Gayle R. Bessenoff, Department of Psychology, University of Connecticut.
Thanks to Julie Keen and Sarah Copeland for their help with data collection and to Wendi Gardner, Dave Kenny, Gloria Cowan, and the research lab group at the Yale Center for Eating and Weight Disorders for their thoughtful comments on this research.
Address correspondence and reprint requests to: Gayle R. Bessenoff, Department of Psychology, University of Connecticut, 406 Babbidge Road, Storrs, CT 06269-1020. E-mail: firstname.lastname@example.org
The current study explored body image self-discrepancy as moderator and social comparison as mediator in the effects on women from thin-ideal images in the media. Female undergraduates (N= 112) with high and low body image self-discrepancy were exposed to advertisements either with thin women (thin ideal) or without thin women (neutral-advertisement control). Exposure to thin-ideal advertisements increased body dissatisfaction, negative mood, and levels of depression and lowered self-esteem. In addition, social comparison processes mediated the relationship between exposure to thin-ideal advertisements and negative self-directed effects. Notably, self-discrepancy moderated this mediation. Women with high levels of body image self-discrepancy were more likely to engage in social comparison from exposure to thin-ideal advertisements, as well as more likely to have those comparison processes induce self-directed negative consequences. This research provides support for an individual difference variable (body image self-discrepancy) that moderates the mediating effect of social comparison from exposure to thin-ideal media.
Interestingly, a common factor in both body dissatisfaction and thin-ideal internalization is that of social comparison. For example, body dissatisfaction is associated with the tendency to compare one's body to others' bodies (Stormer & Thompson, 1996). In addition, measures of thin-ideal internalization include statements assessing level of comparison to thin ideals, such as “I wish I looked like a swimsuit model” and “I often read magazines like Cosmopolitan, Vogue, and Glamour and compare my appearance to the models” (Heinberg et al., 1995). Perhaps this tendency to compare oneself to others possessing the thin ideal is the main source of the negative effects produced by exposure to thin-ideal media.
Social comparison theory states that we seek to compare ourselves to others we believe are similar to ourselves, particularly to determine our own levels of abilities and successes (Festinger, 1954). There are two major types of social comparison. Downward social comparison, comparison to others we perceive to be less fortunate than ourselves in some domain, generally serves to enhance mood and feelings of self-worth (Wills, 1991). Upward social comparison, comparison to others we perceive to be socially better than ourselves, generally leads to negative mood and can threaten self-evaluation (e.g., Gibbons & Gerard, 1989; Wheeler & Miyake, 1992). Upward comparisons serve to enhance the self by eliciting behaviors to improve oneself; when discrepancies between the self and the comparison standard arise, people are motivated to change the self to be more like the comparison standard (Higgins, 1987).
With regard to the thin ideal, Stormer and Thompson (1996) found that women who were more likely to compare their body to others' bodies had higher levels of body dissatisfaction. Furthermore, when study participants engaged in social comparison to thin models (upward comparison), both adolescent girls and college-aged women suffered increased negative mood (e.g., Tiggemann & McGill, 2004), greater body dissatisfaction and eating disorder symptoms (e.g., Field, Carmago, Taylor, Berkey, & Colditz, 1999), as well as diminished beliefs of their own physical attraction (e.g., Martin & Gentry, 1997) than controls who did not explicitly compare. When participants were encouraged to compare themselves to the models, the relationship between media exposure and negative affect and increased weight concerns was mediated by the amount of social comparison (Tiggemann & McGill, 2004).
Possessing a self-discrepancy may moderate the likelihood of making social comparisons to thin-ideal media. Self-awareness research (e.g., Duval & Wicklund, 1972) suggests that self-relevance increases the likelihood for social comparison. When a standard is made salient, individuals who deem the standard's domain as relevant (either chronically or by temporarily activating the self ) will use it as a basis for self-judgment, inducing self-comparison and self-regulatory processes (e.g., Scheier & Carver, 1988). Individuals possessing a body image self-discrepancy associate their failure to reach their body ideals with their self-concept (Bessenoff, 2004). Thus, salience of the thin ideal (e.g., looking at thin-ideal media) should elicit social comparison to these images. However, self-discrepancy may moderate not only likelihood of social comparison, but also the negative impacts of this comparison. Posavac et al. (1998) found that when exposed to thin-ideal media, only those individuals who were dissatisfied with their bodies reported greater weight concerns from comparison to the ideal. Posavac et al. suggested that women with low body dissatisfaction were not affected by the thin imagery because they either had a body similar to the models or they grounded their self-worth in areas unrelated to body image. In other words, the low body dissatisfaction participants did not possess a body image self-discrepancy. However, participants who did report increased weight concerns may have perceived themselves as failing in the self-relevant domain of body image, such that they believed their current physique to be discrepant from their ideal physique. By possessing such a self-discrepancy, social comparison to the thin ideal may have been induced in those individuals with high levels of body dissatisfaction. In addition, the failure to meet their own body ideal may also have increased concerns with weight from this comparison, perhaps as a self-regulatory mechanism. Because Posavac et al. (1998) did not measure social comparison processes in their study, it is not clear who actually engaged in comparison processes when exposed to thin-ideal media and how these processes affected weight concerns.
