Body Image 19 (2016) 37–43

Contents lists available at ScienceDirect

Body Image
journal homepage: www.elsevier.com/locate/bodyimage

Attractive celebrity and peer images on Instagram: Effect on women’s

mood and body image
Zoe Brown ∗ , Marika Tiggemann
School of Psychology, Flinders University, Adelaide, Australia

a r t i c l e i n f o a b s t r a c t

Article history: A large body of research has documented that exposure to images of thin fashion models contributes
Received 3 March 2016 to women’s body dissatisfaction. The present study aimed to experimentally investigate the impact of
Received in revised form 19 August 2016 attractive celebrity and peer images on women’s body image. Participants were 138 female undergrad-
Accepted 19 August 2016
uate students who were randomly assigned to view either a set of celebrity images, a set of equally
Available online 3 September 2016
attractive unknown peer images, or a control set of travel images. All images were sourced from public
Instagram profiles. Results showed that exposure to celebrity and peer images increased negative mood
Keywords:
and body dissatisfaction relative to travel images, with no significant difference between celebrity and
Body image
Celebrity
peer images. This effect was mediated by state appearance comparison. In addition, celebrity worship
Peer moderated an increased effect of celebrity images on body dissatisfaction. It was concluded that exposure
Instagram to attractive celebrity and peer images can be detrimental to women’s body image.
Media © 2016 Elsevier Ltd. All rights reserved.
Social comparison

Introduction Celebrities are well known individuals who receive excessive

public attention, typically for their involvement in entertainment
While there are a number of sociocultural factors that influ- or sport. Celebrities are prevalent in all forms of media, from
ence body image, including family and peers (Thompson, Heinberg, magazines and television to Facebook and Instagram. It has been
Altabe, & Tantleff-Dunn, 1999), it has been argued that the most suggested that the constant presentation throughout the media of
pervasive and influential are the mass media (Grabe, Ward, & thin and attractive female celebrity images, just like fashion mod-
Hyde, 2008; Groesz, Levine, & Murnen, 2001; Levine & Murnen, els, reinforces an unachievable thin ideal (Maltby, Giles, Barber,
2009; Thompson et al., 1999; Tiggemann, 2011; Want, 2009). & McCutcheon, 2005). Supporting this suggestion, a recent trend
Women encounter media images of thin and attractive models in cosmetic surgery sees women requesting specific body parts
and celebrities on a daily basis. Meta-analyses of both experi- of celebrities, for example, Kim Kardashian’s eyes and jawline
mental and correlational studies have concluded that there is a (PRNewswire, 2015). A few correlational studies have investigated
consistent and significant small-to-medium negative effect of thin some aspects of celebrities and their relationship to body image.
ideal media exposure on women’s body image (Grabe et al., 2008; Shorter, Brown, Quinton, and Hinton (2008) found that the discrep-
Groesz et al., 2001; Want, 2009). In addition, research has found ancy between participants’ views of their own size and their views
that thin ideal media has a similar negative effect on women’s of celebrities’ sizes was positively correlated with disordered eat-
mood (Bessenoff, 2006; Tiggemann & McGill, 2004; Tiggemann & ing. Harrison (1997) similarly found that interpersonal attraction
Polivy, 2010; Tiggemann, Polivy, & Hargreaves, 2009). However, to thin celebrities, rather than to average or heavy celebrities, was
the existing experimental research has primarily examined expo- positively associated with eating disorder symptoms.
sure to fashion models in magazines or television commercials. However, as yet no studies have experimentally investigated the
Although celebrities also feature heavily in contemporary media, effect of celebrity exposure on body image. Rather, experimental
as yet there has been little experimental research examining the research has predominately examined the effect of fashion models.
effect of attractive celebrities on women’s mood and body image. Celebrities are different from fashion models because they are not
anonymous figures and their personal lives are widely publicised
in the media. Consequently, people often form strong connections
with celebrities and sometimes begin to imagine having special
∗ Corresponding author at: School of Psychology, Flinders University, GPO Box relationships with them, a connection referred to as celebrity wor-
2100, Adelaide 5001, South Australia, Australia. ship (Maltby et al., 2005). A small amount of research has found
E-mail address: zoe.brown@flinders.edu.au (Z. Brown).

http://dx.doi.org/10.1016/j.bodyim.2016.08.007
1740-1445/© 2016 Elsevier Ltd. All rights reserved.
38 Z. Brown, M. Tiggemann / Body Image 19 (2016) 37–43

