mediators controlling for T1 levels of body dissatisfaction

to estimate the specific indirect effects). However, before exam-ining whether social comparison and body surveillance mediatedthe relation between thin-ideal internalization and body dissatis-faction half-longitudinally, it was necessary to examine whether

E.E. Fitzsimmons-Craft et al. / Body Image 11 (2014) 488500 495

Fig. 3. Standardized path coefficients and factor loadings for the elaborated sociocultural model of disordered eating structural model. PSPS = Perceived SocioculturalPressure Scale. SATAQ-4 = Sociocultural Attitudes Toward Appearance Questionnaire-4. IBSS-R = Ideal-Body Stereotype Scale-Revised. BAAR = Beliefs About AttractivenessScale-Revised; we note that we are using the Importance of Being Attractive and Thin subscale. BEECOM = Body, Eating, and Exercise Comparison Orientation Mea-sure. OBCS = Objectified Body Consciousness Scale. BSQ = Body Shape Questionnaire. EDI-BD = Eating Disorder Inventory-Body Dissatisfaction. EDE-Q-6 = Eating DisorderE tudes

tplactaivbm.pgFtstd(b.

pdlab

xamination-Questionnaire-6. BULIT-R = Bulimia Test-Revised. EAT-26 = Eating Atti

hese constructs prospectively predicted one another and if theyrospectively predicted one another when controlling for baseline

evels of the outcome variable. A single analysis using latent vari-bles (using the same indicators as shown in Fig. 3 but at T2 in thease of social comparison and body surveillance) was conductedo assess the relationships between the thin-ideal internalizationt T1 and social comparison and body surveillance at T2. Resultsndicated that thin-ideal internalization at T1 predicted significantariance in both social comparison ( = .65, p < .001; R2 = .43) andody surveillance ( = .54, p < .001; R2 = .29) at T2. However, thisodel did not provide a good fit to the data (RMSEA: .131; SRMR:

076; CFI: .832; TLI: .789), and thus, it was unclear whether thesearameter estimates could be meaningfully interpreted. Next, a sin-le analysis using latent variables (using the indicators shown inig. 3 but at T2 in the case of body dissatisfaction) was conductedo assess the relationships between social comparison and bodyurveillance at T1 and body dissatisfaction at T2. Results indicatedhat T1 social comparison predicted unique variance in T2 bodyissatisfaction ( = .67, p < .001), while body surveillance did not = .01, p = .957). This model explained 45.2% of the variance in T2ody dissatisfaction and provided a modest fit to the data (RMSEA:

086; SRMR: .056; CFI: .929; TLI: .911).We then investigated whether these constructs prospectively

redicted one another when controlling for baseline levels of the

ependent variable. Examining these effects in a single model with

atent variables, results indicated that thin-ideal internalizationt T1 did not predict T2 social comparison ( = .01, p = .936) orody surveillance ( = .01, p = .952) after controlling for baseline

Test-26. **p < .01. ***p < .001.

levels of these constructs. Additionally, this model did not providea good fit to the data (RMSEA: .102; SRMR: .069; CFI: .814; TLI:.788). Likewise, social comparison and body surveillance at T1 didnot predict body dissatisfaction at T2 after controlling for baselinelevels ( = .03, p = .781; = .07, p = .361, respectively). This modelprovided a possibly acceptable fit to the data (RMSEA: .100; SRMR:.055; CFI: .911; TLI: .891). Given that these constructs were notfound to predict change in one another over the course of 3 months,it was not possible that social comparison and body surveillancewould mediate the thin-ideal internalization-body dissatisfactionrelation prospectively. Thus, a half-longitudinal mediation modelwas not investigated.

Examination of the prospective relations between socialcomparison/body surveillance and disordered eating. Finally, weinvestigated the relations between social comparison and bodysurveillance at T1 and disordered eating at T2 both without andwith controlling for baseline levels of disordered eating. A singleanalysis using latent variables was conducted to assess the rela-tionships between social comparison and body surveillance at T1and disordered eating at T2. Results indicated that T1 social com-parison predicted unique variance in T2 disordered eating ( = .74,p < .001), while body surveillance did not ( = .12, p = .328). Thismodel explained 43.3% of the variance in T2 disordered eating andprovided a modest fit to the data (RMSEA: .086; SRMR: .060; CFI:

.912; TLI: .890). A separate analysis revealed that social comparisonand body surveillance at T1 did not predict disordered eating at T2after controlling for baseline levels of disordered eating ( = .08,p = .577; = .14, p = .143, respectively). This model did not

496 E.E. Fitzsimmons-Craft et al. / BodyTa

ble

2C

orre

lati

ons

amon

gth

em

easu

red

vari

able

sat

T1(n

=22

6).