The current study examined body image self-discrepancy as a possible moderator in social comparison processes from exposure to thin-ideal media, as well as in the negative effects of such comparisons. It was predicted that reactions to advertisements depicting the thin ideal should differ for individuals depending on their level of self-discrepancy. In addition, this research examined social comparison as a mediator in the relationship between exposure to thin-ideal advertisements and negative outcomes, such that thin-ideal exposure should induce comparison to the ideal (either the body ideal explicit in the advertisements or one's own personal body image ideal), which should then elicit lowered self-evaluation. Much of the research in this area has either had participants explicitly compare to an ideal or made assumptions about comparison without actually measuring how these processes were involved. Moreover, most research has examined body dissatisfaction or weight concerns solely as consequences of thin-ideal media exposure. However, effects of upward social comparison include depression of mood and threats to self-evaluation; thus, media exposure may lead to other negative consequences for the self beyond weight-related concerns. Consequently, both a moderator (self-discrepancy) and a mediator (social comparison) will be examined simultaneously in the present study, examining affective and self-evaluative measures, as well as weight-related responses from exposure to thin-ideal advertising.
Moderators affect the strength or direction of an effect of the predictor variable. Mediators account for the relationship between the independent and dependent variables (Baron & Kenny, 1986). Thus, self-discrepancy as a moderator variable would mean that negative effects from thin-ideal exposure are stronger for individuals possessing greater levels of body image self-discrepancy than for those with lower levels of this self-discrepancy. Social comparison as a mediator variable would mean that the relationships between thin-ideal exposure and various outcomes (such as weight concerns and negative self-evaluation) may be explained by the presence of social comparison processes, such that when you remove these processes, the relationship no longer holds. Moderated mediation suggests a dual role of both self-discrepancy and social comparison, such that the effect of social comparison on the relationship between thin-ideal exposure and negative outcomes may be stronger for high versus low self-discrepant women. As can be seen in Figure 1, the predicted moderated mediation can work in two possible ways: the effect of thin-ideal exposure on social comparison may differ by level of self-discrepancy, and social comparison may interact with level of self-discrepancy to affect weight concerns and self-evaluation.
In the current study, women possessing either high or low levels of body image self-discrepancy were exposed to advertisements depicting the thin ideal to examine differences in the consequences of salience of thin-ideal norms in media. Body image self-discrepancy was determined by differences between perceptions of one's own body physique and one's ideal body physique. Thin-ideal salience was manipulated by exposing participants to two types of advertising: clothing advertisements with thin women (thin-ideal prime) and non-clothing product advertisements without thin women (neutral-ad control). Dependent measures included indices of body dissatisfaction, drive for thinness, and bulimic tendencies, as well as levels of mood, depression, and state self-esteem incorporating three subareas of self-esteem: appearance, social, and performance self-esteem. In addition, participants indicated to what extent they were thinking comparison-related, weight-related, and weight-regulatory thoughts. The following three hypotheses were tested:
Hypothesis 1 : Self-discrepancy will moderate the relationship between exposure to thin-ideal advertisements and negative consequences, such that women with high levels of body image self-discrepancy exposed to thin-ideal advertisements should experience greater levels of body dissatisfaction, depression, negative mood, and negative self-esteem compared to low self-discrepant women.
Hypothesis 2 : Social comparison (extent of thinking comparison-related thoughts) will mediate the relationship between exposure to thin-ideal advertisements and negative consequences, such that individuals exposed to thin-ideal advertisements will be more likely to engage in social comparison and, in turn, experience negative effects such as body dissatisfaction and lowered self-esteem due to a failed comparison.
Hypothesis 3 : Self-discrepancy will moderate this mediation, such that possessing a body image self-discrepancy will not only increase the likelihood of comparison from exposure to thin-ideal advertisements but also increase the risk of negative consequences from this comparison.
One hundred twelve female participants (N= 112) were recruited from an Introductory Psychology participant pool at a mid-sized northeastern university and were given course credit for participation. Ages ranged from 17 to 39 years (M= 18.68, SD= 2.16, median age = 18). The majority of the participants were Caucasian/European American (n= 101); others were Asian (n= 3), African American (n= 2), Hispanic (n= 2), Native American (n= 1), or other (n= 2). One participant chose not to specify race. Participants were preselected based on level of body image self-discrepancy, which was determined by a body image survey administered during group testing at the beginning of the term. Participants were chosen based on their body image self-discrepancy to form two groups—high self-discrepant and low self-discrepant—based on a median split of all scores from the mass testing session (median = 10). Of those who participated in the study, scores ranged from 0 to 40 (N= 53, M= 16.33, SD= 7.06 for high self-discrepants; N= 59, M= 0.36, SD= 1.08 for low self-discrepants).
Participants were randomly assigned to one of two experimental conditions: thin-ideal prime or neutral-advertisement control. The experimenter was blind to participants' self-discrepant status. Experimental sessions included groups of one to four persons.