a significant relationship between celebrity worship and negative As of September 2015, Instagram had more than 400 million
body image (Maltby et al., 2005; Swami, Taylor, & Carvalho, 2011). active users (Kharpal, 2015). Instagram is a unique social network-
The detrimental effect of exposure to thin idealised media ing site because its prime focus is on sharing images. Research
images on women’s body satisfaction has generally been attributed shows that individuals generally choose to only upload images in
to the process of social comparison (Levine & Murnen, 2009; which they look thin and attractive (Manago, Graham, Greenfield,
Want, 2009). Social comparison theory (Festinger, 1954) argues & Salimkhan, 2008). Instagram users can also edit and filter their
that humans have an innate drive to compare themselves with oth- photographs to achieve an “ideal” look, and then share them with
ers in order to make evaluations about their abilities and opinions, their friends or the public. This creates a new opportunity for users
especially when objective measures do not exist. When compar- to be exposed to idealised images of their peers. As Instagram
ing their appearance with that of thin and attractive models, most users on average follow 100–200 people (Statista, 2015), it is likely
women will inevitably fall short, resulting in feeling negative about that the followed people will include both close friends and peo-
themselves and their bodies. Indeed, experimental research has ple they barely know. The participants in Fardouly and Vartanian’s
confirmed that the observed negative effect of media exposure (2015) study reported that they compared their appearance on
on negative mood and body dissatisfaction is at least partly medi- social media more with distant peers than with friends or celebri-
ated by social comparison processing (Bessenoff, 2006; Tiggemann ties. As yet, no studies have experimentally examined the effect of
& McGill, 2004; Tiggemann & Polivy, 2010; Tiggemann et al., Instagram images of unknown attractive peers on mood and body
2009). satisfaction.
Social comparison theory also suggests that comparisons will be Although Instagram’s prime focus is on peers, in fact celebrities
most readily made with targets who are perceived as similar to one- hold the top followed accounts (The Telegraph, 2016). Celebrities
self, as this comparison provides the most information. In support, use Instagram to share pictures of their personal and professional
research has found that women report comparing more with their lives with their fans. As of May 2016, the top 8 out of 10 most
peers than with thin ideal media (Heinberg & Thompson, 1992; followed Instagram users were thin, toned, and attractive female
Wheeler & Miyake, 1992) and that these peer comparisons result in celebrities (Justin Bieber and Cristiano Ronaldo being the only two
greater thoughts about dieting than comparisons to media (Leahey men in the top 10, both of whom are also thin and toned). Selena
& Crowther, 2008). On the other hand, Ridolfi, Myers, Crowther, and Gomez has the most followers on Instagram, with more than 79.2
Ciesla (2011) and a meta-analysis by Myers and Crowther (2009) million, followed by Taylor Swift with 77.3 million followers, and
found that there was no difference between comparison to peers Ariana Grande with 72 million followers (The Telegraph, 2016).
and thin media images. Some recent studies have also included The combination of celebrity images with peer images, presented
celebrities as a comparison category. Strahan, Wilson, Cressman, equally on the same platform, may result in users viewing celebri-
and Buote (2006) found women reported that they compared them- ties more like their peers.
selves to peers when evaluating their social skills but to fashion The main aim of the present study was to examine the effect
models and celebrities when evaluating their appearance. Likewise, of celebrity images, compared with equally attractive (unknown)
Jones (2001) found that girls reportedly compared more to peers peer, and control (travel) images, on women’s body image. Travel
for height and personality, but equally to celebrities/fashion models images were chosen because they are also a common type of
and peers for weight and shape. Strahan et al. (2006) suggest that image posted on Instagram. We predicted that exposure to celebrity
women will make comparisons to models or celebrities when they images would result in greater negative mood and body dissatis-
are confronted with cultural norms of thinness because models and faction compared to exposure to travel images. We also tentatively
celebrities set the standard for these norms and women know that predicted that exposure to attractive peer images would result in
they will be judged against them. To provide further insight into this greater negative mood and body dissatisfaction compared to expo-
issue, the present study included a comparison group of unknown sure to celebrity images. In addition, we expected that the effects
peers to determine whether the effect of comparison with celebri- of image type would be mediated by state appearance comparison.
ties differs from comparison with equally attractive peers. Lastly, we hypothesised that the effect of image type on negative
The great majority of research in the field of media and body mood and body dissatisfaction would be moderated by celebrity
image has examined traditional media formats such as television worship. Specifically, the effect of exposure to celebrity images was
and magazines. However, the rise of the Internet, in particular expected to be greater for participants high on celebrity worship.
the use of social networking sites, has led researchers to begin to
examine the effect of Internet sites on women’s body image. Cor-
relational research has found that time spent on the Internet, and Method
more specifically time spent on social networking sites, is positively
correlated with negative mood (Fardouly, Diedrichs, Vartanian, & Participants
Halliwell, 2015) and body dissatisfaction (Fardouly & Vartanian,
2015; Tiggemann & Miller, 2010; Tiggemann & Slater, 2013, 2014). Participants were 138 female undergraduate students from
One experimental study has found that women’s preoccupation Flinders University, South Australia. They were aged between 18
with their weight and shape decreased at a lower rate after using and 30 years, with a mean age of 20.10 (SD = 2.61), and were pre-
Facebook than after viewing a control website (Mabe, Forney, & dominantly Caucasian (77.4%). Their mean body mass index (BMI)
Keel, 2014). Further, a more detailed examination of Facebook use of 22.61 (SD = 4.35) was within the “normal weight” range (Garrow
by Meier and Gray (2014) concluded that it was time spent interac- & Webster, 1984). Participants received course credit for their par-
ting with the photographic functions of Facebook, rather than total ticipation.
Facebook time, that was related to body dissatisfaction. Thus, one of
the aims of the present study was to contribute to social media and
body image research by using the photo-based site of Instagram Design
to source stimulus images. To our knowledge, the only experimen-
tal body image research to use Instagram images (Tiggemann & A between-subjects experimental design was used to inves-
Zaccardo, 2015) found that exposure to fitspiration images leads to tigate the effect of image type (celebrity, peer, travel) on the
increased negative mood and body dissatisfaction in comparison to dependent variables of negative mood and body dissatisfac-
travel images. tion (after controlling for baseline scores). In addition, state
 Z. Brown, M. Tiggemann / Body Image 19 (2016) 37–43 39