12

34

56

78

910

1112

1314

1516

17

1.PS

PS

2.SA

TAQ

-4,P

ress

ure

.82*

**

3.IB

SS-R

.17*

.26*

**

4.SA

TAQ

-4,I

nte

rnal

izat

ion

.40*

**.4

9***

.42*

**

5.B

AA

R,I

mp

orta

nce

ofB

ein

gA

ttra

ctiv

ean

dTh

in.3

2***

.36*

**.4

1***

.45*

**

6.B

EEC

OM

,Bod

y.4

9***

.52*

**.4

5***

.61*

**.4

2***

7.

BEE

CO

M,E

atin

g.4

4***

.50*

**.3

8***

.56*

**.4

2***

.73*

**

8.B

EEC

OM

,Exe

rcis

e.3

9***

.43*

**.4

1***

.52*

**.4

3***

.69*

**.6

9***

9.

BEE

CO

M,T

otal

.50*

**.5

5***

.47*

**.6

3***

.48*

**.9

0***

.90*

**.8

9***

10

.OB

CS,

Surv

eill

ance

.28*

**.3

0***

.34*

**.5

2***

.39*

**.6

5***

.47*

**.4

7***

.60*

**

11.B

SQ.6

1***

.66*

**.2

8***

.57*

**.4

6***

.70*

**.6

5***

.62*

**.7

4***

.54*

**

12.E

DI-

BD

.56*

**.6

1***

.20*

*.4

7***

.44*

**.6

0***

.53*

**.4

9***

.61*

**.4

5***

.85*

**

13.E

DE-

Q-6

,Wei

ght

Con

cern

/Sh

ape

Con

cern

.49*

**.5

7***

.26*

**.6

0***

.48*

**.6

7***

.61*

**.5

5***

.68*

**.5

2***

.89*

**.8

1***

14

.BU

LIT-

R.4

1***

.45*

**.2

5***

.46*

**.4

1***

.54*

**.5

6***

.49*

**.6

0***

.38*

**.7

4***

.58*

**.7

1***

15

.ED

E-Q

-6,R

estr

ain

t.3

0***

.36*

**.1

8*.4

2***

.30*

**.3

3***

.46*

**.4

0***

.45*

**.2

9***

.56*

**.4

2***

.61*

**.5

4***

16

.EA

T-26

.34*

**.3

9***

.27*

**.5

7***

.41*

**.5

0***

.54*

**.4

9***

.57*

**.4

4***

.62*

**.4

8***

.64*

**.6

5***

.56*

**

17.B

MI

.32*

**.4

0***

.18

**.0

2.1

0.2

0**

.22*

*.1

8**

.22*

*.0

9.4

8***

.47*

**.3

8***

.31*

**.1

4*.1

4

Not

e.PS

PS=

Perc

eive

dSo

cioc

ult

ura

lPr

essu

reSc

ale.

SATA

Q-4

=So

cioc

ult

ura

lA

ttit

ud

esTo

war

dA

pp

eara

nce

Qu

esti

onn

aire

-4.

IBSS

-R=

Idea

l-B

ody

Ster

eoty

pe

Scal

e-R

evis

ed.

BA

AR

=B

elie

fsA

bou

tA

ttra

ctiv

enes

sSc

ale-

Rev

ised

.B

EEC

OM

=B

ody,

Eati

ng,

and

Exer

cise

Com

par

ison

Ori

enta

tion

Scal

e.O

BC

S=

Obj

ecti

fied

Bod

yC

onsc

iou

snes

sSc

ale.

BSQ

=B

ody

Shap

eQ

ues

tion

nai

re.

EDI-

BD

=Ea

tin

gD

isor

der

Inve

nto

ry-B

ody

Dis

sati

sfac

tion

.ED

E-Q

-6=

Eati

ng

Dis

ord

erEx

amin

atio

n-Q

ues

tion

nai

re-6

.BU

LIT-

R=

Bu

lim

iaTe

st-R

evis

ed.E

AT-

26=

Eati

ng

Att

itu

des

Test

-26.

BM

I=B

ody

mas

sin

dex

.Var

iabl

esar

eco

nti

nu

ous,

wit

hh

igh

erva

lues

refl

ecti

ng

hig

her

leve

lsof

the

con

stru

ct.

*p