Body image survey This survey measured level of body image self-discrepancy. The survey was given to participants during a mass testing session at the beginning of the school term, approximately 4 to10 weeks before the experimental session. The survey incorporated the Pictorial Body Image Scale developed by Stunkard, Sorensen, and Schulsinger (1983). This scale is widely used in body image research (e.g., Fallon & Rozin, 1985; Harrison, 1993; Tiggemann & Lynch, 2001). The scale consists of nine female figures that range in sizes from extremely thin (10) to extremely overweight (90). The scale was used to determine participants' actual and ideal body images. Participants were asked to make ratings based on the following: “the figure of your own sex that best represents your current appearance” (actual body image) and “the figure that best represents what you would ideally like to look like” (own-ideal body image). An actual–ideal discrepancy score was determined by subtracting each participant's ideal body image score from their actual body image score. Larger numbers indicate a greater distance between actual and ideal in the direction of desiring a thinner body. Participants were also asked to indicate their height and weight, which was then used to calculate Body Mass Index (BMI; weight [in lbs]× 703/height2[in inches]).
Advertisements Two sets of stimulus materials were constructed. Each set consisted of eight advertisements assembled into a packet. The thin-ideal set of advertisements consisted of clothing/fashion advertisements depicting three-fourths to full-body pictures of thin female models. Pretesting in a sample of women confirmed that the women in these advertisements were perceived to be quite thin, M= 2.67 (SD= 1.23) on a scale ranging from 1 (extremely thin) to 9 (overweight). The non–thin-ideal (neutral-ad control) set of advertisements consisted of non-clothing product advertisements (e.g., watches, cars, furniture) that did not depict thin female models. Although people appeared in three of the eight advertisements (one with a young boy, two depicting a middle-aged male), these ads did not include any women. The stimuli were selected from full-page color advertisements appearing in popular women's magazines (e.g., Glamour, Vogue). Ads were pretested with a sample of women as to how the advertisements made them feel about their bodies. This pretest also assessed the sample's perception of how the advertisements would make women in general feel about their bodies. These two questions were responded to on a scale ranging from 1 (insecure) to 9 (secure). The pretest sample felt more insecure about their bodies from viewing the thin-ideal advertisements than the neutral advertisements, t(31) = 3.78, p < .01 (M= 6.59 for neutral advertisements, M= 5.34 for thin-ideal advertisements). The sample also perceived that the thin-ideal advertisements would be more likely to make women in general feel more insecure about their bodies than the non–thin-ideal advertisements, t(31) = 4.70, p < .001 (M= 6.08 for neutral advertisements, M= 4.36 for thin-ideal advertisements).1
Mood Scale To assess mood, the scale consisted of 12 affective adjectives, 6 relating to dejection (e.g., dejected, sad) and 6 to agitation (e.g., agitated, frustrated; see Higgins, 1987). Participants were asked to rate the descriptiveness of each adjective for how they felt at that moment on a 4-point Likert scale ranging from 1 (not at all) to 4 (very much). Larger numbers indicate experiencing that particular mood to a greater degree. The scales exhibit satisfactory internal consistency (alphas of .83 for Dejection and .74 for Agitation).
State Self-Esteem Scale The 20-item scale by Heatherton and Polivy (1991) was used to assess three subareas of self-esteem: seven questions related to appearance self-esteem (e.g., “I am pleased with my appearance”), seven questions related to social self-esteem (e.g., “I worry about what other people think of me”), and six questions related to performance self-esteem (e.g., “I feel confident about my abilities”). Participants rated their degree of agreement with each item on a 5-point Likert scale ranging from 1 (not at all) to 5 (extremely). Larger numbers indicate higher levels of self-esteem. All three subscales, as well as the total self-esteem scale, exhibit satisfactory internal consistency (alphas of .85 for Appearance, .82 for Performance, .89 for Social, and .90 for the total scale).
Automatic Thoughts Questionnaire (ATQ)Hollon and Kendall's (1980) 30-item measure was used to determine level of depression-related cognitions. Participants were asked to rate on a 5-point Likert scale ranging from 1 (not at all) to 5 (all the time) how frequently they had experienced particular thoughts while participating in the current experiment, such as “I'm worthless,”“I'm a failure,” and “I wish I were a better person.” Larger scores indicate greater levels of depression-related thoughts. The ATQ exhibits satisfactory internal consistency, an alpha of .93, and the test has been found to reliably discriminate between depressed and nondepressed subjects.2
Eating Disorder Inventory-2 (EDI-2S)Garner's (1991) 91-item self-report questionnaire measures anorexic and bulimic symptomatology and consists of nine subscales: Drive for Thinness (DFT), Body Dissatisfaction (BD), Bulimia, Ineffectiveness, Perfectionism, Interpersonal Distrust, Interoceptive Awareness, Maturity Fears, and Social Insecurity. However, because this questionnaire is quite long, a 30-item version validated by Rienecke-Hoste, Mineka, Martinovich, and Fitzgibbon (2004) was used instead. Participants rated 30 statements on a 6-point scale ranging from 1 (never) to 6 (always). Example items include “I am terrified of gaining weight,”“I think my hips are too big,” and “I have gone on eating binges where I felt that I could not stop.” Larger numbers indicate greater levels of symptomatology. For purposes of this study, the DFT, BD, and Bulimia subscales were used in analyses (alphas of .74 for DFT, .80 for BD, and .83 for Bulimia).
Extent Thoughts Questionnaire Thoughts relating to social comparison, weight, and weight-reduction activities were assessed with a questionnaire constructed for this study that measured the extent to which participants engaged in certain thoughts while looking at and rating the advertisements. Participants responded on a 5-point Likert scale ranging from 1 (not at all) to 5 (to an extreme degree) to the following questions: “To what extent were your thoughts related to comparing yourself to something or someone in the ads?”, “To what extent did you think about thoughts related to aspects of body weight?”, and “To what extent did you think about weight reduction activities (such as dieting and exercising)?” Additional questions, such as “To what extent did you think about the products and their features?” were added to help bolster the cover story. Larger numbers indicate having experienced these thoughts to a greater extent while looking at and rating the advertisements.