appearance comparison was tested as a mediating variable and Negative mood and body dissatisfaction. Following Heinberg
celebrity worship was tested as a potential moderating variable. and Thompson (1995), visual analogue scales (VAS) were used
to measure state mood and body dissatisfaction before and after
viewing the images. There were five mood dimensions (anx-
Materials
iety, depression, happiness, anger, and confidence) and three
body dissatisfaction dimensions (weight dissatisfaction, appear-
Stimulus materials. Three sets of images were constructed for
ance dissatisfaction, and facial features dissatisfaction). The facial
the study: the celebrity condition contained 15 attractive celebrity
component was added as faces are an important aspect of Insta-
images, the peer condition contained 15 attractive peer images,
gram images. The scales consisted of 100-mm horizontal lines
and the travel condition contained 15 images of travel destinations.
(0 = none, 100 = very much). Participants were instructed to indicate
Each image set consisted of three photographs of five different peo-
how they feel “right now” by placing a mark at the appropriate point
ple or places. For the celebrity and peer image sets, the photographs
of each line. The scales were measured from the ‘none’ anchor to the
included only the target individual and ranged from full body shots
nearest millimetre. The overall scores for negative mood and body
to close up shots of the person’s face, as these are common types of
dissatisfaction were calculated by averaging the items related to the
photographs found on Instagram. The women in the celebrity and
construct (some reverse coded). This produced a scale ranging from
peer images were of similar age to the participants and wearing rel-
0 (no negative mood/no body dissatisfaction) to 100 (high negative
atively tight or revealing clothing. Five of the travel images included
mood/high body dissatisfaction). Heinberg and Thompson (1995)
a person to justify the appearance comparison measures for par-
showed that VAS are a valid and reliable measure of mood and body
ticipants in that condition. The people in the travel images were
dissatisfaction. In the current study, the internal reliability was high
of similar age to the participants, wearing non-revealing clothing,
for both negative mood (˛ = .85) and body dissatisfaction (˛ = .94).
and were not the focus of the image.
All the stimulus material photos were taken from public Insta-
gram profiles, but without the Instagram border or associated State appearance comparison. The State Appearance Compar-
comments. The five celebrities (Kim Kardashian, Taylor Swift, ison Scale (Tiggemann & McGill, 2004) was used to examine the
Selena Gomez, Rihanna, and Beyoncé) were chosen from the top amount of appearance comparison participants engaged in. Par-
10 most followed celebrities on Instagram. The five travel destina- ticipants were asked to indicate on three 7-point Likert scales
tions (Australia, England, France, Germany, and Egypt) were found how much they thought about their appearance when viewing
using the #travel hashtag. The five peers were found on Instagram the images (1 = no thought about appearance, 7 = a lot of thought),
using the #selfie hashtag and were chosen to be equally attractive and the extent to which they compared their overall appearance
as the celebrities and to have less than 200 followers (to ensure and specific body parts with the people they saw in the images
that they were not models/celebrities). Images were presented (1 = no comparison, 7 = a lot of comparison). A total score for state
to participants on an Apple iPad via a slideshow using the Apple appearance comparison was calculated by averaging the items.
“Photos” application. A title slide with the name of the person or Items in this scale have been shown to be highly inter-correlated
place appeared for 10 seconds before each set of three images. The (r = .71–.82) (Tiggemann & McGill, 2004). In the current sample, the
participants viewed each image for 10 seconds. measure had good internal reliability (˛ = .94).
The stimulus materials were chosen from a pool of 40 celebrity,
40 peer, and 40 travel images. A pilot study was conducted with Celebrity worship. The 17-item version of the Celebrity Atti-
five independent female raters from the target age group (M = 22 tude Scale (CAS) developed by McCutcheon, Lange, and Houran
years, SD = 1.22) to test the attractiveness and thinness of the (2002) was used to measure participants’ level of celebrity worship.
celebrity and peer images, to equate the image sets on overall The scale was adapted for the present study to measure worship of
visual appeal, and to test whether the celebrities and travel destina- celebrities in general rather than worship of a specific celebrity.
tions were well known. Participants rated attractiveness (1 = very Using a 5-point scale (1 = strongly disagree, 5 = strongly agree), par-
unattractive, 5 = very attractive), thinness (1 = very heavy, 5 = very ticipants were asked to indicate their views about celebrities. Items
thin), and visual quality (1 = very poor, 5 = excellent) using 5-point ranged from a general interest in celebrities to borderline patho-
Likert scales. The final five celebrities and the five travel desti- logical (e.g., “My friends and I like to discuss what celebrities have
nations were chosen because the results showed that 100% of done” [Item 2]; “When my favourite celebrity dies I will feel like
participants knew who/where they were. None (0%) of the partici- dying too” [Item 9]). A total score for celebrity worship was cal-
pants recognised the chosen peers. Although the celebrity images culated by averaging the items. The CAS has been shown to have
were marginally more attractive (celebrity: M = 4.00, SD = 0.00; good internal consistency and convergent validity (McCutcheon
peer: M = 3.92, SD = 0.27) and thin (celebrity: M = 4.08, SD = 0.11; et al., 2002). In the current sample, the measure had high internal
peer: M = 3.92, SD = 0.36) than the peer images, the differences reliability (˛ = .91).
were not significant (attractive: t(4) = 0.67, p = .54, d = 0.42; thin:
t(4) = 1.00, p = .37, d = 0.60). Lastly, there were no significant differ- Manipulation check. Each set of three images was preceded
ences between the rated quality of the travel (M = 3.37, SD = 0.42), by title cards with the name of the person or place. At the end of
celebrity (M = 3.49, SD = 0.53), and peer images (M = 3.59, SD = 0.45), the study, participants rated how many celebrities, peers, or places
F(2, 135) = 2.55, p = .82. they had recognised (0–5) and how many of their names they had
heard previously (0–5). Participants were also asked to recall the
Social networking site usage. Participants were asked five names of the people or places shown in the images. The rea-
whether or not they had an Instagram account, how much time soning was that if celebrities were well known, their names should
they spent using Instagram per day (less than 10 min, 10–30 min, be better remembered than those of the unknown peers. This recall
31–60 min, 1–2 hrs, over 2 hrs), and how many people/celebrities task was used as an objective measure of whether the experimental
they follow on Instagram. Participants also indicated how much manipulation was successful.
importance they placed on the quality of a photograph (e.g., blur-
riness, composition) when posted by others and by themselves Demographic information. Participants were asked their age
on two 7-point Likert scales (1 = not important at all, 7 = extremely and ethnicity. Measures of participants’ current weight and height
important). were also taken to calculate body mass index (BMI).
40 Z. Brown, M. Tiggemann / Body Image 19 (2016) 37–43

