- Email: [email protected]
An evaluation of the Tripartite Influence Model of body dissatisfaction and eating disturbance with adolescent girls
Body Image 1 (2004) 237–251
An evaluation of the Tripartite Influence Model of body dissatisfaction and eating disturbance with adolescent girls Helene Keery a,1 , Patricia van den Berg b , J. Kevin Thompson b,∗ b
a Division of General Pediatrics and Adolescent Health, University of Minnesota, Minneapolis, MN, USA Department of Psychology, University of South Florida, 4202 E. Fowler Avenue, Tampa, FL 33620-8200, USA
Received 22 August 2003; received in revised form 8 March 2004; accepted 8 March 2004
Abstract The Tripartite Influence Model of body image and eating disturbance proposes that three formative influences (peer, parents, and media) affect body image and eating problems through two mediational mechanisms: internalization of the thin-ideal and appearance comparison processes. The current study evaluated this model in a sample of 325 sixth through eighth grade girls. Simple path analyses indicated that internalization and comparison fully mediated the relationship between parental influence and body dissatisfaction and partially mediated the relationship between peer influence and body dissatisfaction. Additionally, internalization and comparison partially mediated the relationship between media influence and body dissatisfaction. Six a priori SEM models based on the full Tripartite Influence Model were also evaluated. A resulting model was found to be an adequate fit to the data, supporting the viability of the Tripartite Model as a useful framework for understanding processes that may predispose young women to develop body image disturbances and eating dysfunction. © 2004 Elsevier B.V. All rights reserved. Keywords: Body image; Eating disturbance; Risk factors; Adolescence; Internalization; Tripartite Influence Model
Introduction Eating disorders occur at an alarming frequency in adolescent girls and are associated with high levels of emotional distress and physical problems (Thompson & Smolak, 2001). The prevalence of anorexia nervosa and bulimia nervosa in adolescence is similar to that ∗ Corresponding author. Tel.: +1-813-9740367; fax: +1-813-9744617. E-mail address: [email protected] (J.K. Thompson). 1 Present address: Eating Disorders Institute, Methodist Hospital, St. Louis Park, Minnesota, USA.
found in adults (i.e., approximately 0.5 and 1–3.0%, respectively) and subclinical levels of eating disturbance range as high as 10% of adolescent females (American Psychiatric Association, 2000; Thompson & Smolak, 2001). Accordingly, researchers in the past few years have invested an enormous amount of effort in attempts to define, measure, and evaluate formative influences that may lead to the onset of such disturbances (Shisslak & Crago, 2001; Stice, 2001; Stice & Hoffman, 2004; Wertheim, Paxton, & Blaney, 2004). Many variables have been hypothesized to play a role in the onset of body image problems and eating disturbances, including: negative affect, low self-
1740-1445/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.bodyim.2004.03.001
238
H. Keery et al. / Body Image 1 (2004) 237–251
esteem, media pressures, social comparison processes, peer influences, internalization of the societallybased thin-ideal, and parental pressures (Thompson, Heinberg, Altabe, & Tantleff-Dunn, 1999; Wertheim et al., 2004). In fact, investigations into the role of potential risk factors for body image and eating disturbances have proliferated in recent years, however, studies have often been limited by the lack of a theoretical model to guide selection of measures and testing of hypotheses. In an attempt to develop a model that would incorporate many of the variables hypothesized to have an effect on body image and eating disturbances, Thompson et al. (1999) developed the Tripartite Influence Model. This model is composed of three primary sources of influence (peers, parents, media) that are hypothesized to exert their effect on body image and eating disturbance via two primary mechanisms—appearance comparison and internalization of the thin-ideal. An additional aspect of the model includes a proposed directional link from the restricting component of eating disturbance to bulimia, and a link between bulimia and psychological functioning (e.g., self-esteem, depression). Fig. 1 (Model 1) details all of the hypothesized relations among variables predicted by this model, as originally outlined by Thompson et al. (1999). Work in our laboratory and that of other investigators has supported the role of specific peer, parental, and me-
dia factors as a predictor of body image and eating problems, using path models and prospective studies (Field, Camargo, Taylor, Berkey, & Colditz, 1999; Field, Camargo, Taylor, Berkey, Roberts, & Colditz, 2002; McKnight Investigators, 2003; Thompson & Stice, 2001). Additionally, the Tripartite Influence Model was supported in a covariance structure modeling study of college women (van den Berg, Thompson, Obremski-Brandon, & Coovert, 2002), however, only appearance comparison was evaluated as a mediator in this study. To date, the model has yet to be evaluated in an adolescent sample. Therefore, the current study sought to examine the Tripartite Influence Model as a potential explanatory framework for understanding formative factors that contribute to the development of body image and eating problems. Based on our previous work with adults, it was hypothesized that sociocultural influences (peers, parents, media) and mediational factors (internalization of the thin-ideal, appearance comparison) would predict body dissatisfaction and eating disturbances. Additionally, it was hypothesized that body dissatisfaction would have a direct influence on restriction, which would lead to bulimic symptoms. Several variations of the basic Tripartite Influence Model were also evaluated, along with a specific comparison to another prominent theoretical model (Stice, 2001), in order to provide for an evaluation of competing models.
Fig. 1. Hypothetical SEM Models 1 and 2 based on Thompson et al. (1999).
H. Keery et al. / Body Image 1 (2004) 237–251
Method Participants The initial sample consisted of 433 female adolescents in the sixth, seventh, and eighth grades who where attending a single middle school in Pasco County, Florida. Parental assent was obtained and no parents declined the participation of their children. All participants were tested on 1 day, as part of a science class project. The initial sample was reduced to 325 participants, primarily because of incomplete data on forms assessing parental influences (see “Data reduction and analysis” section). Their ages ranged from 11 to 15 years (M = 12.6 years, SD = 0.90 years). Thirty-six percent of the participants were in sixth grade, 36% were in seventh grade, and 27% were in eighth grade. Eighty-five percent of the participants identified themselves as Caucasian, 5% as Hispanic, 3% as African-American, 2% as native American, 1% were Asian, and 4% identified themselves as other. Measures Each of the sociocultural influence factors (peer, parents, media) might best be thought of as a multidimensional construct, consisting of several subcomponents. For instance, both peer and parental influence can be broken down into the components of criticism or teasing, modeling of dieting or body image concerns, and investment in thinness. Media influences might include information provided by the media related to appearance norms and/or dieting practices and internalization of media images and messages (i.e., thin-ideal internalization). In a previous study, Shroff and Thompson (2003) developed a comprehensive questionnaire for the assessment of peer, parent and media influences. Factor analysis of their measure yielded a three-factor solution consistent with the tripartite influence (peers, parents, media) and guided selection of measures and items for the current study. The specific scales for the three-component sociocultural influence construct are detailed further and sample items are contained in Table 1 (Table 1 also contains the factor loadings from the factor analysis performed on the peer, parent, and media items, which is discussed in “Results” section).
239
Parental influence The Perception of Teasing Scale-Weight Teasing Frequency Subscale (Thompson, Cattarin, Fowler, & Fisher, 1995) is a six-item scale that assesses general weight teasing. Schutz, Paxton, and Wertheim (2002) found that the scale has an alpha of 0.95 in a sample of 12–15-year-old girls. To examine parental influence, two items were adapted to index father and mother teasing (four total items). The alpha in the current sample was 0.75. The Perceived Family Preoccupation with Weight and Dieting Scale is a nine-item scale designed to assess the frequency of different behaviors that suggest a preoccupation with weight and dieting, such as talking about weight and shape, dieting behavior, and the importance of one another’s weight and shape among family members. The alpha of this measure in a sample of girls aged 12–16 years was 0.87 (Schutz et al., 2002). The alpha in the current sample was 0.86. Family Influence Scale is an adaptation of a threeitem scale developed by Levine, Smolak, and Hayden (1994). The three-item scale assesses participants’ perceptions of family influence on weight and appearance. The scale had an alpha of 0.80 in a sample of sixth to eighth grade girls (Levine et al., 1994). The items were modified to ask about both maternal and paternal influence (six total items). An alpha of 0.74 was obtained for the sample in the current study. The Parental Involvement Scale is a four-item measure that assesses the daughter’s view of her parents’ investment in her appearance (Levine et al., 1994). The scale had an alpha of 0.80 in a sample of sixth to eighth grade girls (Levine et al., 1994). The alpha for the current sample was 0.77. Peer influence The Perception of Teasing Scale-Weight Teasing Frequency Subscale (Thompson, Cattarin, et al., 1995) is a six-item scale that assesses general weight teasing. Schutz et al. (2002) found that the scale has an alpha of 0.95 in a sample of 12–15-year-old girls. To assess peer influence, two items were used and the words “friends, siblings and other boys and girls” replaced “people” in each item. The alpha for the current study was 0.73. The Perceived Friend Preoccupation with Weight and Dieting Scale is a nine-item scale designed to assess the frequency of different behaviors that suggest a
240
H. Keery et al. / Body Image 1 (2004) 237–251
Table 1 Factor loadings of peer, parent, and media items Item The magazines I read and TV shows I watch emphasize that it is important to be thin The magazines I read and the TV shows I watch emphasize the importance of appearance (shape, weight, clothing) The magazines I read and TV shows I watch emphasize dieting to lose weight I have felt pressure from the media to lose weight I would be interested in watching a new TV show if the topic is dieting I would be interested in watching a new TV show if the topic is fitness and exercise I would be interested in watching a new TV show if the topic is fashion I would be interested in reading a new magazine if the topic is dieting I would be interested in reading a new magazine if the topic is fitness and exercise I would be interested in reading a new magazine if the topic is fashion How concerned is your mother about whether you weigh too much or are too fat or might become too fat? How important is it to your mother that you be thin? How concerned is your father about whether you weigh too much or are too fat of might become too fat? How important is it to your father that you be thin? My father is on a diet to lose weight It is important to my father that he be as thin as possible My father’s physical appearance (shape, weight, clothing) is important to him My mother is on a diet to lose weight It is important to my mother that she be as thin as possible My mother’s physical appearance (shape, weight, clothing) is important to her Your father made comments or teased you about your appearance Your mother made comments or teased you about your appearance How often do your parents comment on each other’s weight? How often do your parents encourage each other to lose weight? How often do your parents talk about weight or dieting? How often do your parents worry about their weight? How often do your parents worry about what they eat? How often do your parents diet? Do you think your parents take a lot of notice of each other’s weight and shape? Are weight and shape important to your parents? One or more of my friends and classmates is on a diet to lose weight It is important to my friends and classmates that they be as thin as possible My friends and classmates’ physical appearance (shape, weight, clothing) is important to them Your friends and classmates made comments or teased you about your appearance How often do your friends and classmates comment on each other’s weight? How often do your friends and classmates encourage each other to lose weight? How often do your friends and classmates talk about weight or dieting? How often do your friends and classmates worry about their weight? How often do your friends and classmates worry about what they eat? How often do your friends and classmates diet? How often do your friends and classmates skip meals? Do you think that your friends and classmates take a lot of notice in each other’s weight and shape? Are weight and shape important to your friends and classmates?