Participants worked in individual rooms, and all instructions were given orally by the experimenter. Participants were told the experiment was “a marketing research study that explores different attitudes related to the self and others.” Participants were first given a marketing survey that asked them to rate their frequency of purchasing of a variety of products to bolster the cover story. They were then given the packet containing the advertisements. Half of the participants received the clothing advertisements with thin women (thin-ideal prime) and the other half received the non-clothing product advertisements without thin women (neutral-ad control). Participants completed a questionnaire assessing their attitudes toward the products shown in the advertisements. In the thin-ideal condition, the questionnaire consisted of six questions about each advertisement (“How much do you like this outfit?”, “How much do you like this advertisement?”, “How likely would you be to buy this outfit if you saw it in a store?”, “Would you wear this outfit to a party?”, “How flattering would this outfit be on you?”, “How flattering would other people think this outfit is on you?”). In the neutral-ad condition, the questionnaire consisted of four questions about each advertisement (“How much does this product appeal to you?”, “How much do you like this advertisement?”, “How likely would you be to buy this product?”, “How likely would your friends be to buy this product?”). Participants were told to look at each picture carefully and answer each question as honestly as possible and were given 10 minutes to rate all eight advertisements.
After rating the advertisements, participants were told that the study pertained to the different types of feelings and thoughts that advertisements can invoke in their audience. Participants were given a packet titled “Current Thoughts,” which included, in order, the mood survey, depression (ATQ) scale, and state self-esteem scale. Participants were asked to answer the questions in response to how they were feeling at the moment. After completing this packet, participants were given the extent thoughts questionnaire, which they were told would examine more specific thoughts they might have had while they were rating the advertisements. Participants were specifically told that if a particular thought occurred to them during the study, but not while they were rating the advertisements, they were to indicate “not at all” on the scale. The EDI-2S was then administered, ostensibly for a clinical researcher who needed some extra surveys completed. After completing the study, participants were thoroughly probed for suspicion, then debriefed and given course credit for their participation.
The data of 18 participants were excluded from the following analyses; 3 were excluded because their body mass index data were incomplete and 15 were excluded due to possible awareness of the experimental hypothesis. Of those 15, all were in the experimental condition and 8 were high self-discrepant (final total, N= 94; high self-discrepant, n= 48; low self-discrepant, n= 46; thin-ideal condition, n= 42; control condition, n= 52).
BMI differed significantly for high self-discrepant women as compared to low self-discrepant women in the sample, F(1, 90) = 41.57, p < .001, d= 1.34, such that high self-discrepant women had larger physiques (M= 24.20, SD= 2.98) than low self-discrepant women (M= 20.18, SD= 3.02). Although participants were randomly assigned to condition, women in the thin-ideal condition had marginally higher BMI scores than those in the neutral-ad condition, F(1, 90) = 3.28, p= .07, d= .38 (M= 22.76, SD= 3.03 for thin-ideal condition; M= 21.62, SD= 2.98 for neutral-advertisement condition). To control for group differences in BMI, as well as to examine perceptions of discrepancy rather than objective overweight status, BMI was used as a covariate in all of the following analyses.
Self-Discrepancy as Moderator Between Thin-Ideal Exposure and Negative Effects
The first set of analyses examined the moderating effect of self-discrepancy on the effects of thin-ideal exposure on various outcome variables. In particular, the prediction was that thin-ideal exposure effects would be stronger for high body image self-discrepant participants. The dependent variables studied were weight concerns, mood, self-esteem, and depression.
Weight concerns To examine the effects of exposure to thin-ideal advertisements on various aspects of weight concerns, a 2 × 2 (thin-ideal vs. neutral-ad condition by high vs. low self-discrepancy) multivariate analysis of covariance (MANCOVA) was conducted on the Body Dissatisfaction (BD), Drive for Thinness (DFT), and Bulimia subscales of the EDI-2S using BMI as a covariate. The analysis revealed no statistically significant main effect of condition, a main effect of self-discrepancy, F(3, 87) = 13.38, p < .001, and no significant condition × self-discrepancy interaction. As seen in Table 1, univariate analyses revealed main effects of self-discrepancy for BD, DFT, and Bulimia scores, such that women with high body image self-discrepancy exhibited greater affective, cognitive, and behavioral weight concerns than women with low body image self-discrepancy.3
Table 1. Adjusted Means (and Standard Errors) as a Function of Self-Discrepancy (Controlling for BMI)
High Self-Discrepancy(n = 48)
Low Self-Discrepancy(n = 46)
Note. EDI: Eating Disorder Inventory; ATQ: Automatic Thoughts Questionnaire; SSE: State Self-Esteem.
*p < .05. **p < .01. ***p < .001.
EDI-Drive for Thinness
Because much of the previous research has examined media effects solely on body dissatisfaction, a separate analysis of covariance (ANCOVA) was conducted with just the BD subscale of the EDI-2S using BMI as a covariate. As seen in Table 2, a statistically significant main effect of condition was found for BD, such that women exposed to the thin-ideal advertisements expressed greater body dissatisfaction than controls exposed to neutral advertisements. The condition × self-discrepancy interaction was not significant (see Table 3).