Procedure Table 2
Means (SD) for manipulation check (recognition, heard of name, and correct recall)
by image type.
After obtaining ethics approval from the institutional review
committee, participants were recruited for a study entitled “Insta- Image type
gram Study: People and Places”. Participants were randomly Celebrity Peer Travel
assigned to one of three experimental conditions (celebrities, peers,
Recognition 4.26 (1.54) 0.24 (0.60) 3.69 (1.29)
or travel), subject to equal numbers (n = 46) in each experimen- Heard of name 4.71 (1.04) 0.30 (0.59) 4.44 (1.08)
tal condition. Participants were tested individually or in small Correct recall 4.63 (0.97) 0.98 (0.88) 4.13 (0.98)
groups of two or three, and testing sessions lasted approximately
30 minutes. After reading the Letter of Introduction, Information
Sheet, and signing the consent form, participants completed the Table 2 also shows that at the end of the study participants who
social networking use questions and pre-exposure measures of viewed celebrity images recalled almost all of the names of the five
mood and body dissatisfaction. Participants were then provided celebrities (M = 4.63, SD = 0.97), in contrast to those who viewed
with an iPad and told that they would be watching a slideshow peer images, who recalled less than one of the names of five peers
of typical images from Instagram. As each photo appeared, par- (M = 0.98, SD = 0.88). A t-test confirmed that celebrity names were
ticipants were asked to rate the quality of the photo using a significantly better recalled than peer names, t(89) = 18.62, p < .001,
5-point scale (1 = very poor quality to 5 = very good quality). This d = 3.94. Thus, on all three measures, the experimental manipula-
was to ensure that participants attended to the images. In all condi- tion was judged as successful.
tions, participants viewed 15 images for 10 seconds each. Following
exposure to the Instagram images, participants completed post- Instagram Usage
exposure measures of mood and body dissatisfaction, as well as
measures of state appearance comparison and celebrity worship. The majority of participants reported that they had an
Finally, participants completed a manipulation check and then their account on Instagram (86.2%). Their median use of Instagram was
height and weight were measured. 10–30 minutes per day. They reported following a median number
of 100–200 people on Instagram, of which the median number of
Results celebrities was 5–10.

Characteristics of the Sample Effect of Image Type on Negative Mood and Body
Dissatisfaction
The three experimental image conditions were compared to
determine if there were any differences in initial baseline char- In order to determine whether there were significant group
acteristics. As can be seen in Table 1, there were no significant differences on negative mood and body dissatisfaction, ANCOVAs
differences between groups on the demographic variables of age, with two orthogonal planned comparisons were performed using
F(2, 134) = 0.68, p = .51, 2p = .01, or BMI, F(2, 126) = 0.63, p = .53, the Lmatrix subcommand. For each ANCOVA, pre-exposure scores
2p = .01. Nor was there any significant difference on initial nega- were entered as the covariate to control for individual differences.
The first orthogonal planned comparison compared the combined
tive mood, F(2, 135) = 0.11, p = .90, 2p = .13, or body dissatisfaction,
celebrity and peer group against the travel group [contrast: +1
F(2, 135) = 0.72, p = .49, 2p = .01. Thus, random allocation of partici-
+1 −2]. The second orthogonal planned comparison compared the
pants to the three image groups was deemed successful. Table 1 also
celebrity and peer image groups [contrast: +1 −1 0].
shows that the groups did not differ significantly on celebrity wor-
Table 3 shows the adjusted means (controlling for pre-
ship, F(2, 135) = 0.66, p = .52, 2p = .01. This indicates that celebrity
exposure scores) for negative mood and body dissatisfaction. The
worship was not affected by the experimental manipulation.
first planned comparison showed that participants who viewed
celebrity and peer images experienced significantly greater neg-
Manipulation Check ative mood, F(1, 134) = 10.76, p < .001, 2p = .08, and greater body
dissatisfaction, F(1, 134) = 5.90, p = .02, 2p = .04, than partici-
Table 2 shows that participants who viewed celebrity images
pants who viewed travel images. The second planned comparison
recognised and had heard of more than four of the five celebrities.
showed that there was no significant difference between the
In contrast, peer condition participants recognised and had heard
celebrity and peer images on negative mood, F(1, 134) = 0.61, p = .44,
of less than one of the five peers shown. Independent samples t-
d = 0.08, or body dissatisfaction, F(1, 134) = 0.01, p = .94, d < 0.01.
tests confirmed that the celebrities were recognised, t(90) = 16.73,
In support of our prediction, the results showed that in com-
p < .001, d = 3.44, and had been heard of significantly more than the
parison to travel images, participants exposed to celebrity images
peers, t(89) = 24.98, p < .001, d = 5.22.
reported greater negative mood and body dissatisfaction. In con-
trast to our prediction, however, there was no difference between
Table 1 celebrity and peer images on negative mood and body dissatisfac-
Means (SD) for age, BMI, initial negative mood, initial body dissatisfaction, and tion.
celebrity worship by image type.

Image type Table 3

Means (SD) for negative mood, body dissatisfaction, and state appearance compar-
Celebrity Peer Travel
ison by image type.
Demographic
Age 19.72 (2.43) 20.18 (2.61) 20.33 (2.80) Image type
BMI 22.00 (3.62) 22.71 (4.53) 23.05 (4.78) Celebrity Peer Travel
Initial measures
Negative mood 27.22 (12.76) 26.30 (14.56) 27.55 (11.98) Negative mooda 28.11 (0.96) 29.18 (0.96) 24.80 (0.96)
Body dissatisfaction 48.41 (24.43) 42.85 (22.89) 43.84 (23.87) Body dissatisfactiona 45.76 (1.62) 45.94 (1.62) 41.04 (1.62)
Trait measures State appearance comparison 3.88 (1.79) 4.09 (1.72) 1.89 (1.84)
Celebrity worship 2.15 (0.62) 2.31 (0.67) 2.19 (0.73) a
Note. Adjusted means (SE).
 Z. Brown, M. Tiggemann / Body Image 19 (2016) 37–43 41