Factor 1 Parent
Factor 2 Peer
Factor 3 Media
0.05
0.15
0.56
0.10
0.14
0.47
0.04 0.13 −0.04 −0.04
0.09 0.08 −0.06 −0.06
0.56 0.42 0.77 0.64
0.01 −0.02 −0.03
0.06 −0.09 −0.05
0.48 0.83 0.71
0.02 0.33
0.12 0.07
0.43 −0.01
0.38 0.35
−0.01 0.06
−0.00 −0.03
0.38 0.49 0.41 0.41 0.56 0.53 0.50 0.35 0.38 0.54 0.66 0.72 0.72 0.55 0.77 0.58
−0.06 −0.14 −0.02 0.03 0.00 0.05 0.10 0.02 0.01 0.08 −0.03 −0.03 −0.00 0.10 −0.02 0.08
0.03 0.00 0.02 0.03 0.02 0.07 0.02 0.06 −0.06 0.04 0.07 0.06 0.02 −0.12 −0.11 0.04
0.67 0.09 0.08 0.09
−0.09 0.49 0.49 0.47
0.04 0.05 0.13 0.14
0.01
0.34
−0.04
−0.03 0.00 −0.08 0.03 0.03 0.09 −0.05 −0.04
0.69 0.55 0.80 0.80 0.74 0.71 0.56 0.60
−0.16 −0.03 0.01 −0.03 0.02 −0.02 0.02 0.04
−0.10
0.64
0.16
H. Keery et al. / Body Image 1 (2004) 237–251
preoccupation with weight and dieting. These include talk about weight and shape, dieting behavior, and importance of one another’s weight and shape among friends. The alpha of this measure in a sample of girls aged 12–16 years was 0.87 (Schutz et al., 2002). The alpha for this sample was 0.89. Peer Influence Scale is an adaptation of a three-item scale developed by Levine et al. (1994). The scale assesses participants’ perceptions of peer influence on weight and appearance. Levine et al. found an alpha of 0.73 in sample of sixth to eighth grade girls. An alpha of 0.70 was obtained in the current sample. The McKnight Risk Factor Survey is a multidimensional measure used to assess risk factors for eating disturbance in fourth through twelfth grade girls (Shisslak et al., 1999). The two-item Weight Rejection Scale was used to assess girls’ interpersonal experiences related to rejection caused by their weight. The alpha for the current sample was 0.72. Media influence Media Influence Scale is an adaptation of a threeitem scale developed by Levine et al. (1994). The items were modified to assess participants’ recollection of media influence on weight and appearance. The alpha in the current sample was 0.83. The Interest Scale (Harrison, 2000) is a six-item scale that assesses a girl’s interest in watching television shows or reading magazines that cover topics such as dieting, fitness and exercise, and fashion. The alpha for the current sample was 0.82. Two items from the Perceived Sociocultural Pressure Scale (Stice, Ziemba, Margolis, & Flick, 1996) were used to assess perceived pressure from the media to lose weight. The alpha for this sample was 0.72. The media scales were preceded by a brief survey that asked participants to list their three favorite television shows, magazines, and movies. This was included in order to prime their recall of media images and messages in order to facilitate the responses to these items. This scale was not included in analyses. Mediational variables Internalization. The Sociocultural Internalization of Appearance Questionnaire–Adolescents (SIAQ-A) was used to assess internalization. This scale is a downward extension of the Internalization Subscale of the Sociocultural Attitudes Towards Appearance
241
Questionnaire-3 (Thompson, van den Berg, Roehrig, Guarda, & Heinberg, 2004). Internalization items capture an incorporation of the media images and messages of attractiveness into one’s own self-identity. A sample item is: Reading magazines makes me want to change my appearance. This five-item measure has been validated in samples from three countries and a high internal consistency (>0.80) has been found in six independent samples of 10–15 year-old girls (Keery, Shroff, Thompson, Wertheim, & Smolak, in press). The alpha for the current sample was 0.88. Appearance comparison. Two measures of social comparison were included, in order to provide for an index of general and specific appearance comparison. The Physical Appearance Comparison Scale (Thompson, Heinberg & Tantleff, 1991) is a five-item measure that assesses an overall tendency to compare one’s own appearance to the appearance of generic others in social situations (e.g., parties and social gatherings). In the current study, one of the items, which is the only reverse-coded item on the scale, had a negative correlation with the item total. Therefore, this item was deleted and the resulting alpha was 0.80. The Social Comparison Questionnaire (Schutz et al., 2002) is a 14-item scale that assesses the frequency of both appearance and more general comparison tendencies and the targets for social comparisons (i.e., friends, mother, father, fashion models). The items related to comparing oneself to different targets in terms of appearance were used in this study. The alpha for the current sample was 0.80. Outcome measures Body image. Two measures were used to index body dissatisfaction, in order to provide for a broader test of the mediational hypotheses in the simple paths. An attempt was made to use a measure that captured the specific dissatisfaction with body size and weight, along with a measure that assessed a more global aspect of appearance evaluation (Thompson, 2004). The Eating Disorder Inventory–Body Dissatisfaction Scale (EDI-BD; Garner, 1991) contains nine items that measure satisfaction with the size of specific body sites such as the waist, thighs, and hips. Shore and Porter (1990) found that the scale has an alpha of 0.91 for 11–18-year-old females. The alpha for the EDI-BD was 0.88 for the current sample. The Self-Image Ques-
242
H. Keery et al. / Body Image 1 (2004) 237–251
tionnaire for Young Adolescents–Body Image Subscale (SIQYA; Peterson, Schulenberg, Abramowitz, Offer, & Jarcho, 1984) is an 11-item subscale designed to assess broad, positive feelings toward the body in 10–15-year olds. The alpha for this sample was 0.79. Eating disturbance. The Eating Disorder Inventory– Drive for Thinness Scale (EDI-DT; Garner, 1991) measures restricting tendencies, desire to lose weight, and fear of weight gain. It has an internal consistency of 0.83 for a combined sample of eating disordered individuals and 0.81–0.91 for four samples of nonpatient female controls (Garner, 1991). Shore and Porter (1990) found an alpha of 0.81 for 11–18-yearold females. The alpha for the current sample was 0.91. The Eating Disorder Inventory–Bulimia Scale (EDI-B; Garner, 1991) assesses the tendency toward overeating, binging, and the impulse to engage in self-induced vomiting. Shore and Porter (1990) found an alpha of 0.69 for 11–18-year-old females. The alpha for the EDI-B was 0.83 for the current sample. Psychological functioning As with our previous work (e.g., van den Berg et al., 2002), we sought to provide a composite index of psychological functioning. Self-esteem and depression were included as indicators of psychological functioning (e.g., Thompson, Coovert, Richards, Johnson, & Cattarin, 1995) and, for the first time, we also included for provisional evaluation a measure of perfectionism, given work detailing its potential connection to body image and eating dysfunction (McKnight Investigators, 2003; Santonastaso, Friederici, & Favaro, 1999; Tyrka, Walden, Graber, & Brooks-Gunn, 2002). Self-esteem. The Rosenberg Self-Esteem Inventory is a 10-item measure of general feelings of self-worth (Rosenberg, 1965). The instrument has been shown to be reliable and valid for use with adolescent samples. For example, the alpha was 0.84 in a study of 12–16year-old girls (Schutz et al., 2002). The alpha for the current sample was 0.86. Depression. The Center for Epidemiologic Studies Depression Scale (CESD; Radloff, 1977) was used to
assess depressive symptoms. Items use a four point response format ranging from 0 = never to 4 = most of the time. The reliability of this scale has been well documented (Robert, Lewinsohn, & Seeley, 1991). The alpha for the current sample was 0.88. Perfectionism. The Eating Disorder Inventory–Perfectionism Scale (EDI-P; Garner, 1991) contains six items that measure expectations for achievement and attitudes associated with eating disturbance. Schutz et al. (2002) found an alpha of 0.83 in a sample of 12–16-year-old girls. The alpha for the EDI-P was 0.79 for the current sample. Body mass index Quetlet’s Index of Fatness (Garrow & Webster, 1985) was utilized as a measure of body mass index (BMI). This index is a ratio of weight (in kilograms) to squared height (in meters): W/H2 . Height and weight were obtained through self-report. Past research has examined the reliability of self-report measures of height and weight (Brooks-Gunn, Warren, Rosso, & Gargiulo, 1987). Findings indicate that although adolescents, particularly females, tend to underreport their weight by several pounds and their height by just over an inch, the correlations between actual and reported figures are relatively high (0.98 and 0.75) (Brooks-Gunn et al., 1987). Procedure This study was approved by the Institutional Review Board at the University of South Florida. Specific consent procedures were based on the policies of the Pasco County School District. One week prior to data collection, students were given a letter explaining the purpose of the study to take home to their parents. The letter stated that parents should contact the school if they did not want their child to participate. Participants watched a short video of the instructions for the survey on the closed circuit television system in which they were informed that they could withdraw from the study at any time without penalty. All individuals attending a single middle school in grades 6–8 on the testing day were able to participate in the study. Following survey administration, teachers read the students a short debriefing about the purpose of the study.
H. Keery et al. / Body Image 1 (2004) 237–251
Data reduction and analysis Of the original sample of 425, data from 108 were not included in the primary analyses. Most of the deleted participants did not answer several questions on one or more of the parental influence measures (in particular, the Parental Involvement Scale was not completed by 66 participants). Most likely, these children were from single parent homes, although this information was not formally obtained. Additionally, participants who were missing responses on 5% of the items (i.e., 8 from the total item pool of 143) were also deleted from analyses Finally, one case was deleted due to missing all data on the EDI-P Scale. This resulted in a final sample of 325 participants. In order to determine if there was any consistent difference between the retained versus discarded sample, t-tests were conducted on all variables other than the parental influence measures. There was only one significant effect that resulted from these 11 analyses (scores on the depression scale were significantly lower for the retained sample). In cases wherein fewer than 5% of the items were not answered, participants’ own data were used to create a mean to be imputed for individual scales. For the newly created composite variables from the factor analyses (i.e., peer, parent, media, comparison) the individual’s mean for the individual composite variable was imputed. On all other scales, missing data were replaced by the individual’s mean on that scale. This created a sample of 325 participants with no missing data points on any of the scales administered. However, 43 of these individuals were missing data on height or weight. Preliminary and planned analyses Internal consistencies (Cronbach’s alpha), means, and standard deviations were calculated for all the scales. Principal axis factor analysis using a Promax oblique rotation was performed to determine the factor structure of the sociocultural influence scales (e.g., peer, parental, media). This strategy was also used to ensure that the mediational variables (internalization, comparison) were distinct. Analysis of the scree plot, eigenvalues, theory, and ease of interpretation determined the number of factors retained. Standardized regression coefficients were used as factor loadings.
243
Three a priori criteria for item retention were used. First, items had to load ±0.30 on the primary factor. Second, items could not load more than ±0.20 on alternative factors. Third, a spread of ±0.20 had to be present between loadings on alternative factors. Simple path analyses were tested using the set of three regression equations as recommended by Baron and Kenny (1986). Structural equation modeling (SEM) was used to evaluate the full Tripartite Model (specific models tested are discussed later).
Results Factor analysis: sociocultural influence items This factor analysis examined the original 50 items included in all of the peer, parent, and media scales in the study. Examination of the scree plot indicated that three or four factors should be retained. Factor analyses were run for three and four factor solutions. Visual inspection of the scree plot, ease of interpretation, and theory indicated that the three-factor solution was the best fit. Of the original 50 items, 1 item on the media factor, 3 items on the parent factor and 3 items on the peer factor were deleted, primarily because they cross-loaded on other factors. A factor analysis on the remaining 43 items resulted in a 43-item scale with a three-factor solution. Inspection of the item content reveals that the three factors derived from this scale reflect the dimensions of peer, parent, and media influence. The 43-item scale includes 10 media items, 20 parent items, and 13 peer items. The internal consistency for each factor and scale total were: parent = 0.88; peer = 0.89; media = 0.86; total = 0.87. Table 1 contains the items and factor loadings. Factor analysis: internalization and comparison items This factor analysis examined the original 18 items included in the internalization and comparison scales. Examination of the scree plot indicated that two or three factors should be retained. Factor analyses were run for two- and three-factor solutions. Visual inspection of the scree plot, ease of interpretation, and theory indicated that the three-factor solution was the best fit. Of the original 18 items, 4 comparison items were deleted because they cross-loaded on the internaliza-
244
H. Keery et al. / Body Image 1 (2004) 237–251
tion factor. One item that cross-loaded on the two comparison factors was retained. A factor analysis on the remaining 14 items resulted in a 14-item scale with a three-factor solution. Inspection of the item content reveals that the three factors derived from this scale reflect the dimensions of internalization, peer comparison, and family comparison. The 14-item scale includes 5 internalization items, 6 peer comparison items, and 3 family comparison items. None of the original internalization items were deleted, therefore the entire SIAQ-A scale was used for the internalization variable. The internal consistency for each factor was: internalization = 0.88, peer comparison = 0.90, family comparison = 0.70. This procedure established that internalization and comparison were distinct constructs and could each be used as mediators. As such, the peer and family comparison factors were combined into one composite comparison variable for further analyses. The alpha for the composite comparison factor was 0.86. Correlational and mediational analyses Table 2 contains the intercorrelations among all measures. These correlations were used as the basis for conducting mediational analyses, discussed later. Additionally, an evaluation of the correlations was used to consider whether there was support for the inclusion of the perfectionism indicator in the full model as part of a psychological functioning latent variable (with self-esteem and depression). As Table 2 illustrates, the shared variance between perfectionism and self-esteem was 1.2% and the shared variance between perfectionism and depression was 4.4%. In contrast, the shared variance between self-esteem and depression was 32.5%, therefore perfectionism was not included in the path models tested later. Scores for the self-esteem and depression scale were transformed into z-scores and summed to provide a single indicator of psychological functioning. Internalization and comparison as mediators An evaluation of the correlation matrix reveals significant relationships among the influence variables (peer, parents, media) internalization, comparison,
and both measures of body dissatisfaction. Additionally, the pattern of correlations for both measures of body dissatisfaction and the sociocultural influence and mediational variables were very similar. This fact, along with the shared variance of 0.46 between the two SIQYA and EDI-BD supported the creation of a composite body dissatisfaction variable for further analyses. Simple path models were tested with internalization and comparison as mediators between the sociocultural influence variables (peer, parent, media) and composite body dissatisfaction using the method outlined by Baron and Kenny (1986). The results indicated that internalization and comparison fully mediated the relationship between parental influence and body dissatisfaction (see Fig. 2). Internalization and comparison partially mediated the relationship between peer influence and body dissatisfaction (see Fig. 3) and also partially mediated the relationship between media influence and body dissatisfaction (see Fig. 4). Because of the consistency of the findings in the simple paths indicating that each of the sociocultural influence measures had a significant effect, either partially or fully mediated, on body dissatisfaction, a composite influence variable (sum of peer, parents, and media) was used for subsequent SEM models. Use of a composite also has the advantage of limiting the number of paths, producing a more parsimonious model for SEM testing.