Table 2. Adjusted Means (and Standard Errors) as a Function of Condition (Controlling for BMI)
Thin-IdealCondition(n = 42)
Neutral-AdCondition(n = 52)
Note. EDI: Eating Disorder Inventory; ATQ: Automatic Thoughts Questionnaire; SSE: State Self-Esteem.
*p < .05. **p < .01. ***p < .001.
EDI-Drive for Thinness
Table 3. Adjusted Means (and Standard Errors) as a Function of the Interaction of Condition and Self-Discrepancy (Controlling for BMI)
Thin-Ideal Condition (n = 24)
Neutral-Ad Condition (n = 24)
Thin-Ideal Condition (n = 18)
Neutral-Ad Condition (n = 28)
Note. EDI: Eating Disorder Inventory; ATQ: Automatic Thoughts Questionnaire; SSE: State Self-Esteem. Means having the same subscript are not significantly different at p < .05 using a Bonferroni corrected difference comparison.
*p < .05. **p < .01. ***p < .001.
EDI-Drive for Thinness
To examine more generalized weight concerns, ANCOVAs (controlling for BMI) were conducted for responses to the following questions from the extent thoughts questionnaire: “To what extent did you think about thoughts related to aspects of body weight?” and “To what extent did you think about weight reduction activities (such as dieting and exercising)?” Analyses revealed a main effect of condition for both weight-related and weight-regulatory thoughts, such that these thoughts occurred to a greater extent in the thin-ideal condition than the control condition (see Table 2). There was also a main effect of self-discrepancy for both types of thoughts, such that women with high body image self-discrepancy experienced weight-related and weight-regulatory thoughts to a greater degree than low self-discrepant women (see Table 1). Importantly, the condition effect on these thought variables was moderated by self-discrepancy; the condition by self-discrepancy interaction yielded significance for both weight-related and weight-regulatory thoughts. Bonferonni adjusted pairwise comparisons revealed that in the thin-ideal condition high self-discrepant women experienced these thoughts to a greater extent than low self-discrepant women, ts(40) > 4.90, ps < .001, ds > 1.57, but there were no differences between groups in the control condition (see Table 3).
Beyond weight: Mood, self-esteem, and depression To examine the generalizability of negative consequences from exposure to thin-ideal advertising, analyses were conducted using the mood, self-esteem, and depression (ATQ) scales as dependent measures.
A condition × self-discrepancy MANCOVA (controlling for BMI) on dejection- and agitation-related moods yielded a main effect of condition, F(2, 88) = 12.29, p < .001, but no significant main effect of self-discrepancy. Univariate analyses yielded an effect of condition for both dejection- and agitation-related moods, such that women exposed to thin-ideal advertisements had greater dejection-related and agitation-related moods than controls exposed to advertisements without thin women (see Table 2). Notably, this analysis also yielded a significant condition × self-discrepancy interaction (see Table 3 for interaction effects). This interaction was statistically significant for both dejection and agitation. Bonferonni adjusted pairwise comparisons revealed that, for high self-discrepant women, exposure to thin-ideal advertisements led to greater dejection- and agitation-related mood than exposure to neutral advertisements, ts(46) > 4.81, ps < .001, ds > 1.42, but low self-discrepant participants did not show these differences between conditions (see Table 3).
ANCOVAs (controlling for BMI) were conducted on the total self-esteem and depression (ATQ) measures. Analyses yielded a main effect of condition for both total self-esteem and ATQ, such that participants exposed to advertisements depicting thin women had lower self-esteem and greater levels of depression-related thoughts than women exposed to neutral advertisements without thin women (see Table 2). There was also a main effect of self-discrepancy on total self-esteem, but not ATQ (see Table 1). Importantly, the condition × self-discrepancy interaction was significant for total self-esteem and demonstrated a marginal trend for ATQ (p= .08). As Table 3 shows, high self-discrepant women were more likely to exhibit lower self-esteem and greater depressive thoughts when exposed to thin-ideal advertisements as compared to neutral advertisements, ts(46) > 2.98, ps < .05, ds > .88; low self-discrepant women did not differ in their self-evaluations and self-esteem in the thin-ideal condition as compared to control, ts(44) < .45, ds < .13.
To examine the specific domains of self-esteem affected by exposure to thin-ideal advertisements, a MANCOVA (controlling for BMI) was conducted on the three subscales of state self-esteem (appearance, social, performance). This analysis revealed significant effects of condition, F(3, 87) = 2.94, p < .05, and self-discrepancy, F(3, 87) = 4.95, p < .01. The multivariate condition × self-discrepancy interaction was not significant. Univariate analyses revealed a significant main effect of condition for appearance self-esteem and social self-esteem, such that exposure to thin-ideal advertisements generated greater self-criticism and lower self-esteem in appearance- and social-related domains than exposure to control advertisements (see Table 2). A significant main effect of self-discrepancy was found for appearance self-esteem, such that high self-discrepant women had lower self-esteem in the domain of appearance than low self-discrepant women (see Table 1).