The Role of State Appearance Comparison as a Mediator mood was significant (celebrity: B = 1.12, ˇ = .04, SE = 1.48, CI [−1.81,
4.04]; peer: B = 1.92, ˇ = .07, SE = 1.52, CI [−1.08, 4.93]). Therefore,
It was predicted that state appearance comparison would state appearance comparison mediated the effect of image type on
mediate the effect of image type on negative mood and body dis- negative mood.
satisfaction. Table 3 shows that participants who viewed celebrity Fig. 1(b) shows the same pattern of results for body dissatisfac-
and peer images engaged in more appearance comparison than tion. The indirect effects between both celebrity images, B = 4.19,
participants who viewed travel images, t(135) = 7.42, p < .001, ˇ = .08, SE = 1.21, CI [2.29, 7.17] and peer images, B = 4.99, ˇ = .10,
d = 1.28, with no difference between celebrity and peer condi- SE = 1.27, CI [2.79, 7.82], on body dissatisfaction via appearance
tions, t(135) = 0.67, p = .51, d = 0.12. Across groups, state appearance comparison were significant. Neither of the direct effects were
comparison was itself significantly correlated with post-exposure significant (celebrity: B = 0.53, ˇ = .01, SE = 2.49, CI [−4.39, 5.46];
negative mood (r = .24, p < .001) and post-exposure body dissatis- peer: B = −0.08, ˇ < −.01, SE = 2.60, CI [−5.22, 5.05]). Therefore, state
faction (r = .44, p < .001). appearance comparison mediated the effect of image type on body
The potential mediating role of state appearance comparison dissatisfaction.
was tested using the Preacher and Hayes’ (2008) PROCESS method.
This method is a bootstrapping non-parametric approach that Moderation by Celebrity Worship
allows more confident conclusions to be made regarding the signif-
icance of the direct and indirect effects. In this approach, effects are It was predicted that celebrity worship would moderate the
deemed significant if the 95% confidence interval does not include effect of image type. Significant relationships were found between
zero. For this analysis, indirect effects were computed for each of celebrity worship and pre-exposure dissatisfaction (r = .23, p = .01),
5000 bootstrapped samples. Using the PROCESS method, media- and state appearance comparison (r = .23, p = .01). To test whether
tion is established if the indirect effect of the independent variable celebrity worship moderated the effect of image type, Preacher and
on the dependent variable via the mediator is significant. If the Hayes’ (2008) PROCESS method was again used. As recommended
direct pathway is non-significant after controlling for the media- by Aiken and West (1991), celebrity worship was centred. Dummy
tor, this shows that the mediator significantly explains the effect of coding was again used for image type, with the travel condition
the independent variable on the dependent variable. Because the as the reference group. The two dummy variables were entered
independent variable of image condition had three levels (celebrity, as independent variables, pre-exposure scores were added as a
peer, and travel), dummy coding needed to be used. Two dummy covariate and celebrity worship was the moderator. Two product
variables (celebrity: 1, 0, 0; peer: 0, 1, 0) were computed with travel terms were created by the PROCESS macro to represent interactions
as the reference group. The two dummy variables were entered between the image types and celebrity worship. A significant inter-
as independent variables, pre-exposure scores were added as a action is found when the product term explains additional variance
covariate and state appearance comparison was the mediator. above the main effects.
Fig. 1(a) shows that the pathways between both image types The results showed that the interaction between celebrity wor-
and appearance comparison were significant, as was the pathway ship and image type did not explain significant additional variance
between appearance comparison and negative mood. The indirect 2
in negative mood (celebrity: Rchange = .01, Fchange (1, 131) = 2.43,
effect between celebrity images and negative mood via appearance 2
p = .12; peer: Rchange = .01, Fchange (1, 131) < 0.01, p = .97). How-
comparison was found to be significant, B = 2.20, ˇ = .10, SE = 0.11,
CI [0.93, 3.95], as was the indirect effect between peer images and ever, the interaction did explain significant additional variance in
2
body dissatisfaction for celebrity images (Rchange = .01, Fchange (1,
negative mood, B = 2.45, ˇ = .11, SE = 0.81, CI [1.12, 4.37]. Neither of
2
132) = 4.98, p = .03), but not peer images (Rchange < .01, Fchange (1,
the remaining direct effects of celebrity or peer images on negative
132) = 0.01, p = .94). Therefore, the effect of celebrity images on body
dissatisfaction was moderated by celebrity worship.
To determine the nature of the interaction, the results were
graphed (see Fig. 2). Following the recommendation of Aiken and
West (1991), low celebrity worship was calculated as one standard
deviation below the centred mean and high celebrity worship was
one standard deviation above the centred mean. Fig. 2 shows that
for participants who viewed travel images, celebrity worship lev-
els made little difference to their body dissatisfaction. However,