Fig. 2. Mediation of parental influence. The number in the parentheses is the standardized beta weight for the direct relationship between the influence and outcome measure. Evaluation of this number against the standardized beta weight from the regression equation allows for an interpretation of mediation. ∗∗ p < 0.01.
Table 2 Correlations among variables INT – 0.15∗∗ 0.48∗∗ 0.36∗∗ 0.71∗∗ 0.40∗∗ 0.40∗∗ 0.65∗∗ 0.54∗∗ 0.51∗∗ 0.49∗∗ 0.56∗∗ 0.59∗∗ 0.37∗∗ 0.44∗∗ 0.61∗∗ 0.02
RSE
CESD
MEDIA
PARENT
PEER
COMP
DISS
SIQYA
EDI-BD
EAT
EDI-DT
EDI-BU
PSYCH
Total SI
BMI
– 0.57∗∗ 0.35∗∗ 0.25∗ 0.26∗∗ 0.46∗∗ 0.66∗∗ 0.65∗∗ 0.55∗∗ 0.59∗∗ 0.55∗∗ 0.48∗∗ 0.79∗∗ 0.36∗∗ 0.14∗
– 0.23∗∗ 0.13∗ 0.28∗∗ 0.42∗∗ 0.48∗∗ 0.48∗∗ 0.38∗∗ 0.45∗∗ 0.41∗∗ 0.37∗∗ 0.95∗∗ 0.26∗∗ 0.17∗∗
– 0.44∗∗ 0.47∗∗ 0.56∗∗ 0.46∗∗ 0.43∗∗ 0.40∗∗ 0.55∗∗ 0.55∗∗ 0.40∗∗ 0.30∗∗ 0.75∗∗ 0.08
– 0.36∗∗ 0.34∗∗ 0.26∗∗ 0.25∗∗ 0.23∗∗ 0.43∗∗ 0.44∗∗ 0.31∗∗ 0.19∗∗ 0.83∗∗ 0.02
– 0.42∗∗ 0.36∗∗ 0.32∗∗ 0.34∗∗ 0.46∗∗ 0.47∗∗ 0.33∗∗ 0.30∗∗ 0.76∗∗ 0.14∗
– 0.58∗∗ 0.57∗∗ 0.49∗∗ 0.58∗∗ 0.54∗∗ 0.47∗∗ 0.48∗∗ 0.54∗∗ 0.08
– 0.93∗∗ 0.90∗∗ 0.69∗∗ 0.65∗∗ 0.55∗∗ 0.60∗∗ 0.44∗∗ 0.31∗∗
– 0.68∗∗ 0.65∗∗ 0.59∗∗ 0.55∗∗ 0.60∗∗ 0.41∗∗ 0.21∗∗
– 0.61∗∗ 0.61∗∗ 0.45∗∗ 0.49∗∗ 0.39∗∗ 0.39∗∗
0.61∗∗ 0.45∗∗ 0.49∗∗ 0.39∗∗ 0.39∗∗
– 0.55∗∗ 0.51∗∗ 0.60∗∗ 0.17∗∗
– 0.45∗∗ 0.43∗∗ 0.10∗∗
– 0.33∗∗ 0.17∗∗
– 0.09
–
–
0.11∗ 0.21∗∗ 0.19∗∗ 0.16∗∗ 0.21∗∗ 0.24∗∗ 0.17∗∗ 0.21∗∗ 0.09 0.30∗∗ 0.27∗∗ 0.27∗∗ 0.20∗∗ 0.24∗∗ −0.01
Note: INT = internalization; EDI-P = Eating Disorder Inventory—Perfectionism; RSE = self-esteem; CESD = Center for Epidemiologic Studies Depression Scale; MEDIA = media influence; PARENT = parent influence; PEER = peer influence; COMP = comparison; DISS = composite body dissatisfaction; SIQYA = Self-Image Questionnaire for Young Adolescents; EDI-BD = Eating Disorder Inventory—Body Dissatisfaction Scale; EDI-DT = Eating Disorder Inventory—Drive for Thinness Scale; EDI-BU = Eating Disorder Inventory—Bulimia Scale; PSYCH = total depression and self-esteem; Total SI = total sociocultural influence (peer, parent, media influence); BMI = body mass index. ∗ p < 0.05. ∗∗ p < 0.01.
H. Keery et al. / Body Image 1 (2004) 237–251
INT EDI-P RSE CESD MEDIA PARENT PEER COMP DISS SIQYA EDI-BD EAT EDI-DT EDI-BU PSYCH Total SI BMI
EDI-P
245
246
H. Keery et al. / Body Image 1 (2004) 237–251
Fig. 3. Mediation of peer influence. The number in the parentheses is the standardized beta weight for the direct relationship between the influence and outcome measure. Evaluation of this number against the standardized beta weight from the regression equation allows for an interpretation of mediation. ∗ p < 0.05, ∗∗ p < 0.01.
Fig. 4. Mediation of media influence. The number in the parentheses is the standardized beta weight for the direct relationship between the influence and outcome measure. Evaluation of this number against the standardized beta weight from the regression equation allows for an interpretation of mediation. ∗ p < 0.05, ∗∗ p < 0.01.
SEM models Initially, six models were hypothesized to potentially fit the data (see Figs. 1, 5 and 6). The first four models are based on the primary and secondary models initially presented by Thompson et al. (1999). In Model 1, sociocultural influence has a directional influence on internalization and comparison. Internalization and comparison in turn are hypothesized to influence body dissatisfaction, which has a directional influence on restriction and bulimia. There is also a path from bulimia to psychological functioning. In Model 2, a path from sociocultural influences to body dissatisfaction is added to Model 1 (see Fig. 1).
In Model 3, global psychological functioning is hypothesized to mediate the relationship between sociocultural influence and internalization and between sociocultural influence and comparison. All other paths are identical to Model 1. Model 4 adds the path from sociocultural influences to body dissatisfaction to Model 3 (see Fig. 5). Models 3 and 4 are essentially the secondary model originally hypothesized by Thompson et al. (1999) and also evaluated by van den Berg et al. (2002) in a sample of adults. Models 5 and 6 are slight variations of Models 3 and 4. In Model 5, psychological functioning is moved alongside sociocultural influences and conceptualized
Fig. 5. Hypothetical SEM Models 3 and 4 based on Thompson et al. (1999).
H. Keery et al. / Body Image 1 (2004) 237–251
247
Fig. 6. Hypothetical SEM Models 5 and 6 based on Thompson et al. (1999).
as a second formative influence on comparison and internalization. The covariance between sociocultural influences and global psychological functioning is also included in this model. The remaining paths are the same as those in Models 1–4. In Model 6, direct paths from sociocultural influences and psychological functioning to body dissatisfaction are added to Model 5 (see Fig. 6). An SEM evaluation of all six hypothetical models demonstrated poor fit overall. See Table 3 for fit indices. Given that Models 1 and 2 had lower RMSEA values than Models 3–6, the latter models were not further tested. A chi-square difference test revealed that there was no significant difference in fit between Models 1 and 2 (χ2 = 2.12, d.f. = 1, ns). Therefore, Model 1was chosen as the best fitting model be-
cause it had the lowest RMSEA. However, given the model’s poor fit, a specification search was undertaken. Examination of Wald tests and t-values indicated that all paths in the model were significant and none should be removed. The Lagrange Multiplier test indicated that adding several paths would improve the fit of the model. Based on data and theory, the path from sociocultural influence to restriction was freed. This modification to the model (Model 1A) resulted in a χ2 (11) = 126.35, which remained significant (see Table 2). In addition, the RMSEA was lowered to 0.18 and the CFI improved to 0.89. Following this modification, the Lagrange Multiplier test continued to indicate that the addition of paths would improve the fit of the model. Based on theory and data, the path from body dissatisfaction to global psychological function-
Table 3 Fit indices of six proposed models and modifications to Model 1
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 1–influence to restriction Model 1–body dissatisfaction to global psychological functioning Model 1–internalization to restriction Stice Model
χ2
df
p
RMSEA
CFI
ECVI
206.61 204.49 314.34 312.21 177.22 175.10 126.35 47.25 35.96 126.70
12 11 12 11 9 8 11 10 9 7
<0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
0.2237 0.2330 0.2789 0.2907 0.2402 0.2539 0.1799 0.1072 0.0962 0.2315
0.82 0.82 0.72 0.72 0.85 0.85 0.89 0.97 0.97 0.86
0.74 0.74 1.07 1.07 0.67 0.67 0.50 0.26 0.23 0.48
248
H. Keery et al. / Body Image 1 (2004) 237–251
Fig. 7. Final model with standardized path coefficients.
ing was added (Model 1B). This modification resulted in a significant χ2 (10) = 47.25, RMSEA = 0.11, and CFI = 0.97. One final path, from internalization to restriction, was freed based on results of the Lagrange Multiplier test and theory. The addition of this path resulted in a χ2 (9) = 35.96, RMSEA = 0.096, and CFI = 0.94. At this point in the analysis the model (Fig. 7) had achieved acceptable fit (Li-tze & Bentler, 1999), and no further theoretically justified paths were recommended by the Lagrange Multiplier test. One method of determining the usefulness of a relatively new theoretical model is to compare it to extant models. Stice’s (2001) Dual Pathway Model has perhaps received the most extensive evaluation in recent years. This model proposes that two formative influences (pressure to be thin; thin-ideal internalization) lead to body dissatisfaction, which then leads to both dieting and negative affect. Consequently negative affect and dieting are thought to produce bulimic symptoms. Stice (2001) compared this model to the Tripartite Influence model, finding that his model was a better fit to the data in a sample of adolescent girls. Our test of the Dual Pathway Model suggested that it was not as good a fit to the data as the Tripartite Influence Model for the current sample (see Table 2). The RMSEA was 0.23 and CFI was 0.86, indicating a poor fit to the data.2 2 For this analysis, we used our composite influence measure as the indicator for Stice’s pressures variable. In Stice (2001), he found that the data were a better fit to his Dual Pathway Model
Discussion The goal of the current study was to evaluate the Tripartite Influence Model in a sample of adolescent girls. Simple path analyses indicated that internalization and appearance comparison fully mediated the relationship between parental influence and body dissatisfaction. Internalization and comparison also partially mediated the relationships between peer influence and body dissatisfaction and between media influence and body dissatisfaction. The results of the mediational analyses are similar to results from a similar study with pre-adolescent girls (Shroff & Thompson, 2003). The SEM analysis of the Tripartite Influence Model and several variations on the basic model revealed that the original model (Thompson et al., 1999), with a few minor modifications, was an adequate fit to the data. One of the most interesting findings, and perhaps one that will lead to a revision of the Tripartite Influence Model, is the direct path from the sociocultural influences to restriction. This variable than the Tripartite Influence Model. However, it is unclear how he could have tested the full Tripartite Influence Model because there was no indicator of appearance comparison used in that study. Additionally, the measure of internalization used in Stice (2001) has been shown to index not internalization, but awareness of media images and messages (Thompson et al., 2004). Additionally, there were other differences in the measures used in the two studies. Certainly, any demonstration of the differential predictive utility of competing models requires replication.