Again, because research on thin-ideal media effects has focused mainly on weight concerns, a separate ANCOVA was conducted on the appearance self-esteem subscale. This analysis yielded a significant condition × self-discrepancy interaction. Bonferonni adjusted pairwise comparisons revealed that high self-discrepant women had lower levels of appearance-related self-esteem when exposed to thin-ideal advertisements as compared to controls, t(46) = 3.61, p < .01, d=.77, whereas low self-discrepant participants did not show this difference (see Table 3).4
Summary of main effects of thin-ideal media exposure and moderation of body image self-discrepancy Participants exposed to advertising depicting thin-ideal women, as compared to controls exposed to advertisements not depicting thin-ideal women, exhibited significantly greater body dissatisfaction, weight-related thoughts, weight-regulatory thoughts, levels of depression, and dejection- and agitation-related mood, as well as lower appearance-related and total self-esteem. Body image self-discrepancy moderated the effect of thin-ideal exposure on weight-related thoughts, weight-regulatory thoughts, mood, appearance, and total self-esteem, such that advertisement condition had stronger effects for high than low self-discrepant participants.
Mediation by Social Comparison
Following the findings of previous social comparison research (e.g., Field et al., 1999; Tiggemann & McGill, 2004), it was predicted that exposure to thin-ideal advertisements would elicit comparison processes, resulting in various negative effects. Thus, social comparison was expected to mediate the relationship between thin-ideal exposure and such outcomes. Responses from the social comparison question from the extent thoughts scale (“To what extent were your thoughts related to comparing yourself to something or someone in the ads?”) were used as the mediator in the following analyses. As previously discussed and seen in Table 2, thin-ideal exposure significantly affected body dissatisfaction, weight-related and weight-regulatory thoughts, depression (ATQ), total self-esteem, appearance and social self-esteem, and negative mood (both agitation and dejection). Thus, the following analysis examines social comparison as a possible mediator in each of these effects. BMI was controlled in all analyses.
Condition was effect-coded as 1 for thin-ideal media and −1 for neutral-ad control in the regression analyses. As needed to test for mediation (Baron & Kenny, 1986), condition was found to be a significant predictor of social comparison, β= .69, p < .001, and social comparison was a significant predictor of weight-related thoughts (β= .68, p < .001), weight-regulatory thoughts (β= .56, p < .001), general negative mood (β= .35, p < .01), ATQ (β= .31, p < .01), total self-esteem (β=−.29, p < .01), and appearance self-esteem (β=−.25, p < .05). When social comparison was entered into the model, the effect of condition was reduced to nonsignificance for three of the eight dependent measures: ATQ (from β= .24, p < .01, to β= .06, ns), total self-esteem (from β=−.25, p < .01, to β= .04, ns), and appearance self-esteem (from β=−.25, p < .01, to β=−.15, ns). However, social comparison was found to fully mediate only the relationship between exposure to thin-ideal advertisements and symptoms of depression, Sobel's Z= 1.99, p < .05. Social comparison was also found to partially mediate the relationships between exposure to thin-ideal advertisements and weight-related thoughts (from β= .83, p < .001, to β= .62, p < .001, Sobel's Z= 3.44, p < .001) as well as weight-regulatory thoughts (from β= .60, p < .001, to β= .39, p < .01, Sobel's Z= 2.66, p < .05).
Self-Discrepancy and Social Comparison: Moderated Mediation
Because self-discrepancy moderated the effects of thin-ideal exposure on many of the negative outcomes, differences in the mediating effect of social comparison across levels of self-discrepancy (moderated mediation) were examined. Moderated mediation is similar to mediation except that interactions of the predictor variables are included in the analyses. Regression analyses were conducted to examine whether body image self-discrepancy moderated the mediation of social comparison in the relationship between exposure to thin-ideal advertisements and negative effects (i.e., if social comparison mediation is stronger for women with high levels of self-discrepancy as compared to women with low levels of self-discrepancy). Note that in mediation, there are two paths: one from condition to social comparison and one from social comparison to negative outcomes (see Figure 1). The moderation of both of these effects by self-discrepancy was examined.
First, the effect of self-discrepancy on the relationship between exposure to thin-ideal advertisements and social comparison was tested. BMI (as covariate) was entered as Step 1, then condition, self-discrepancy, and the condition × self-discrepancy interaction. Support for moderation was found; the condition × self-discrepancy interaction was significant, β= .17, p < .05. The results indicate that low self-discrepant participants reacted to exposure to thin-ideal advertisements differently from high self-discrepant participants; high self-discrepant participants were nearly twice as likely to engage in social comparison to the thin-ideal advertisements (mean difference = 2.29) as low self-discrepant participants (mean difference = 1.39).