Fig. 1. Standardised regression coefficients for the effect of images on (a) negative
mood and (b) body dissatisfaction as mediated by state appearance comparison.
Note. The standardised regression coefficient in parentheses is the direct effect not Fig. 2. The effect of image type on body dissatisfaction moderated by celebrity
controlling for the mediator; *p < .05. worship.
42 Z. Brown, M. Tiggemann / Body Image 19 (2016) 37–43

for participants who viewed celebrity images, celebrity worship reports (Jones, 2001; Strahan et al., 2006). However, the finding
levels were positively associated with body dissatisfaction. Put dif- that participants compared their appearance similarly to celebri-
ferently, for participants low on celebrity worship, image type had ties as to peers runs counter to the logic of social comparison theory
little effect on their body dissatisfaction. But for participants high (Festinger, 1954), which would predict greater comparison with
on celebrity worship, image type made a large difference to their similar targets, in this case, peers. Perhaps attractive images, no
body dissatisfaction; in particular, celebrity images had the greatest matter what their source, set the standard against which women
effect. will be judged (Strahan et al., 2006). Relatedly, more complex
accounts of social comparison suggest that while women make
frequent comparisons (Leahey & Crowther, 2008; Ridolfi et al.,
Discussion 2011), they are often unintentional and automatic (Gilbert, Giesler,
& Morris, 1995; Want, 2009). Women may only consider the rel-
The present study adds to our knowledge of the effect of evance or appropriateness of such comparisons after they have
attractive celebrity and peer images on women’s body image and already been made. In addition, participants may have compared
also contributes to current understandings of social comparison similarly to celebrity and peer images because Instagram presents
processing and of celebrity worship. Importantly, it extends mass both on an equal platform. Instagram allows celebrities to post pic-
media research into Instagram, a visually based social media plat- tures about their private daily lives, which may make viewers feel
form with little current research. First, as predicted, exposure to more personally connected to them.
Instagram images of attractive celebrities led to greater negative The finding of a significant correlation between celebrity wor-
mood and body dissatisfaction than exposure to travel images. ship and pre-exposure body dissatisfaction supports a small
The same effect was found for attractive peer images. Second, the amount of previous research (Maltby et al., 2005; Swami et al.,
effect of image type on negative mood and body dissatisfaction was 2011). However, the finding that celebrity worship moderated
mediated by appearance comparison processing. Third, celebrity the effect of exposure to celebrity images on body dissatisfaction
worship was found to moderate the effect of exposure to celebrity is novel. Specifically, women who had higher levels of celebrity
images on women’s body dissatisfaction. worship felt more dissatisfied after viewing celebrity images than
The previous research on celebrity media has been primarily women with lower levels of celebrity worship. Thus, celebrity wor-
correlational in nature (Harrison, 1997; Shorter et al., 2008). To the ship may be conceptualised as an additional vulnerability factor
best of our knowledge, the present study represents the first exper- for body dissatisfaction and disordered eating. Most likely, women
imental examination of the effect of exposure to celebrity images high on celebrity worship consider celebrities to be both impor-
on women’s body image. The findings establish that acute exposure tant and self-relevant, and thus may feel more dissatisfied when
to thin and attractive female celebrity images has an immediate they themselves are not as attractive. In contrast, women low on
negative effect on women’s mood and body image. Interestingly, celebrity worship likely consider the lives and bodies of celebrities
exposure to images of thin and attractive unknown peers had a as not relevant to them.
similar detrimental impact. This finding is consistent with the find- The results of the current study have important practical impli-
ings from in vivo peer studies which also show negative effects cations. The findings suggest that women might be advised to limit
on women’s body dissatisfaction (Krones, Stice, Batres, & Orjada, their exposure to celebrity and peer images. However, because of
2005; Trottier, Polivy, & Herman, 2007; Wasilenko, Kulik, & Wanic, their abundance on social networking sites, this would present
2007). While celebrity stories and images are common content of a daunting challenge. Rather, the findings suggest that women
traditional media such as women’s magazines, peer images are should be educated about the potential dangers of exposure to
not. However, contemporary women are likely to be exposed to celebrity and peer images. Whereas celebrity images and fashion
an increasing number of attractive peer images with the rise of model images have generally been recognised as problematic, lit-
the Internet, particularly social networking sites. Many of these tle research or thinking has addressed the effect of peer images.
attractive peers are likely to be unknown, in contrast to previous Indeed, the recent increased access and exposure to multiple peer
research with known peers (Fardouly & Vartanian, 2015; Wheeler images afforded by social networking sites may constitute a sig-
& Miyake, 1992). The present study’s use of the social networking nificant shift in the media landscape. Accordingly, media literacy
site of Instagram afforded the opportunity to test the effect of both programmes, which have shown some success in combatting neg-
celebrity and unknown peer images on the same platform. For both, ative body image (Levine & Murnen, 2009; Wilksch et al., 2014),
the observed effect sizes (small to medium) are consistent with need to be expanded to include both celebrity and peer media on
meta-analytic findings from media research using fashion model the Internet and elsewhere.
images as stimuli (Grabe et al., 2008; Groesz et al., 2001; Want, As with all studies, the present findings should be interpreted
2009). These results suggest that exposure to thin ideal images in the context of some limitations. First, the sample consisted of
may have a similar negative effect on body satisfaction, regardless primarily white Australian university students. Thus, the findings
of the context or whose body is featured. Future research might may not generalise to other groups of women. Second, the study
explicitly investigate this suggestion. It is also important to note took place in a laboratory setting. Therefore, participants may have
that although individual effects may be small, many women are attended to the images differently from the way they would inter-
exposed to attractive celebrity, peer, and fashion images on a daily act with images in a more naturalistic environment. However, real
basis. While Instagram feeds may vary between users, our findings Instagram images were presented on an ecologically valid format
show that participants followed 100–200 people, including 5–10 (iPad). Third, the results may be subject to some demand effects
celebrities on Instagram. It is certainly possible that there will be and further research might use a stronger cover story. Fourth,
a larger cumulative detrimental impact of viewing these types of although the celebrity and peer images were equated on attractive-
images frequently. ness and thinness, it might be argued that the level of attractiveness
The finding that the effect of image type was mediated by state portrayed does not accurately represent the non-celebrity peer
appearance comparison extends previous research using fashion population. However, the peer images were pictures of real people
magazine advertisements (Bessenoff, 2006; Tiggemann & Polivy, taken from public Instagram accounts. Such attractive peer images
2010; Tiggemann et al., 2009) to celebrity and unknown peer are not uncommon due to the editing features of Instagram and
images. It also shows that women do compare to both celebri- other social networking sites, which allow people to post idealised
ties and peers when assessing appearance, consistent with some pictures of themselves online.
 Z. Brown, M. Tiggemann / Body Image 19 (2016) 37–43 43