H. Keery et al. / Body Image 1 (2004) 237–251
had indirect effects via internalization, comparison, and body dissatisfaction on restriction, yet there was also a direct effect. This finding replicates our work with an adult sample (van den Berg et al., 2002) and was also recently replicated in an adolescent sample in our laboratory (Shroff & Thompson, 2003). In effect, the power of these influences appears to be one of directly leading young girls to begin restriction, yet also setting the stage for restriction (possibly later) by enhancing internalization of societal messages, comparison, and body dissatisfaction. The original Tripartite Influence Model did not include BMI as a factor, therefore this variable was not included herein for model testing. An evaluation of the correlation between BMI and our influence and mediational variables indicates very little relationship (see Table 2). The only significant r was between BMI and peer influence (0.14). However, as might be expected BMI was significantly associated with body dissatisfaction (0.31), therefore, future modifications of the Tripartite Influence Model might include adding BMI, possibly as an exogenous predictor of body dissatisfaction. There are several limitations to the current study. First, the sample consisted of primarily Caucasian girls from a middle-class socioeconomic background. Second, at least 15% of the original sample were dropped because of missing data on parental measures, therefore, individuals who may have been from single parent homes were not included in analyses. Both of these sampling issues limit the generalizability of the current study. Third, the final model required modifications to achieve adequate fit. Although these modifications were theoretically plausible, they may be unique to the current sample, although as noted earlier, the current SEM model was replicated by Shroff and Thompson (2003). Another limitation of this study is its use of crosssectional data to model a developmental process. A prospective design would allow for a greater understanding of the development of body dissatisfaction and eating disturbance over time and is essential for the evaluation of causality and the determination of whether a specific variable is a causal risk factor (e.g., Field, 2004; Kraemer, Stice, Kazdin, Offord, & Kupfer, 2001; Stice & Hoffman, 2004). Such a design would also allow for greater confidence in directional hypotheses proposed by the Tripartite Influence
249
Model. For instance, the current model suggests that comparison leads to internalization rather than the converse, however, because both variables are mediators in the current model, the design is an equivalent model, which means that the models cannot be distinguished mathematically (e.g., MacCallum, Wegener, Uchino, & Fabrigar, 1993).3 Despite the limitations of the current study, the results suggest that interventions that address issues such as appearance related teasing and the modeling of unhealthy diet and exercise attitudes and behaviors by family members and peers may be useful in both prevention and treatment. The results of this study also highlight the need for continued work in the area of prevention interventions that teach young girls to be critical consumers of media. Interventions aimed at reducing the internalization of the sociocultural ideal of thinness and challenging girls to resist comparing themselves to others may help to interrupt the development of body image and eating disturbance (Stice & Hoffman, 2004).
References American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders—IV text revision. Washington, DC: American Psychiatric Association.
3 An argument can be made to support either comparison leading to internalization or the converse. Susan Paxton pointed out to us that Festinger’s (1954) original comparison theory framework supposed that comparison would only occur for an attribute of high importance to the individual, therefore, perhaps internalization of societal images and messages might precede elevations in appearance comparison tendencies. However, our view is that sociocultural influences spark a search for information regarding one’s own appearance in social and interpersonal contexts, driving an initial comparison process as a means of seeking information and clarification. If such a comparison process results in discouraging feedback (e.g., negative comparison experiences exacerbate appearance concern), the end result is a deeper and more nefarious modification of one’s self-view, leading to a more accepting belief in the veracity of the sociocultural and comparison information/feedback. Such a process eventuates in a more basic modification of one’s belief system, an internalization of the thin-ideal, leading the individual to seek appearance modification through multiple means, such as dieting, fashion, make-up, cosmetic surgery, etc. Of course, prospective data will ultimately clarify the directionality question.
250
H. Keery et al. / Body Image 1 (2004) 237–251
Baron, R. M., & Kenny, D. A. (1986). The moderatormediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Brooks-Gunn, J., Warren, M. P., Rosso, J., & Gargiulo, J. (1987). Validity of self-report measures of girls’ pubertal status. Child Development, 58, 829–841. Festinger, L. (1954). A theory of social comparison processes. Human Relations, 1, 117–140. Field, A. E. (2004). Risk factors for eating disorders: A evaluation of the evidence. In J. K. Thompson (Ed.), Handbook of eating disorders and obesity (pp. 17–32). New York: Wiley. Field, A. E., Camargo, C. A., Taylor, C. B., Berkey, C. S., & Colditz, G. A. (1999). Relation of peer and media influences to the development of purging behaviors among preadolescent and adolescent girls. Archives of Pediatrics and Adolescent Medicine, 153, 1184–1189. Field, A. E., Camargo, C. A., Taylor, C. B., Berkey, C. S., Roberts, S. B., & Colditz, G. A. (2001). Peer, parent, and media influences on the development of weight concerns and frequent dieting among preadolescent and adolescent girls and boys. Pediatrics, 107(1), 54–60. Garner, D. M. (1991). Eating disorders inventory-2: Professional manual. Odessa, FL: Psychological Assessment Resources. Garrow, J. S., & Webster, J. (1985). Quetlet’s index (W/H2) as a measure of fatness. International Journal of Obesity, 9, 147– 153. Harrison, K. (2000). The body electric: Thin-ideal media and eating disorders in adolescence. Journal of Communication, 50(3), 119–143. Keery, H., Shroff, H., Thompson, J. K., Wertheim, E., & Smolak, L. (in press). The Sociocultural Internalization of Appearance Questionnaire—Adolescents (SIAQ-A): Psychometric analysis and normative data for three countries. Eating and Weight Disorders: Studies on Anorexia, Bulimia and Obesity. Kraemer, H. C., Stice, E., Kazdin, A., Offord, D., & Kupfer, D. (2001). How do risk factors work together? Mediators, moderators, and independent, overlapping and proxy risk factors. American Journal of Psychiatry, 158, 848– 856. Levine, M. P., Smolak, L., & Hayden, H. (1994). The relation of sociocultural factors to eating attitudes and behaviors among middle school girls. Journal of Early Adolescence, 14, 471– 490. Li-tze, H., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis. Structural Equation Modeling, 6, 1–55. MacCallum, R. C., Wegener, D. T., Uchino, B. N., & Fabrigar, L. R. (1993). The problem of equivalent models in applications of covariance structure analysis. Psychological Bulletin, 114, 185–199. McKnight Investigators. (2003). Risk factors for the onset of eating disorders in adolescent girls: Results of the McKnight longitudinal risk factor study. American Journal of Psychiatry, 160, 248–254. Peterson, A. C., Schulenberg, J. E., Abramowitz, R. H., Offer, D., & Jarcho, H. D. (1984). A Self-Image Questionnaire for
Young Adolescents (SIQYA): Reliability and validity studies. Journal of Youth and Adolescence, 13, 93–111. Radloff, L. S. (1977). A CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385–401. Robert, R. E., Lewinsohn, P. M., & Seeley, J. R. (1991). Screening for adolescent depression: A comparison of depression scales. Journal of the American Academy of Child and Adolescent Psychiatry, 30, 58–66. Rosenberg, M. (1965). Measurement of self-esteem. In M. Rosenberg (Ed.), Society and the adolescent self-image (pp. 297–307). New York: Princeton University Press. Santonastaso, P., Friederici, S., & Favaro, A. (1999). Full and partial syndromes in eating disorders: A 1-year prospective study of risk factors among female students. Psychopathology, 32, 50–56. Schutz, H. K., Paxton, S. J., & Wertheim, E. H. (2002). Investigation of body comparison among adolescent girls. Journal of Applied Social Psychology, 32(9), 1906–1937. Shisslak, C. M., & Crago, M. (2001). Risk and protective factors in the development of eating disorders. In J. K. Thompson, & L. Smolak (Eds.), Body image, eating disorders, and obesity in youth: Assessment, prevention, and treatment (pp. 103–125). Washington, DC: American Psychological Association. Shisslak, C. M., Renger, R., Sharpe, T., Crago, M., McKnight, K. M., & Gray, N. et al., (1999). Development and evaluation of the McKnight Risk Factor Survey for assessing potential risk and protective factors for disordered eating in preadolescent and adolescent girls. International Journal of Eating Disorders, 25, 195–214. Shore, R. A., & Porter, J. E. (1990). Normative and reliability data for 11 to 18 year olds on the Eating Disorder Inventory. International Journal of Eating Disorders, 9, 201–207. Shroff, H., & Thompson, J. K. (2003). An evaluation of the Tripartite Influence Model with adolescent girls. Unpublished manuscript. Stice, E. (2001). A prospective test of the dual-pathway model of bulimic pathology: Mediating effects of dieting and negative affect. Journal of Abnormal Psychology, 110, 124–135. Stice, E., & Hoffman, E., 2004. Eating disorder prevention programs. In J. K. Thompson (Ed.), Handbook of eating disorders and obesity (pp. 33–57). New York: Wiley. Stice, E., Ziemba, C., Margolis, J., & Flick, P. (1996). The Dual Pathway Model differentiates bulimics, subclinical bulimics, and controls: Testing the continuity hypothesis. Behavior Therapy, 27, 531–549. Thompson, J. K. (2004). The (mis)measurement of body image: Ten strategies to improve assessment for applied and research purposes. Body Image: An International Journal of Research, 1, 7–14. Thompson, J. K., Cattarin, J., Fowler, B., & Fisher, E. (1995). The Perception of Teasing Scale (POTS): A revision and extension of the Physical Appearance Related Teasing Scale (PARTS). Journal of Personality Assessment, 65, 146–157. Thompson, J. K., Coovert, M. D., Richards, K. J., Johnson, S., & Cattarin, J. (1995). Development of body image, eating disturbance, and general psychological functioning in female
H. Keery et al. / Body Image 1 (2004) 237–251 adolescents: Covariance structure modeling and longitudinal investigations. International Journal of Eating Disorders, 18, 221–236. Thompson, J. K., Heinberg, L. J., Altabe, M., & Tantleff-Dunn, S. (1999). Exacting beauty: Theory, assessment, and treatment of body image disturbance. Washington, DC: American Psychological Association. Thompson, J. K., Heinberg, L. J., & Tantleff, S. (1991). The Physical Appearance Comparison Scale (PACS). The Behavior Therapist, 14, 174. Thompson, J. K., & Smolak, L. (Eds.). (2001). Body image, eating disorders, and obesity in youth: Assessment, prevention, and treatment. Washington, DC: American Psychological Association. Thompson, J. K., & Stice, E. (2001). Thin-ideal internalization: Mounting evidence for a new risk factor for body image disturbance and eating pathology. Current Directions in Psychological Science, 10(5), 181–183.
251
Thompson, J. K., van den Berg, P., Roehrig, M., Guarda, A., & Heinberg, L. J. (2004). The Sociocultural Attitudes Towards Appearance Questionnaire-3. International Journal of Eating Disorders, 35, 293–304. Tyrka, A. R., Waldron, I., Graber, J. A., & Brooks-Gunn, J. (2002). Prospective predictors of the onset of anorexic and bulimic syndromes. International Journal of Eating Disorders, 32, 282–290. van den Berg, P., Thompson, J. K., Obremski-Brandon, K., & Coovert, M. (2002). The Tripartite Influence Model of body image and eating disturbance: A covariance structure modeling investigation testing the meditational role of appearance comparison. Journal of Psychosomatic Research, 53(5), 1007– 1020. Wertheim, E. H., Paxton, S. J., & Blaney, S. (2004). Risk factors for the development of body image disturbance. In J. K. Thompson (Ed.), Handbook of eating disorders and obesity (pp. 463–494). New York: Wiley.