Second, the role of self-discrepancy on the effects of social comparison was examined. It was also determined whether the effect of the moderation of thin-ideal exposure by self-discrepancy is explained by the mediation of social comparison. Mediators can be obscured by moderators, such that responses are cancelled out by opposite reactions. Therefore, the following analyses tested the effects of condition, social comparison, self-discrepancy, and the interaction of social comparison and self-discrepancy on all outcome variables that were affected by exposure to thin-ideal advertisements: weight-related thoughts, weight-regulatory thoughts, depression (ATQ), total self-esteem, appearance self-esteem, and mood. As predicted, the social comparison × self-discrepancy interaction was significant for weight-regulatory thoughts (β= .20, p < .01), depression (β= .22, p < .05), and negative mood (β= .26, p < .01), but not for total self-esteem (β=−.15), appearance self-esteem (β=−.13), or weight-related thoughts (β= .08). High self-discrepant participants experienced a much larger increase in weight-regulatory thoughts from comparison to the thin-ideal than low self-discrepant participants (B= .949 vs. B= .157, respectively). High self-discrepant participants experienced an increase in depressive thoughts from social comparison to the thin-ideal (B= 1.48), whereas there was no relation between social comparison and depression for low self-discrepant participants (B= .00). Although not significant, results for self-esteem were in the same direction; social comparison led to a decrease in self-esteem for high self-discrepant participants (B=−.17) but had no effect on self-esteem for low self-discrepant participants (B= .00). For negative mood, high and low self-discrepant participants reacted differently; for high self-discrepant participants, comparison to thin-ideal advertisements led to an increase in negative mood (B= .14), whereas for low self-discrepant participants, this comparison led to a decrease in negative mood (or increase in positive mood, B=−.11). See Figure 2 for path model inclusive of all difference scores from the regression analyses.
Also of interest was the effect of self-discrepancy on the negative effects from social comparison, controlling for its effects on social comparison from exposure to thin-ideal advertisements. The interaction of condition × self-discrepancy was reduced to nonsignificance when entered into the above models. Thus, the moderating effects of self-discrepancy on the relationship between thin-ideal exposure and the negative outcomes were explained by social comparison processes.
Exposure to thin-ideal advertisements affected weight concerns, mood, self-esteem, and depression. As predicted, body image self-discrepancy generally moderated these effects. Women with high levels of body image self-discrepancy experienced higher levels of dejection- and agitation-related mood; lowered self-esteem, particularly in the domain of appearance; and increased levels of depressive thoughts when they viewed thin-ideal advertisements than when they viewed advertisements that did not depict the thin ideal. Importantly, women with low levels of body image self-discrepancy did not show these differences between conditions. Thus, it appears that women with high body image self-discrepancy are at greater risk for negative consequences from exposure to thin-ideal media.
As predicted, social comparison processes mediated the effect of exposure to thin-ideal advertising on symptoms of depression, weight-related thoughts, and weight-regulatory thoughts.5 Notably, body image self-discrepancy moderated this effect. In the thin-ideal exposure condition, women with high body image self-discrepancy were almost twice as likely to engage in social comparison than women with low body image self-discrepancy. Furthermore, high self-discrepant women were more likely to experience depressive symptoms, negative mood, and weight-regulatory thoughts from this comparison. It is also important to note that the mediating effect of social comparison on mood was obscured by differing reactions from women with high levels of body image self-discrepancy and women with low levels of body image self-discrepancy. Interestingly, whether women felt better or worse after viewing thin-ideal advertisements depended on level of self-discrepancy. Although women with high levels of self-discrepancy experienced the predicted lowering of mood from comparison to the thin ideal, women with low levels of self-discrepancy felt better after this comparison. Such self-enhancement effects have been demonstrated in other studies (e.g., Henderson-King & Henderson-King, 1997; Myers & Biocca, 1992; Wilcox & Laird, 2000) and may be due to assimilation processes (Mussweiler & Strack, 2000), perhaps caused by identification with the model (Wilcox & Laird, 2000).
In addition, women with high levels of body image self-discrepancy were more likely to think about weight-reduction behaviors from exposure to thin-ideal advertisements. This result supports Scheier and Carver's (1988) theory that associations with the self can induce social comparison and self-regulatory processes. Also, self-discrepancies have been found to be linked with various types of eating disorders, such as anorexia and bulimia, which include extreme weight-related regulatory behaviors such as dieting, purging, and extreme exercise regimes (e.g., Cash & Green, 1986; Snyder, 1997; Strauman et al., 1991; Szymanski & Cash, 1995). Perhaps these tendencies are brought to light when norms related to weight are made salient (e.g., when exposed to thin-ideal media). When failure to live up to one's ideal becomes salient, this should motivate the individual to change her behavior to reduce the discrepancy. In the domain of weight, this should lead to actions that will allow one to become thinner, such as diet and exercise.
The current study also replicated previous findings that exposure to thin-ideal media leads to a general increase in body dissatisfaction (e.g., Cusumano & Thompson, 1987; Harrison & Cantor, 1997; Heinberg et al., 1995); however, there was no evidence for increases in the cognitive (drive for thinness) or behavioral (bulimia) aspects of eating disorder symptomatology. Perhaps affective dimensions (body dissatisfaction) may be more reactive to norm salience than cognitive or behavioral aspects. Surprisingly, the predicted interaction of exposure and self-discrepancy was not found for any of the three eating disorder subscales. However, exposure to thin-ideal advertisements did increase responses to the weight-related and weight-regulatory thoughts questions on the extent thoughts questionnaire, and this effect was also moderated by body image self-discrepancy. Thus, although exposure to thin-ideal media does appear to increase body and weight-related concerns, perhaps the specific content of these concerns may manifest in different ways. Moreover, the EDI may be less vulnerable to transitory states than weight-related thoughts. These results suggest the importance of studies to examine a wide range of weight concerns in addition to body dissatisfaction.