Despite the above limitations, the present study has clearly Meier, E. P., & Gray, J. (2014). Facebook photo activity associated with body image
demonstrated that exposure to celebrity and peer images on Insta- disturbance in adolescent girls. Cyberpsychology, Behavior, and Social Networking,
17, 199–206. http://dx.doi.org/10.1089/cyber.2013.0305
gram can have a negative effect on women’s immediate mood and Myers, T. A., & Crowther, J. H. (2009). Social comparison as a predictor of body dissat-
body dissatisfaction. These findings considerably extend previous isfaction: A meta-analytic review. Journal of Abnormal Psychology, 118, 683–698.
research, which has largely investigated exposure to fashion model http://dx.doi.org/10.1037/a0016763
Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for
images in women’s fashion magazines, in terms of both content (to assessing and comparing indirect effects in multiple mediator models. Behavior
celebrities and peers) and format (to Instagram). Accordingly, the Research Methods, 40, 879–891. http://dx.doi.org/10.3758/brm.40.3.879
results illustrate the ever-expanding role played by the mass media PRNewswire. (2015). Annual AAFPRS survey reveals celebrity look-alike surgery on
the rise. Retrieved from http://www.prnewswire.com/news-releases/annual-
in contemporary body image.
aafprs-survey-reveals-celebrity-look-alike-surgery-on-the-rise-300024032.
html
References Ridolfi, D. R., Myers, T. A., Crowther, J. H., & Ciesla, J. A. (2011). Do appearance focused
cognitive distortions moderate the relationship between social comparisons to
peers and media images and body image disturbance? Sex Roles, 65, 491–505.
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interac-
http://dx.doi.org/10.1007/s11199-011-9961-0
tions. Thousand Oaks, CA: Sage Publications.
Shorter, L., Brown, S. L., Quinton, S. J., & Hinton, L. (2008). Relationships between
Bessenoff, G. R. (2006). Can the media affect us? Social comparison, self-discrepancy,
body-shape discrepancies with favored celebrities and disordered eating in
and the thin ideal. Psychology of Women Quarterly, 30, 239–251. http://dx.doi.
young women. Journal of Applied Social Psychology, 38, 1364–1377. http://dx.
org/10.1111/j.1471-6402.2006.00292.x
doi.org/10.1111/j.1559-1816.2008.00351.x
Fardouly, J., Diedrichs, P. C., Vartanian, L. R., & Halliwell, E. (2015). Social compar-
Statista. (2015). Average number of Instagram followers of teenage users in United States
isons on social media: The impact of Facebook on young women’s body image
as of March 2015. Retrieved from http://www.statista.com/statistics/419326/us-
concerns and mood. Body Image, 13, 38–45. http://dx.doi.org/10.1016/j.bodyim.
teen-instagram-followers-number/
2014.12.002
Strahan, E. J., Wilson, A. E., Cressman, K. E., & Buote, V. M. (2006). Comparing to
Fardouly, J., & Vartanian, L. R. (2015). Negative comparisons about one’s appearance
perfection: How cultural norms for appearance affect social comparisons and
mediate the relationship between Facebook usage and body image concerns.
self-image. Body Image, 3, 211–227. http://dx.doi.org/10.1016/j.bodyim.2006.
Body Image, 12, 82–88. http://dx.doi.org/10.1016/j.bodyim.2014.10.004
07.004
Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7,
Swami, V., Taylor, R., & Carvalho, C. (2011). Body dissatisfaction assessed by the
117–140. http://dx.doi.org/10.1177/001872675400700202
Photographic Figure Rating Scale is associated with sociocultural, personality,
Garrow, J. S., & Webster, J. (1984). Quetelet’s index (W/H2 ) as a measure of fatness.
and media influences. Scandinavian Journal of Psychology, 52, 57–63. http://dx.
International Journal of Obesity, 9(2), 147–153.
doi.org/10.1111/j.1467-9450.2010.00836.x
Gilbert, D. T., Giesler, R. B., & Morris, K. A. (1995). When comparisons arise. Journal of
The Telegraph. (2016). Most popular Instagram accounts: In pictures. Retrieved from
Personality and Social Psychology, 69, 227–236. http://dx.doi.org/10.1037/0022-
http://www.telegraph.co.uk/technology/2015/12/14/the-ten-most-popular-
3514.69.2.227
instagram-accounts-in-pictures/nicki-minaj-instagram-account/
Grabe, S., Ward, L. M., & Hyde, J. S. (2008). The role of the media in body image con-
Thompson, J. K., Heinberg, L. J., Altabe, M., & Tantleff-Dunn, S. (1999). Exacting beauty:
cerns among women: A meta-analysis of experimental and correlational studies.
Theory, assessment, and treatment of body image disturbance. Washington, DC:
Psychological Bulletin, 134, 460–476. http://dx.doi.org/10.1037/0033-2909.134.
American Psychological Association.
3.460
Tiggemann, M. (2011). Sociocultural perspectives on human appearance and body
Groesz, L. M., Levine, M. P., & Murnen, S. K. (2001). The effect of experimental pre-
image. In T. F. Cash & L. Smolak (Eds.), Body image: A handbook of science, practice,
sentation of thin media images on body satisfaction: A meta-analytic review.
and prevention (pp. 12–19). New York: Guilford.
International Journal of Eating Disorders, 31, 1–16. http://dx.doi.org/10.1002/eat.
Tiggemann, M., & McGill, B. (2004). The role of social comparison in the effect of
10005
magazine advertisements on women’s mood and body dissatisfaction. Journal
Harrison, K. (1997). Does interpersonal attraction to thin media personalities pro-