An evaluation of the Tripartite Influence Model of body dissatisfaction and eating disturbance with adolescent girls Helene Keery a,1 , Patricia van den Berg b , J. Kevin Thompson b,∗ b
a Division of General Pediatrics and Adolescent Health, University of Minnesota, Minneapolis, MN, USA Department of Psychology, University of South Florida, 4202 E. Fowler Avenue, Tampa, FL 33620-8200, USA
Received 22 August 2003; received in revised form 8 March 2004; accepted 8 March 2004
Abstract The Tripartite Influence Model of body image and eating disturbance proposes that three formative influences (peer, parents, and media) affect body image and eating problems through two mediational mechanisms: internalization of the thin-ideal and appearance comparison processes. The current study evaluated this model in a sample of 325 sixth through eighth grade girls. Simple path analyses indicated that internalization and comparison fully mediated the relationship between parental influence and body dissatisfaction and partially mediated the relationship between peer influence and body dissatisfaction. Additionally, internalization and comparison partially mediated the relationship between media influence and body dissatisfaction. Six a priori SEM models based on the full Tripartite Influence Model were also evaluated. A resulting model was found to be an adequate fit to the data, supporting the viability of the Tripartite Model as a useful framework for understanding processes that may predispose young women to develop body image disturbances and eating dysfunction. © 2004 Elsevier B.V. All rights reserved. Keywords: Body image; Eating disturbance; Risk factors; Adolescence; Internalization; Tripartite Influence Model
Introduction Eating disorders occur at an alarming frequency in adolescent girls and are associated with high levels of emotional distress and physical problems (Thompson & Smolak, 2001). The prevalence of anorexia nervosa and bulimia nervosa in adolescence is similar to that ∗ Corresponding author. Tel.: +1-813-9740367; fax: +1-813-9744617. E-mail address: [email protected] (J.K. Thompson). 1 Present address: Eating Disorders Institute, Methodist Hospital, St. Louis Park, Minnesota, USA.
found in adults (i.e., approximately 0.5 and 1–3.0%, respectively) and subclinical levels of eating disturbance range as high as 10% of adolescent females (American Psychiatric Association, 2000; Thompson & Smolak, 2001). Accordingly, researchers in the past few years have invested an enormous amount of effort in attempts to define, measure, and evaluate formative influences that may lead to the onset of such disturbances (Shisslak & Crago, 2001; Stice, 2001; Stice & Hoffman, 2004; Wertheim, Paxton, & Blaney, 2004). Many variables have been hypothesized to play a role in the onset of body image problems and eating disturbances, including: negative affect, low self-
1740-1445/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.bodyim.2004.03.001
238
H. Keery et al. / Body Image 1 (2004) 237–251
esteem, media pressures, social comparison processes, peer influences, internalization of the societallybased thin-ideal, and parental pressures (Thompson, Heinberg, Altabe, & Tantleff-Dunn, 1999; Wertheim et al., 2004). In fact, investigations into the role of potential risk factors for body image and eating disturbances have proliferated in recent years, however, studies have often been limited by the lack of a theoretical model to guide selection of measures and testing of hypotheses. In an attempt to develop a model that would incorporate many of the variables hypothesized to have an effect on body image and eating disturbances, Thompson et al. (1999) developed the Tripartite Influence Model. This model is composed of three primary sources of influence (peers, parents, media) that are hypothesized to exert their effect on body image and eating disturbance via two primary mechanisms—appearance comparison and internalization of the thin-ideal. An additional aspect of the model includes a proposed directional link from the restricting component of eating disturbance to bulimia, and a link between bulimia and psychological functioning (e.g., self-esteem, depression). Fig. 1 (Model 1) details all of the hypothesized relations among variables predicted by this model, as originally outlined by Thompson et al. (1999). Work in our laboratory and that of other investigators has supported the role of specific peer, parental, and me-
dia factors as a predictor of body image and eating problems, using path models and prospective studies (Field, Camargo, Taylor, Berkey, & Colditz, 1999; Field, Camargo, Taylor, Berkey, Roberts, & Colditz, 2002; McKnight Investigators, 2003; Thompson & Stice, 2001). Additionally, the Tripartite Influence Model was supported in a covariance structure modeling study of college women (van den Berg, Thompson, Obremski-Brandon, & Coovert, 2002), however, only appearance comparison was evaluated as a mediator in this study. To date, the model has yet to be evaluated in an adolescent sample. Therefore, the current study sought to examine the Tripartite Influence Model as a potential explanatory framework for understanding formative factors that contribute to the development of body image and eating problems. Based on our previous work with adults, it was hypothesized that sociocultural influences (peers, parents, media) and mediational factors (internalization of the thin-ideal, appearance comparison) would predict body dissatisfaction and eating disturbances. Additionally, it was hypothesized that body dissatisfaction would have a direct influence on restriction, which would lead to bulimic symptoms. Several variations of the basic Tripartite Influence Model were also evaluated, along with a specific comparison to another prominent theoretical model (Stice, 2001), in order to provide for an evaluation of competing models.
Fig. 1. Hypothetical SEM Models 1 and 2 based on Thompson et al. (1999).
H. Keery et al. / Body Image 1 (2004) 237–251
Method Participants The initial sample consisted of 433 female adolescents in the sixth, seventh, and eighth grades who where attending a single middle school in Pasco County, Florida. Parental assent was obtained and no parents declined the participation of their children. All participants were tested on 1 day, as part of a science class project. The initial sample was reduced to 325 participants, primarily because of incomplete data on forms assessing parental influences (see “Data reduction and analysis” section). Their ages ranged from 11 to 15 years (M = 12.6 years, SD = 0.90 years). Thirty-six percent of the participants were in sixth grade, 36% were in seventh grade, and 27% were in eighth grade. Eighty-five percent of the participants identified themselves as Caucasian, 5% as Hispanic, 3% as African-American, 2% as native American, 1% were Asian, and 4% identified themselves as other. Measures Each of the sociocultural influence factors (peer, parents, media) might best be thought of as a multidimensional construct, consisting of several subcomponents. For instance, both peer and parental influence can be broken down into the components of criticism or teasing, modeling of dieting or body image concerns, and investment in thinness. Media influences might include information provided by the media related to appearance norms and/or dieting practices and internalization of media images and messages (i.e., thin-ideal internalization). In a previous study, Shroff and Thompson (2003) developed a comprehensive questionnaire for the assessment of peer, parent and media influences. Factor analysis of their measure yielded a three-factor solution consistent with the tripartite influence (peers, parents, media) and guided selection of measures and items for the current study. The specific scales for the three-component sociocultural influence construct are detailed further and sample items are contained in Table 1 (Table 1 also contains the factor loadings from the factor analysis performed on the peer, parent, and media items, which is discussed in “Results” section).
239
Parental influence The Perception of Teasing Scale-Weight Teasing Frequency Subscale (Thompson, Cattarin, Fowler, & Fisher, 1995) is a six-item scale that assesses general weight teasing. Schutz, Paxton, and Wertheim (2002) found that the scale has an alpha of 0.95 in a sample of 12–15-year-old girls. To examine parental influence, two items were adapted to index father and mother teasing (four total items). The alpha in the current sample was 0.75. The Perceived Family Preoccupation with Weight and Dieting Scale is a nine-item scale designed to assess the frequency of different behaviors that suggest a preoccupation with weight and dieting, such as talking about weight and shape, dieting behavior, and the importance of one another’s weight and shape among family members. The alpha of this measure in a sample of girls aged 12–16 years was 0.87 (Schutz et al., 2002). The alpha in the current sample was 0.86. Family Influence Scale is an adaptation of a threeitem scale developed by Levine, Smolak, and Hayden (1994). The three-item scale assesses participants’ perceptions of family influence on weight and appearance. The scale had an alpha of 0.80 in a sample of sixth to eighth grade girls (Levine et al., 1994). The items were modified to ask about both maternal and paternal influence (six total items). An alpha of 0.74 was obtained for the sample in the current study. The Parental Involvement Scale is a four-item measure that assesses the daughter’s view of her parents’ investment in her appearance (Levine et al., 1994). The scale had an alpha of 0.80 in a sample of sixth to eighth grade girls (Levine et al., 1994). The alpha for the current sample was 0.77. Peer influence The Perception of Teasing Scale-Weight Teasing Frequency Subscale (Thompson, Cattarin, et al., 1995) is a six-item scale that assesses general weight teasing. Schutz et al. (2002) found that the scale has an alpha of 0.95 in a sample of 12–15-year-old girls. To assess peer influence, two items were used and the words “friends, siblings and other boys and girls” replaced “people” in each item. The alpha for the current study was 0.73. The Perceived Friend Preoccupation with Weight and Dieting Scale is a nine-item scale designed to assess the frequency of different behaviors that suggest a
240
H. Keery et al. / Body Image 1 (2004) 237–251
Table 1 Factor loadings of peer, parent, and media items Item The magazines I read and TV shows I watch emphasize that it is important to be thin The magazines I read and the TV shows I watch emphasize the importance of appearance (shape, weight, clothing) The magazines I read and TV shows I watch emphasize dieting to lose weight I have felt pressure from the media to lose weight I would be interested in watching a new TV show if the topic is dieting I would be interested in watching a new TV show if the topic is fitness and exercise I would be interested in watching a new TV show if the topic is fashion I would be interested in reading a new magazine if the topic is dieting I would be interested in reading a new magazine if the topic is fitness and exercise I would be interested in reading a new magazine if the topic is fashion How concerned is your mother about whether you weigh too much or are too fat or might become too fat? How important is it to your mother that you be thin? How concerned is your father about whether you weigh too much or are too fat of might become too fat? How important is it to your father that you be thin? My father is on a diet to lose weight It is important to my father that he be as thin as possible My father’s physical appearance (shape, weight, clothing) is important to him My mother is on a diet to lose weight It is important to my mother that she be as thin as possible My mother’s physical appearance (shape, weight, clothing) is important to her Your father made comments or teased you about your appearance Your mother made comments or teased you about your appearance How often do your parents comment on each other’s weight? How often do your parents encourage each other to lose weight? How often do your parents talk about weight or dieting? How often do your parents worry about their weight? How often do your parents worry about what they eat? How often do your parents diet? Do you think your parents take a lot of notice of each other’s weight and shape? Are weight and shape important to your parents? One or more of my friends and classmates is on a diet to lose weight It is important to my friends and classmates that they be as thin as possible My friends and classmates’ physical appearance (shape, weight, clothing) is important to them Your friends and classmates made comments or teased you about your appearance How often do your friends and classmates comment on each other’s weight? How often do your friends and classmates encourage each other to lose weight? How often do your friends and classmates talk about weight or dieting? How often do your friends and classmates worry about their weight? How often do your friends and classmates worry about what they eat? How often do your friends and classmates diet? How often do your friends and classmates skip meals? Do you think that your friends and classmates take a lot of notice in each other’s weight and shape? Are weight and shape important to your friends and classmates?