The main effects of self-discrepancy suggest that women with high body image self-discrepancy have lower appearance self-esteem and greater weight-related concerns in general (including possible eating disorder symptomatology) than women with lower levels of this self-discrepancy. This finding is consistent with theories of self-discrepancy and self-awareness, in addition to much of the literature linking self-discrepancies with emotional distress (e.g., Higgins, 1989; Moretti & Higgins, 1990; Strauman, 1989), as well as eating disorders (Snyder, 1997; Strauman et al., 1991; Szymanski & Cash, 1995).
This research provides support for an individual difference variable, body image self-discrepancy, which moderates the mediating effect of social comparison from exposure to thin-ideal media. High body image self-discrepant women were more likely to engage in social comparison processes from exposure to thin-ideal media, as well as more likely to experience from these comparison processes self-directed negative consequences such as negative mood, depressive thoughts, and thoughts about diet and exercise practices.
One limitation of the current study is the loss of 15 participants. It is not completely clear why 15 participants caught on to the purpose of the study. All were in the experimental thin-ideal condition and half possessed high levels of self-discrepancy. One possible explanation is that other studies regarding body image (e.g., self-objectification) were being conducted at the same time as the current study. Perhaps some participants had participated in these studies, and thus were no longer naïve participants. A second possibility is that the cover story was thin and easily discernible: The rationale given was an examination of “different types of feelings and thoughts advertisements can invoke in their audience.” Perhaps this justification was too obvious. Nevertheless, when all 112 participants were included in the analyses, results were in the same direction, although not as strong.
A second limitation is that the participants were primarily college-aged White females, limiting the generalizability of findings in terms of age, educational level, race, and culture. For example, Tiggemann and Lynch (2001) found that older women, as compared to younger women, have less weight-related anxiety and fewer eating disorder symptoms, although equivalent body dissatisfaction. In addition, Quinn and Crocker (1998) found that African American women have higher self-esteem and fewer body image problems than their Caucasian counterparts.
An additional limitation of this study is that the instructions may have made the self relevant for all participants. While examining the advertisements, participants in the thin-ideal condition were asked to think about themselves in relation to the clothing depicted in the advertisements; for example, they were asked to rate whether they would look good in the clothing being advertised. Thus, the self-concept may have been activated in relation to the thin-ideal standard for all participants. This might explain why even women with low levels of self-discrepancy engaged in significantly more social comparison processes from exposure to the thin-ideal advertisements as compared to controls. Further research should examine chronic versus temporary activation of the self in these effects.
Conclusions and Future Directions
The emotional and behavioral dysfunction associated with unattainable norms, such as those surrounding weight and appearance, makes it clear that it is important to understand the mechanisms through which those norms can affect the self. The current study extends the literature by considering the role of the self-concept in the negative consequences from exposure to advertising depicting the thin ideal. The findings not only confirm that social comparison processes play a part in these effects, but also that certain individuals, namely those who have self-failure associated with this particular domain (i.e., possess a self-discrepancy), are more likely to engage in these comparison processes and are more susceptible to negative effects from this comparison. Future exploration of these issues should expand on interventions aimed to circumvent social comparison to media; for example, women trained in media analysis techniques are less likely to engage in social comparison and experience significantly less weight concern than participants who receive a control intervention before exposure to media images (Posavac, Posavac, & Weigel, 2001). However, because these comparison processes may also be automatically activated, perhaps other methods of intervention should be sought as well.
It is not enough to relegate blame to the media, for exposure alone cannot explain such self-directed consequences (e.g., Heinberg et al., 1995; Stice et al., 1994). It is also not enough just to oppose the norm consciously; although we may not endorse a particular idea, we may still be significantly affected by it (e.g., Bessenoff & Sherman, 2000; Dovidio, Kawakami, Johnson, Johnson, & Howard, 1997). Self-discrepancy theory may provide an advance in understanding the mechanisms underlying internalization and use of social norms and how injunctive norms, such as the weight and appearance norms examined in the current study, can so strongly influence self-evaluation and dysfunctional behavior.
The pretest also showed that the thin-ideal advertisements were rated as significantly less interesting and appealing than the non–thin-ideal (neutral) advertisements (mean ratings = 4.33/4.18 for thin-ideal ads vs. 5.31/5.20 for neutral ads, on a scale ranging from 1 (not at all interesting/appealing) to 9 (very interesting/appealing), t(31) > 4.54, p < .01). It is conceivable that these differences could have contributed to the results. Although this difference might account for the mood effects, it is not clear why less appealing advertisements would affect women's self-evaluations. In addition, only high body image self-discrepant women demonstrated these negative effects from the thin-ideal advertising. Thus, it is highly unlikely that this factor can account for the data.
The institutional review board requested alteration to one of the items on the Hollon and Kendall (1980) questionnaire. The item “It's just not worth it” was construed as inappropriate and too closely related to suicide, and thus was changed to “I'm unattractive” (Rienecke-Hoste, 2001). The internal consistency of the scale inclusive of the substituted item is still quite high (alpha of .93).
As seen in Table 3, univariate analyses also revealed a significant condition × self-discrepancy interaction for Bulimia.
These analyses examine data from the appearance self-esteem subscale three separate times—when examining total self-esteem (which is a composite score of all three subscales), the MANCOVA on the three subscales, and the ANCOVA on the appearance subscale—thus increasing the risk of Type I error.
Although the current study was intended to examine the extent of social comparison from thin-ideal media effects without explicit social comparison instructions, participants were asked to imagine themselves in the clothing, which may have implicitly primed social comparison (Tiggemann & McGill, 2004).