of Social and Clinical Psychology, 23, 23–44. http://dx.doi.org/10.1521/jscp.23.1.
mote eating disorders? Journal of Broadcasting & Electronic Media, 41, 478–500.
23.26991
http://dx.doi.org/10.1080/08838159709364422
Tiggemann, M., & Miller, J. (2010). The Internet and adolescent girls’ weight satis-
Heinberg, L. J., & Thompson, J. K. (1992). Social comparison: Gender, target impor-
faction and drive for thinness. Sex Roles, 63, 79–90. http://dx.doi.org/10.1007/
tance ratings, and relation to body image disturbance. Journal of Social Behavior
s11199-010-9789-z
and Personality, 7(2), 335–344.
Tiggemann, M., & Polivy, J. (2010). Upward and downward: Social comparison
Heinberg, L. J., & Thompson, J. K. (1995). Body image and televised images of thinness
processing of thin idealized media images. Psychology of Women Quarterly, 34,
and attractiveness: A controlled laboratory investigation. Journal of Social and
356–364. http://dx.doi.org/10.1111/j.1471-6402.2010.01581.x
Clinical Psychology, 14, 325–338. http://dx.doi.org/10.1521/jscp.1995.14.4.325
Tiggemann, M., Polivy, J., & Hargreaves, D. (2009). The processing of thin ideals in
Jones, D. C. (2001). Social comparison and body image: Attractiveness comparisons
fashion magazines: A source of social comparison or fantasy? Journal of Social
to models and peers among adolescent girls and boys. Sex Roles, 45, 645–664.
and Clinical Psychology, 28, 73–93. http://dx.doi.org/10.1521/jscp.2009.28.1.73
http://dx.doi.org/10.1023/A:1014815725852
Tiggemann, M., & Slater, A. (2013). NetGirls: The Internet, Facebook, and body image
Kharpal, A. (2015). Facebook’s Instagram hits 400M users, beats twitter. CNBC.
concern in adolescent girls. International Journal of Eating Disorders, 46, 630–633.
Retrieved from http://www.cnbc.com/2015/09/23/instagram-hits-400-
http://dx.doi.org/10.1002/eat.22141
million-users-beating-twitter.html
Tiggemann, M., & Slater, A. (2014). NetTweens: The Internet and body image con-
Krones, P. G., Stice, E., Batres, C., & Orjada, K. (2005). In vivo social comparison to a
cerns in preteenage girls. Journal of Early Adolescence, 34, 606–620. http://dx.doi.
thin-ideal peer promotes body dissatisfaction: A randomized experiment. Inter-
org/10.1177/0272431613501083
national Journal of Eating Disorders, 38, 134–142. http://dx.doi.org/10.1002/eat.
Tiggemann, M., & Zaccardo, M. (2015). “Exercise to be fit, not skinny”: The effect of
20171
fitspiration imagery on women’s body image. Body Image, 15, 61–67. http://dx.
Leahey, T. M., & Crowther, J. H. (2008). An ecological momentary assessment of
doi.org/10.1016/j.bodyim.2015.06.003
comparison target as a moderator of the effects of appearance-focused social
Trottier, K., Polivy, J., & Herman, C. P. (2007). Effects of exposure to thin and over-
comparisons. Body Image, 5, 307–311. http://dx.doi.org/10.1016/j.bodyim.2008.
weight peers: Evidence of social comparison in restrained and unrestrained
03.002
eaters. Journal of Social and Clinical Psychology, 26, 155–172. http://dx.doi.org/
Levine, M. P., & Murnen, S. K. (2009). “Everybody knows that mass media are/are
10.1521/jscp.2007.26.2.155
not [pick one] a cause of eating disorders”: A critical review of evidence for
Want, S. C. (2009). Meta-analytic moderators of experimental exposure to media
a causal link between media, negative body image, and disordered eating in
portrayals of women on female appearance satisfaction: Social comparisons
females. Journal of Social and Clinical Psychology, 28, 9–42. http://dx.doi.org/10.
as automatic processes. Body Image, 6, 257–269. http://dx.doi.org/10.1016/j.
1521/jscp.2009.28.1.9
bodyim.2009.07.008
Mabe, A. G., Forney, K. J., & Keel, P. K. (2014). Do you “like” my photo? Facebook
Wasilenko, K. A., Kulik, J. A., & Wanic, R. A. (2007). Effects of social comparisons with
use maintains eating disorder risk. International Journal of Eating Disorders, 47,
peers on women’s body satisfaction and exercise behavior. International Journal
516–523. http://dx.doi.org/10.1002/eat.22254
of Eating Disorders, 40, 740–745. http://dx.doi.org/10.1002/eat.20433
Maltby, J., Giles, D. C., Barber, L., & McCutcheon, L. E. (2005). Intense-personal
Wheeler, L., & Miyake, K. (1992). Social comparison in everyday life. Journal of Person-
celebrity worship and body image: Evidence of a link among female adolescents.
ality and Social Psychology, 62, 760–773. http://dx.doi.org/10.1037/0022-3514.
British Journal of Health Psychology, 10, 17–32. http://
62.5.760
Manago, A. M., Graham, M. B., Greenfield, P. M., & Salimkhan, G. (2008).
Wilksch, S. M., Paxton, S. J., Byrne, S. M., Austin, S. B., McLean, S. A., Thompson,
Self-presentation and gender on MySpace. Journal of Applied Developmental Psy-
K. M.,. . .& Wade, T. D. (2014). Prevention across the spectrum: A randomized
chology, 29, 446–458. http://dx.doi.org/10.1016/j.appdev.2008.07.001
controlled trial of three programs to reduce risk factors for both eating disorders
McCutcheon, L. E., Lange, R., & Houran, J. (2002). Conceptualization and measure-
and obesity. Psychological Medicine, 45, 1811–1823. http://dx.doi.org/10.1017/
ment of celebrity worship. British Journal of Psychology, 93, 67–87. http://dx.doi.
s003329171400289x
org/10.1348/000712602162454