Factor 1 Parent
Factor 2 Peer
Factor 3 Media
0.05
0.15
0.56
0.10
0.14
0.47
0.04 0.13 −0.04 −0.04
0.09 0.08 −0.06 −0.06
0.56 0.42 0.77 0.64
0.01 −0.02 −0.03
0.06 −0.09 −0.05
0.48 0.83 0.71
0.02 0.33
0.12 0.07
0.43 −0.01
0.38 0.35
−0.01 0.06
−0.00 −0.03
0.38 0.49 0.41 0.41 0.56 0.53 0.50 0.35 0.38 0.54 0.66 0.72 0.72 0.55 0.77 0.58
−0.06 −0.14 −0.02 0.03 0.00 0.05 0.10 0.02 0.01 0.08 −0.03 −0.03 −0.00 0.10 −0.02 0.08
0.03 0.00 0.02 0.03 0.02 0.07 0.02 0.06 −0.06 0.04 0.07 0.06 0.02 −0.12 −0.11 0.04
0.67 0.09 0.08 0.09
−0.09 0.49 0.49 0.47
0.04 0.05 0.13 0.14
0.01
0.34
−0.04
−0.03 0.00 −0.08 0.03 0.03 0.09 −0.05 −0.04
0.69 0.55 0.80 0.80 0.74 0.71 0.56 0.60
−0.16 −0.03 0.01 −0.03 0.02 −0.02 0.02 0.04
−0.10
0.64
0.16
H. Keery et al. / Body Image 1 (2004) 237–251
preoccupation with weight and dieting. These include talk about weight and shape, dieting behavior, and importance of one another’s weight and shape among friends. The alpha of this measure in a sample of girls aged 12–16 years was 0.87 (Schutz et al., 2002). The alpha for this sample was 0.89. Peer Influence Scale is an adaptation of a three-item scale developed by Levine et al. (1994). The scale assesses participants’ perceptions of peer influence on weight and appearance. Levine et al. found an alpha of 0.73 in sample of sixth to eighth grade girls. An alpha of 0.70 was obtained in the current sample. The McKnight Risk Factor Survey is a multidimensional measure used to assess risk factors for eating disturbance in fourth through twelfth grade girls (Shisslak et al., 1999). The two-item Weight Rejection Scale was used to assess girls’ interpersonal experiences related to rejection caused by their weight. The alpha for the current sample was 0.72. Media influence Media Influence Scale is an adaptation of a threeitem scale developed by Levine et al. (1994). The items were modified to assess participants’ recollection of media influence on weight and appearance. The alpha in the current sample was 0.83. The Interest Scale (Harrison, 2000) is a six-item scale that assesses a girl’s interest in watching television shows or reading magazines that cover topics such as dieting, fitness and exercise, and fashion. The alpha for the current sample was 0.82. Two items from the Perceived Sociocultural Pressure Scale (Stice, Ziemba, Margolis, & Flick, 1996) were used to assess perceived pressure from the media to lose weight. The alpha for this sample was 0.72. The media scales were preceded by a brief survey that asked participants to list their three favorite television shows, magazines, and movies. This was included in order to prime their recall of media images and messages in order to facilitate the responses to these items. This scale was not included in analyses. Mediational variables Internalization. The Sociocultural Internalization of Appearance Questionnaire–Adolescents (SIAQ-A) was used to assess internalization. This scale is a downward extension of the Internalization Subscale of the Sociocultural Attitudes Towards Appearance
241
Questionnaire-3 (Thompson, van den Berg, Roehrig, Guarda, & Heinberg, 2004). Internalization items capture an incorporation of the media images and messages of attractiveness into one’s own self-identity. A sample item is: Reading magazines makes me want to change my appearance. This five-item measure has been validated in samples from three countries and a high internal consistency (>0.80) has been found in six independent samples of 10–15 year-old girls (Keery, Shroff, Thompson, Wertheim, & Smolak, in press). The alpha for the current sample was 0.88. Appearance comparison. Two measures of social comparison were included, in order to provide for an index of general and specific appearance comparison. The Physical Appearance Comparison Scale (Thompson, Heinberg & Tantleff, 1991) is a five-item measure that assesses an overall tendency to compare one’s own appearance to the appearance of generic others in social situations (e.g., parties and social gatherings). In the current study, one of the items, which is the only reverse-coded item on the scale, had a negative correlation with the item total. Therefore, this item was deleted and the resulting alpha was 0.80. The Social Comparison Questionnaire (Schutz et al., 2002) is a 14-item scale that assesses the frequency of both appearance and more general comparison tendencies and the targets for social comparisons (i.e., friends, mother, father, fashion models). The items related to comparing oneself to different targets in terms of appearance were used in this study. The alpha for the current sample was 0.80. Outcome measures Body image. Two measures were used to index body dissatisfaction, in order to provide for a broader test of the mediational hypotheses in the simple paths. An attempt was made to use a measure that captured the specific dissatisfaction with body size and weight, along with a measure that assessed a more global aspect of appearance evaluation (Thompson, 2004). The Eating Disorder Inventory–Body Dissatisfaction Scale (EDI-BD; Garner, 1991) contains nine items that measure satisfaction with the size of specific body sites such as the waist, thighs, and hips. Shore and Porter (1990) found that the scale has an alpha of 0.91 for 11–18-year-old females. The alpha for the EDI-BD was 0.88 for the current sample. The Self-Image Ques-
242
H. Keery et al. / Body Image 1 (2004) 237–251
tionnaire for Young Adolescents–Body Image Subscale (SIQYA; Peterson, Schulenberg, Abramowitz, Offer, & Jarcho, 1984) is an 11-item subscale designed to assess broad, positive feelings toward the body in 10–15-year olds. The alpha for this sample was 0.79. Eating disturbance. The Eating Disorder Inventory– Drive for Thinness Scale (EDI-DT; Garner, 1991) measures restricting tendencies, desire to lose weight, and fear of weight gain. It has an internal consistency of 0.83 for a combined sample of eating disordered individuals and 0.81–0.91 for four samples of nonpatient female controls (Garner, 1991). Shore and Porter (1990) found an alpha of 0.81 for 11–18-yearold females. The alpha for the current sample was 0.91. The Eating Disorder Inventory–Bulimia Scale (EDI-B; Garner, 1991) assesses the tendency toward overeating, binging, and the impulse to engage in self-induced vomiting. Shore and Porter (1990) found an alpha of 0.69 for 11–18-year-old females. The alpha for the EDI-B was 0.83 for the current sample. Psychological functioning As with our previous work (e.g., van den Berg et al., 2002), we sought to provide a composite index of psychological functioning. Self-esteem and depression were included as indicators of psychological functioning (e.g., Thompson, Coovert, Richards, Johnson, & Cattarin, 1995) and, for the first time, we also included for provisional evaluation a measure of perfectionism, given work detailing its potential connection to body image and eating dysfunction (McKnight Investigators, 2003; Santonastaso, Friederici, & Favaro, 1999; Tyrka, Walden, Graber, & Brooks-Gunn, 2002). Self-esteem. The Rosenberg Self-Esteem Inventory is a 10-item measure of general feelings of self-worth (Rosenberg, 1965). The instrument has been shown to be reliable and valid for use with adolescent samples. For example, the alpha was 0.84 in a study of 12–16year-old girls (Schutz et al., 2002). The alpha for the current sample was 0.86. Depression. The Center for Epidemiologic Studies Depression Scale (CESD; Radloff, 1977) was used to
assess depressive symptoms. Items use a four point response format ranging from 0 = never to 4 = most of the time. The reliability of this scale has been well documented (Robert, Lewinsohn, & Seeley, 1991). The alpha for the current sample was 0.88. Perfectionism. The Eating Disorder Inventory–Perfectionism Scale (EDI-P; Garner, 1991) contains six items that measure expectations for achievement and attitudes associated with eating disturbance. Schutz et al. (2002) found an alpha of 0.83 in a sample of 12–16-year-old girls. The alpha for the EDI-P was 0.79 for the current sample. Body mass index Quetlet’s Index of Fatness (Garrow & Webster, 1985) was utilized as a measure of body mass index (BMI). This index is a ratio of weight (in kilograms) to squared height (in meters): W/H2 . Height and weight were obtained through self-report. Past research has examined the reliability of self-report measures of height and weight (Brooks-Gunn, Warren, Rosso, & Gargiulo, 1987). Findings indicate that although adolescents, particularly females, tend to underreport their weight by several pounds and their height by just over an inch, the correlations between actual and reported figures are relatively high (0.98 and 0.75) (Brooks-Gunn et al., 1987). Procedure This study was approved by the Institutional Review Board at the University of South Florida. Specific consent procedures were based on the policies of the Pasco County School District. One week prior to data collection, students were given a letter explaining the purpose of the study to take home to their parents. The letter stated that parents should contact the school if they did not want their child to participate. Participants watched a short video of the instructions for the survey on the closed circuit television system in which they were informed that they could withdraw from the study at any time without penalty. All individuals attending a single middle school in grades 6–8 on the testing day were able to participate in the study. Following survey administration, teachers read the students a short debriefing about the purpose of the study.
H. Keery et al. / Body Image 1 (2004) 237–251
Data reduction and analysis Of the original sample of 425, data from 108 were not included in the primary analyses. Most of the deleted participants did not answer several questions on one or more of the parental influence measures (in particular, the Parental Involvement Scale was not completed by 66 participants). Most likely, these children were from single parent homes, although this information was not formally obtained. Additionally, participants who were missing responses on 5% of the items (i.e., 8 from the total item pool of 143) were also deleted from analyses Finally, one case was deleted due to missing all data on the EDI-P Scale. This resulted in a final sample of 325 participants. In order to determine if there was any consistent difference between the retained versus discarded sample, t-tests were conducted on all variables other than the parental influence measures. There was only one significant effect that resulted from these 11 analyses (scores on the depression scale were significantly lower for the retained sample). In cases wherein fewer than 5% of the items were not answered, participants’ own data were used to create a mean to be imputed for individual scales. For the newly created composite variables from the factor analyses (i.e., peer, parent, media, comparison) the individual’s mean for the individual composite variable was imputed. On all other scales, missing data were replaced by the individual’s mean on that scale. This created a sample of 325 participants with no missing data points on any of the scales administered. However, 43 of these individuals were missing data on height or weight. Preliminary and planned analyses Internal consistencies (Cronbach’s alpha), means, and standard deviations were calculated for all the scales. Principal axis factor analysis using a Promax oblique rotation was performed to determine the factor structure of the sociocultural influence scales (e.g., peer, parental, media). This strategy was also used to ensure that the mediational variables (internalization, comparison) were distinct. Analysis of the scree plot, eigenvalues, theory, and ease of interpretation determined the number of factors retained. Standardized regression coefficients were used as factor loadings.
243
Three a priori criteria for item retention were used. First, items had to load ±0.30 on the primary factor. Second, items could not load more than ±0.20 on alternative factors. Third, a spread of ±0.20 had to be present between loadings on alternative factors. Simple path analyses were tested using the set of three regression equations as recommended by Baron and Kenny (1986). Structural equation modeling (SEM) was used to evaluate the full Tripartite Model (specific models tested are discussed later).
Results Factor analysis: sociocultural influence items This factor analysis examined the original 50 items included in all of the peer, parent, and media scales in the study. Examination of the scree plot indicated that three or four factors should be retained. Factor analyses were run for three and four factor solutions. Visual inspection of the scree plot, ease of interpretation, and theory indicated that the three-factor solution was the best fit. Of the original 50 items, 1 item on the media factor, 3 items on the parent factor and 3 items on the peer factor were deleted, primarily because they cross-loaded on other factors. A factor analysis on the remaining 43 items resulted in a 43-item scale with a three-factor solution. Inspection of the item content reveals that the three factors derived from this scale reflect the dimensions of peer, parent, and media influence. The 43-item scale includes 10 media items, 20 parent items, and 13 peer items. The internal consistency for each factor and scale total were: parent = 0.88; peer = 0.89; media = 0.86; total = 0.87. Table 1 contains the items and factor loadings. Factor analysis: internalization and comparison items This factor analysis examined the original 18 items included in the internalization and comparison scales. Examination of the scree plot indicated that two or three factors should be retained. Factor analyses were run for two- and three-factor solutions. Visual inspection of the scree plot, ease of interpretation, and theory indicated that the three-factor solution was the best fit. Of the original 18 items, 4 comparison items were deleted because they cross-loaded on the internaliza-
244
H. Keery et al. / Body Image 1 (2004) 237–251
tion factor. One item that cross-loaded on the two comparison factors was retained. A factor analysis on the remaining 14 items resulted in a 14-item scale with a three-factor solution. Inspection of the item content reveals that the three factors derived from this scale reflect the dimensions of internalization, peer comparison, and family comparison. The 14-item scale includes 5 internalization items, 6 peer comparison items, and 3 family comparison items. None of the original internalization items were deleted, therefore the entire SIAQ-A scale was used for the internalization variable. The internal consistency for each factor was: internalization = 0.88, peer comparison = 0.90, family comparison = 0.70. This procedure established that internalization and comparison were distinct constructs and could each be used as mediators. As such, the peer and family comparison factors were combined into one composite comparison variable for further analyses. The alpha for the composite comparison factor was 0.86. Correlational and mediational analyses Table 2 contains the intercorrelations among all measures. These correlations were used as the basis for conducting mediational analyses, discussed later. Additionally, an evaluation of the correlations was used to consider whether there was support for the inclusion of the perfectionism indicator in the full model as part of a psychological functioning latent variable (with self-esteem and depression). As Table 2 illustrates, the shared variance between perfectionism and self-esteem was 1.2% and the shared variance between perfectionism and depression was 4.4%. In contrast, the shared variance between self-esteem and depression was 32.5%, therefore perfectionism was not included in the path models tested later. Scores for the self-esteem and depression scale were transformed into z-scores and summed to provide a single indicator of psychological functioning. Internalization and comparison as mediators An evaluation of the correlation matrix reveals significant relationships among the influence variables (peer, parents, media) internalization, comparison,
and both measures of body dissatisfaction. Additionally, the pattern of correlations for both measures of body dissatisfaction and the sociocultural influence and mediational variables were very similar. This fact, along with the shared variance of 0.46 between the two SIQYA and EDI-BD supported the creation of a composite body dissatisfaction variable for further analyses. Simple path models were tested with internalization and comparison as mediators between the sociocultural influence variables (peer, parent, media) and composite body dissatisfaction using the method outlined by Baron and Kenny (1986). The results indicated that internalization and comparison fully mediated the relationship between parental influence and body dissatisfaction (see Fig. 2). Internalization and comparison partially mediated the relationship between peer influence and body dissatisfaction (see Fig. 3) and also partially mediated the relationship between media influence and body dissatisfaction (see Fig. 4). Because of the consistency of the findings in the simple paths indicating that each of the sociocultural influence measures had a significant effect, either partially or fully mediated, on body dissatisfaction, a composite influence variable (sum of peer, parents, and media) was used for subsequent SEM models. Use of a composite also has the advantage of limiting the number of paths, producing a more parsimonious model for SEM testing.
Fig. 2. Mediation of parental influence. The number in the parentheses is the standardized beta weight for the direct relationship between the influence and outcome measure. Evaluation of this number against the standardized beta weight from the regression equation allows for an interpretation of mediation. ∗∗ p < 0.01.
Table 2 Correlations among variables INT – 0.15∗∗ 0.48∗∗ 0.36∗∗ 0.71∗∗ 0.40∗∗ 0.40∗∗ 0.65∗∗ 0.54∗∗ 0.51∗∗ 0.49∗∗ 0.56∗∗ 0.59∗∗ 0.37∗∗ 0.44∗∗ 0.61∗∗ 0.02
RSE
CESD
MEDIA
PARENT
PEER
COMP
DISS
SIQYA
EDI-BD
EAT
EDI-DT
EDI-BU
PSYCH
Total SI
BMI
– 0.57∗∗ 0.35∗∗ 0.25∗ 0.26∗∗ 0.46∗∗ 0.66∗∗ 0.65∗∗ 0.55∗∗ 0.59∗∗ 0.55∗∗ 0.48∗∗ 0.79∗∗ 0.36∗∗ 0.14∗
– 0.23∗∗ 0.13∗ 0.28∗∗ 0.42∗∗ 0.48∗∗ 0.48∗∗ 0.38∗∗ 0.45∗∗ 0.41∗∗ 0.37∗∗ 0.95∗∗ 0.26∗∗ 0.17∗∗
– 0.44∗∗ 0.47∗∗ 0.56∗∗ 0.46∗∗ 0.43∗∗ 0.40∗∗ 0.55∗∗ 0.55∗∗ 0.40∗∗ 0.30∗∗ 0.75∗∗ 0.08
– 0.36∗∗ 0.34∗∗ 0.26∗∗ 0.25∗∗ 0.23∗∗ 0.43∗∗ 0.44∗∗ 0.31∗∗ 0.19∗∗ 0.83∗∗ 0.02
– 0.42∗∗ 0.36∗∗ 0.32∗∗ 0.34∗∗ 0.46∗∗ 0.47∗∗ 0.33∗∗ 0.30∗∗ 0.76∗∗ 0.14∗
– 0.58∗∗ 0.57∗∗ 0.49∗∗ 0.58∗∗ 0.54∗∗ 0.47∗∗ 0.48∗∗ 0.54∗∗ 0.08
– 0.93∗∗ 0.90∗∗ 0.69∗∗ 0.65∗∗ 0.55∗∗ 0.60∗∗ 0.44∗∗ 0.31∗∗
– 0.68∗∗ 0.65∗∗ 0.59∗∗ 0.55∗∗ 0.60∗∗ 0.41∗∗ 0.21∗∗
– 0.61∗∗ 0.61∗∗ 0.45∗∗ 0.49∗∗ 0.39∗∗ 0.39∗∗
0.61∗∗ 0.45∗∗ 0.49∗∗ 0.39∗∗ 0.39∗∗
– 0.55∗∗ 0.51∗∗ 0.60∗∗ 0.17∗∗
– 0.45∗∗ 0.43∗∗ 0.10∗∗
– 0.33∗∗ 0.17∗∗
– 0.09
–
–
0.11∗ 0.21∗∗ 0.19∗∗ 0.16∗∗ 0.21∗∗ 0.24∗∗ 0.17∗∗ 0.21∗∗ 0.09 0.30∗∗ 0.27∗∗ 0.27∗∗ 0.20∗∗ 0.24∗∗ −0.01
Note: INT = internalization; EDI-P = Eating Disorder Inventory—Perfectionism; RSE = self-esteem; CESD = Center for Epidemiologic Studies Depression Scale; MEDIA = media influence; PARENT = parent influence; PEER = peer influence; COMP = comparison; DISS = composite body dissatisfaction; SIQYA = Self-Image Questionnaire for Young Adolescents; EDI-BD = Eating Disorder Inventory—Body Dissatisfaction Scale; EDI-DT = Eating Disorder Inventory—Drive for Thinness Scale; EDI-BU = Eating Disorder Inventory—Bulimia Scale; PSYCH = total depression and self-esteem; Total SI = total sociocultural influence (peer, parent, media influence); BMI = body mass index. ∗ p < 0.05. ∗∗ p < 0.01.
H. Keery et al. / Body Image 1 (2004) 237–251
INT EDI-P RSE CESD MEDIA PARENT PEER COMP DISS SIQYA EDI-BD EAT EDI-DT EDI-BU PSYCH Total SI BMI
EDI-P
245
246
H. Keery et al. / Body Image 1 (2004) 237–251
Fig. 3. Mediation of peer influence. The number in the parentheses is the standardized beta weight for the direct relationship between the influence and outcome measure. Evaluation of this number against the standardized beta weight from the regression equation allows for an interpretation of mediation. ∗ p < 0.05, ∗∗ p < 0.01.
Fig. 4. Mediation of media influence. The number in the parentheses is the standardized beta weight for the direct relationship between the influence and outcome measure. Evaluation of this number against the standardized beta weight from the regression equation allows for an interpretation of mediation. ∗ p < 0.05, ∗∗ p < 0.01.
SEM models Initially, six models were hypothesized to potentially fit the data (see Figs. 1, 5 and 6). The first four models are based on the primary and secondary models initially presented by Thompson et al. (1999). In Model 1, sociocultural influence has a directional influence on internalization and comparison. Internalization and comparison in turn are hypothesized to influence body dissatisfaction, which has a directional influence on restriction and bulimia. There is also a path from bulimia to psychological functioning. In Model 2, a path from sociocultural influences to body dissatisfaction is added to Model 1 (see Fig. 1).
In Model 3, global psychological functioning is hypothesized to mediate the relationship between sociocultural influence and internalization and between sociocultural influence and comparison. All other paths are identical to Model 1. Model 4 adds the path from sociocultural influences to body dissatisfaction to Model 3 (see Fig. 5). Models 3 and 4 are essentially the secondary model originally hypothesized by Thompson et al. (1999) and also evaluated by van den Berg et al. (2002) in a sample of adults. Models 5 and 6 are slight variations of Models 3 and 4. In Model 5, psychological functioning is moved alongside sociocultural influences and conceptualized
Fig. 5. Hypothetical SEM Models 3 and 4 based on Thompson et al. (1999).
H. Keery et al. / Body Image 1 (2004) 237–251
247
Fig. 6. Hypothetical SEM Models 5 and 6 based on Thompson et al. (1999).
as a second formative influence on comparison and internalization. The covariance between sociocultural influences and global psychological functioning is also included in this model. The remaining paths are the same as those in Models 1–4. In Model 6, direct paths from sociocultural influences and psychological functioning to body dissatisfaction are added to Model 5 (see Fig. 6). An SEM evaluation of all six hypothetical models demonstrated poor fit overall. See Table 3 for fit indices. Given that Models 1 and 2 had lower RMSEA values than Models 3–6, the latter models were not further tested. A chi-square difference test revealed that there was no significant difference in fit between Models 1 and 2 (χ2 = 2.12, d.f. = 1, ns). Therefore, Model 1was chosen as the best fitting model be-
cause it had the lowest RMSEA. However, given the model’s poor fit, a specification search was undertaken. Examination of Wald tests and t-values indicated that all paths in the model were significant and none should be removed. The Lagrange Multiplier test indicated that adding several paths would improve the fit of the model. Based on data and theory, the path from sociocultural influence to restriction was freed. This modification to the model (Model 1A) resulted in a χ2 (11) = 126.35, which remained significant (see Table 2). In addition, the RMSEA was lowered to 0.18 and the CFI improved to 0.89. Following this modification, the Lagrange Multiplier test continued to indicate that the addition of paths would improve the fit of the model. Based on theory and data, the path from body dissatisfaction to global psychological function-
Table 3 Fit indices of six proposed models and modifications to Model 1
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 1–influence to restriction Model 1–body dissatisfaction to global psychological functioning Model 1–internalization to restriction Stice Model
χ2
df
p
RMSEA
CFI
ECVI
206.61 204.49 314.34 312.21 177.22 175.10 126.35 47.25 35.96 126.70
12 11 12 11 9 8 11 10 9 7
<0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
0.2237 0.2330 0.2789 0.2907 0.2402 0.2539 0.1799 0.1072 0.0962 0.2315
0.82 0.82 0.72 0.72 0.85 0.85 0.89 0.97 0.97 0.86
0.74 0.74 1.07 1.07 0.67 0.67 0.50 0.26 0.23 0.48
248
H. Keery et al. / Body Image 1 (2004) 237–251
Fig. 7. Final model with standardized path coefficients.
ing was added (Model 1B). This modification resulted in a significant χ2 (10) = 47.25, RMSEA = 0.11, and CFI = 0.97. One final path, from internalization to restriction, was freed based on results of the Lagrange Multiplier test and theory. The addition of this path resulted in a χ2 (9) = 35.96, RMSEA = 0.096, and CFI = 0.94. At this point in the analysis the model (Fig. 7) had achieved acceptable fit (Li-tze & Bentler, 1999), and no further theoretically justified paths were recommended by the Lagrange Multiplier test. One method of determining the usefulness of a relatively new theoretical model is to compare it to extant models. Stice’s (2001) Dual Pathway Model has perhaps received the most extensive evaluation in recent years. This model proposes that two formative influences (pressure to be thin; thin-ideal internalization) lead to body dissatisfaction, which then leads to both dieting and negative affect. Consequently negative affect and dieting are thought to produce bulimic symptoms. Stice (2001) compared this model to the Tripartite Influence model, finding that his model was a better fit to the data in a sample of adolescent girls. Our test of the Dual Pathway Model suggested that it was not as good a fit to the data as the Tripartite Influence Model for the current sample (see Table 2). The RMSEA was 0.23 and CFI was 0.86, indicating a poor fit to the data.2 2 For this analysis, we used our composite influence measure as the indicator for Stice’s pressures variable. In Stice (2001), he found that the data were a better fit to his Dual Pathway Model
Discussion The goal of the current study was to evaluate the Tripartite Influence Model in a sample of adolescent girls. Simple path analyses indicated that internalization and appearance comparison fully mediated the relationship between parental influence and body dissatisfaction. Internalization and comparison also partially mediated the relationships between peer influence and body dissatisfaction and between media influence and body dissatisfaction. The results of the mediational analyses are similar to results from a similar study with pre-adolescent girls (Shroff & Thompson, 2003). The SEM analysis of the Tripartite Influence Model and several variations on the basic model revealed that the original model (Thompson et al., 1999), with a few minor modifications, was an adequate fit to the data. One of the most interesting findings, and perhaps one that will lead to a revision of the Tripartite Influence Model, is the direct path from the sociocultural influences to restriction. This variable than the Tripartite Influence Model. However, it is unclear how he could have tested the full Tripartite Influence Model because there was no indicator of appearance comparison used in that study. Additionally, the measure of internalization used in Stice (2001) has been shown to index not internalization, but awareness of media images and messages (Thompson et al., 2004). Additionally, there were other differences in the measures used in the two studies. Certainly, any demonstration of the differential predictive utility of competing models requires replication.
H. Keery et al. / Body Image 1 (2004) 237–251
had indirect effects via internalization, comparison, and body dissatisfaction on restriction, yet there was also a direct effect. This finding replicates our work with an adult sample (van den Berg et al., 2002) and was also recently replicated in an adolescent sample in our laboratory (Shroff & Thompson, 2003). In effect, the power of these influences appears to be one of directly leading young girls to begin restriction, yet also setting the stage for restriction (possibly later) by enhancing internalization of societal messages, comparison, and body dissatisfaction. The original Tripartite Influence Model did not include BMI as a factor, therefore this variable was not included herein for model testing. An evaluation of the correlation between BMI and our influence and mediational variables indicates very little relationship (see Table 2). The only significant r was between BMI and peer influence (0.14). However, as might be expected BMI was significantly associated with body dissatisfaction (0.31), therefore, future modifications of the Tripartite Influence Model might include adding BMI, possibly as an exogenous predictor of body dissatisfaction. There are several limitations to the current study. First, the sample consisted of primarily Caucasian girls from a middle-class socioeconomic background. Second, at least 15% of the original sample were dropped because of missing data on parental measures, therefore, individuals who may have been from single parent homes were not included in analyses. Both of these sampling issues limit the generalizability of the current study. Third, the final model required modifications to achieve adequate fit. Although these modifications were theoretically plausible, they may be unique to the current sample, although as noted earlier, the current SEM model was replicated by Shroff and Thompson (2003). Another limitation of this study is its use of crosssectional data to model a developmental process. A prospective design would allow for a greater understanding of the development of body dissatisfaction and eating disturbance over time and is essential for the evaluation of causality and the determination of whether a specific variable is a causal risk factor (e.g., Field, 2004; Kraemer, Stice, Kazdin, Offord, & Kupfer, 2001; Stice & Hoffman, 2004). Such a design would also allow for greater confidence in directional hypotheses proposed by the Tripartite Influence
249
Model. For instance, the current model suggests that comparison leads to internalization rather than the converse, however, because both variables are mediators in the current model, the design is an equivalent model, which means that the models cannot be distinguished mathematically (e.g., MacCallum, Wegener, Uchino, & Fabrigar, 1993).3 Despite the limitations of the current study, the results suggest that interventions that address issues such as appearance related teasing and the modeling of unhealthy diet and exercise attitudes and behaviors by family members and peers may be useful in both prevention and treatment. The results of this study also highlight the need for continued work in the area of prevention interventions that teach young girls to be critical consumers of media. Interventions aimed at reducing the internalization of the sociocultural ideal of thinness and challenging girls to resist comparing themselves to others may help to interrupt the development of body image and eating disturbance (Stice & Hoffman, 2004).
References American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders—IV text revision. Washington, DC: American Psychiatric Association.
3 An argument can be made to support either comparison leading to internalization or the converse. Susan Paxton pointed out to us that Festinger’s (1954) original comparison theory framework supposed that comparison would only occur for an attribute of high importance to the individual, therefore, perhaps internalization of societal images and messages might precede elevations in appearance comparison tendencies. However, our view is that sociocultural influences spark a search for information regarding one’s own appearance in social and interpersonal contexts, driving an initial comparison process as a means of seeking information and clarification. If such a comparison process results in discouraging feedback (e.g., negative comparison experiences exacerbate appearance concern), the end result is a deeper and more nefarious modification of one’s self-view, leading to a more accepting belief in the veracity of the sociocultural and comparison information/feedback. Such a process eventuates in a more basic modification of one’s belief system, an internalization of the thin-ideal, leading the individual to seek appearance modification through multiple means, such as dieting, fashion, make-up, cosmetic surgery, etc. Of course, prospective data will ultimately clarify the directionality question.
250
H. Keery et al. / Body Image 1 (2004) 237–251
Baron, R. M., & Kenny, D. A. (1986). The moderatormediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Brooks-Gunn, J., Warren, M. P., Rosso, J., & Gargiulo, J. (1987). Validity of self-report measures of girls’ pubertal status. Child Development, 58, 829–841. Festinger, L. (1954). A theory of social comparison processes. Human Relations, 1, 117–140. Field, A. E. (2004). Risk factors for eating disorders: A evaluation of the evidence. In J. K. Thompson (Ed.), Handbook of eating disorders and obesity (pp. 17–32). New York: Wiley. Field, A. E., Camargo, C. A., Taylor, C. B., Berkey, C. S., & Colditz, G. A. (1999). Relation of peer and media influences to the development of purging behaviors among preadolescent and adolescent girls. Archives of Pediatrics and Adolescent Medicine, 153, 1184–1189. Field, A. E., Camargo, C. A., Taylor, C. B., Berkey, C. S., Roberts, S. B., & Colditz, G. A. (2001). Peer, parent, and media influences on the development of weight concerns and frequent dieting among preadolescent and adolescent girls and boys. Pediatrics, 107(1), 54–60. Garner, D. M. (1991). Eating disorders inventory-2: Professional manual. Odessa, FL: Psychological Assessment Resources. Garrow, J. S., & Webster, J. (1985). Quetlet’s index (W/H2) as a measure of fatness. International Journal of Obesity, 9, 147– 153. Harrison, K. (2000). The body electric: Thin-ideal media and eating disorders in adolescence. Journal of Communication, 50(3), 119–143. Keery, H., Shroff, H., Thompson, J. K., Wertheim, E., & Smolak, L. (in press). The Sociocultural Internalization of Appearance Questionnaire—Adolescents (SIAQ-A): Psychometric analysis and normative data for three countries. Eating and Weight Disorders: Studies on Anorexia, Bulimia and Obesity. Kraemer, H. C., Stice, E., Kazdin, A., Offord, D., & Kupfer, D. (2001). How do risk factors work together? Mediators, moderators, and independent, overlapping and proxy risk factors. American Journal of Psychiatry, 158, 848– 856. Levine, M. P., Smolak, L., & Hayden, H. (1994). The relation of sociocultural factors to eating attitudes and behaviors among middle school girls. Journal of Early Adolescence, 14, 471– 490. Li-tze, H., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis. Structural Equation Modeling, 6, 1–55. MacCallum, R. C., Wegener, D. T., Uchino, B. N., & Fabrigar, L. R. (1993). The problem of equivalent models in applications of covariance structure analysis. Psychological Bulletin, 114, 185–199. McKnight Investigators. (2003). Risk factors for the onset of eating disorders in adolescent girls: Results of the McKnight longitudinal risk factor study. American Journal of Psychiatry, 160, 248–254. Peterson, A. C., Schulenberg, J. E., Abramowitz, R. H., Offer, D., & Jarcho, H. D. (1984). A Self-Image Questionnaire for
Young Adolescents (SIQYA): Reliability and validity studies. Journal of Youth and Adolescence, 13, 93–111. Radloff, L. S. (1977). A CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385–401. Robert, R. E., Lewinsohn, P. M., & Seeley, J. R. (1991). Screening for adolescent depression: A comparison of depression scales. Journal of the American Academy of Child and Adolescent Psychiatry, 30, 58–66. Rosenberg, M. (1965). Measurement of self-esteem. In M. Rosenberg (Ed.), Society and the adolescent self-image (pp. 297–307). New York: Princeton University Press. Santonastaso, P., Friederici, S., & Favaro, A. (1999). Full and partial syndromes in eating disorders: A 1-year prospective study of risk factors among female students. Psychopathology, 32, 50–56. Schutz, H. K., Paxton, S. J., & Wertheim, E. H. (2002). Investigation of body comparison among adolescent girls. Journal of Applied Social Psychology, 32(9), 1906–1937. Shisslak, C. M., & Crago, M. (2001). Risk and protective factors in the development of eating disorders. In J. K. Thompson, & L. Smolak (Eds.), Body image, eating disorders, and obesity in youth: Assessment, prevention, and treatment (pp. 103–125). Washington, DC: American Psychological Association. Shisslak, C. M., Renger, R., Sharpe, T., Crago, M., McKnight, K. M., & Gray, N. et al., (1999). Development and evaluation of the McKnight Risk Factor Survey for assessing potential risk and protective factors for disordered eating in preadolescent and adolescent girls. International Journal of Eating Disorders, 25, 195–214. Shore, R. A., & Porter, J. E. (1990). Normative and reliability data for 11 to 18 year olds on the Eating Disorder Inventory. International Journal of Eating Disorders, 9, 201–207. Shroff, H., & Thompson, J. K. (2003). An evaluation of the Tripartite Influence Model with adolescent girls. Unpublished manuscript. Stice, E. (2001). A prospective test of the dual-pathway model of bulimic pathology: Mediating effects of dieting and negative affect. Journal of Abnormal Psychology, 110, 124–135. Stice, E., & Hoffman, E., 2004. Eating disorder prevention programs. In J. K. Thompson (Ed.), Handbook of eating disorders and obesity (pp. 33–57). New York: Wiley. Stice, E., Ziemba, C., Margolis, J., & Flick, P. (1996). The Dual Pathway Model differentiates bulimics, subclinical bulimics, and controls: Testing the continuity hypothesis. Behavior Therapy, 27, 531–549. Thompson, J. K. (2004). The (mis)measurement of body image: Ten strategies to improve assessment for applied and research purposes. Body Image: An International Journal of Research, 1, 7–14. Thompson, J. K., Cattarin, J., Fowler, B., & Fisher, E. (1995). The Perception of Teasing Scale (POTS): A revision and extension of the Physical Appearance Related Teasing Scale (PARTS). Journal of Personality Assessment, 65, 146–157. Thompson, J. K., Coovert, M. D., Richards, K. J., Johnson, S., & Cattarin, J. (1995). Development of body image, eating disturbance, and general psychological functioning in female
H. Keery et al. / Body Image 1 (2004) 237–251 adolescents: Covariance structure modeling and longitudinal investigations. International Journal of Eating Disorders, 18, 221–236. Thompson, J. K., Heinberg, L. J., Altabe, M., & Tantleff-Dunn, S. (1999). Exacting beauty: Theory, assessment, and treatment of body image disturbance. Washington, DC: American Psychological Association. Thompson, J. K., Heinberg, L. J., & Tantleff, S. (1991). The Physical Appearance Comparison Scale (PACS). The Behavior Therapist, 14, 174. Thompson, J. K., & Smolak, L. (Eds.). (2001). Body image, eating disorders, and obesity in youth: Assessment, prevention, and treatment. Washington, DC: American Psychological Association. Thompson, J. K., & Stice, E. (2001). Thin-ideal internalization: Mounting evidence for a new risk factor for body image disturbance and eating pathology. Current Directions in Psychological Science, 10(5), 181–183.
251
Thompson, J. K., van den Berg, P., Roehrig, M., Guarda, A., & Heinberg, L. J. (2004). The Sociocultural Attitudes Towards Appearance Questionnaire-3. International Journal of Eating Disorders, 35, 293–304. Tyrka, A. R., Waldron, I., Graber, J. A., & Brooks-Gunn, J. (2002). Prospective predictors of the onset of anorexic and bulimic syndromes. International Journal of Eating Disorders, 32, 282–290. van den Berg, P., Thompson, J. K., Obremski-Brandon, K., & Coovert, M. (2002). The Tripartite Influence Model of body image and eating disturbance: A covariance structure modeling investigation testing the meditational role of appearance comparison. Journal of Psychosomatic Research, 53(5), 1007– 1020. Wertheim, E. H., Paxton, S. J., & Blaney, S. (2004). Risk factors for the development of body image disturbance. In J. K. Thompson (Ed.), Handbook of eating disorders and obesity (pp. 463–494). New York: Wiley.