J a






p p a a t c t s ( R a t a 2 a 2 s t (

h 1

Body Image 27 (2018) 86–92

Contents lists available at ScienceDirect

Body Image

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / b o d y i m a g e

Selfie” harm: Effects on mood and body image in young women

ennifer S. Mills a,∗, Sarah Musto a, Lindsay Williams a, Marika Tiggemann b

Department of Psychology, York University, 4700 Keele St., Toronto, Ontario, M3J 1P3, Canada School of Psychology, Flinders University, Adelaide, South Australia

r t i c l e i n f o

rticle history: eceived 2 February 2018 eceived in revised form 9 August 2018 ccepted 10 August 2018 vailable online 24 August 2018


a b s t r a c t

“Selfies” (self-taken photos) are a common self-presentation strategy on social media. This study exper- imentally tested whether taking and posting selfies, with and without photo-retouching, elicits changes to mood and body image among young women. Female undergraduate students (N = 110) were randomly assigned to one of three experimental conditions: taking and uploading either an untouched selfie, taking and posting a preferred and retouched selfie to social media, or a control group. State mood and body image were measured pre- and post-manipulation. As predicted, there was a main effect of experimental condition on changes to mood and feelings of physical attractiveness. Women who took and posted self-

ocial media elfies ody image nxiety onfidence elf-presentation

ies to social media reported feeling more anxious, less confident, and less physically attractive afterwards compared to those in the control group. Harmful effects of selfies were found even when participants could retake and retouch their selfies. This is the first experimental study showing that taking and posting selfies on social media causes adverse psychological effects for women.

© 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license


. Introduction

Within the past decade, social networking has become a hugely opular form of online communication, especially among young eople (Perloff, 2014). Facebook, Instagram, and Snapchat are mong some of the most widely used social media platforms avail- ble and can be accessed via computer, smartphone, computer ablet, and through other forms of technology (Perloff, 2014). In omparison to conventional mass media, social media are interac- ive, allowing individuals to create their own personal profiles and hare information and photos with users on their social network Stefanone, Lackaff, & Rosen, 2011). A national survey by the Pew esearch Center found that in the U.S., 18- to 29-year-olds who ccess the Internet are the most likely of any demographic group o use a social networking (i.e., social media) site, and that women re more likely than men to use these sites (Duggan & Brenner, 013). Over 95% of college students regularly maintain and man- ge their social networking profiles (Perloff, 2014; Stefanone et al., 011). Women, in particular, have been found to upload photos to ocial media more frequently than do men, and tend to spend more

ime updating, managing, and maintaining their personal profiles Stefanone et al., 2011).

∗ Corresponding author. E-mail address: [email protected] (J.S. Mills).

ttps:// 740-1445/© 2018 The Authors. Published by Elsevier Ltd. This is an open access article u


Emerging evidence provides insight into the effects that social media behaviours may have on users. On one hand, social media use may be beneficial as it allows greater connectedness with oth- ers, leading to an increased sense of well-being (Tiggemann & Miller, 2010). On the other hand, social media use may lead to a preoccupation and focus on physical appearance, such as engage- ment in appearance-related photo activities (Cohen, Newton-John, & Slater, 2017), which could cause appearance concerns and low- ered body image and self-esteem (de Vries, Peter, Nikken, & de Graaf, 2014). As users are frequently exposed to a variety of other profiles, they can compare their own appearance to friends, rel- atives, and strangers (Haferkamp & Kramer, 2011). Hancock and Toma (2009) found that people select their own online dating pro- file photos in an attempt to look as attractive as possible without being judged to be deceptive. Cross-sectional data have revealed that for both women and men, Facebook use is associated with greater (upward) social comparison and self-objectification, which are both related to lower self-esteem, poorer mental health, and body image concerns (Hanna et al., 2017).

1.1. Social media and body image

Various studies have documented widespread body and weight

dissatisfaction among girls and women, and social media has been found to be a significant catalyst for these appearance con- cerns (Brown & Tiggemann, 2016; Holland & Tiggemann, 2016; Tiggemann & Miller, 2010). Given that social media provide the

nder the CC BY license (

y Imag

o t f S a w a ( F m i m a w G f r a T M s i b c d a s t l M s s m


m i t t ( d a t t ( u s & a H p “ s e p F m e a 2 h s 1 s

J.S. Mills et al. / Bod

pportunity for social comparison, as well as exposure to unrealis- ic beauty expectations, body dissatisfaction is likely to result from requent use (Fardouly, Pinkus, & Vartanian, 2017; Tiggemann & later, 2013; Want & Saiphoo, 2017). Social media present innumer- ble idealized images of thin, lean/tone, beautiful, photo-shopped omen, and the “thin ideal” and “athletic ideal” are displayed as

normal, desirable, and attainable body type for every woman Kim & Chock, 2015; Meier & Gray, 2014; Robinson et al., 2017). urthermore, the Internet and social media have been found to pro- ote thinness, dieting behavior, and weight loss through idealized

mages of “perfect” women (Perloff, 2014). Women who use social edia often internalize the “thin ideal,” causing them to strive for

n unrealistic, unnatural standard of beauty and to feel ashamed hen they are unable to achieve it (Kim & Chock, 2015; Meier & ray, 2014; Tiggemann & Slater, 2013). Studies have found that

requent exposure to the Internet and social networking websites esults in high levels of weight dissatisfaction, drive for thinness, nd body surveillance in young women (Tiggemann & Miller, 2010; iggemann & Slater, 2013), regardless of race (Howard, Heron, acIntyre, Myers, & Everhart, 2017). Additionally, Perloff (2014)

uggests that women who have relatively higher levels of thin deal internalization, perfectionism, and/or low self-esteem would e especially likely to spend time on appearance-focused online omparisons and that they probably do not use ‘self-protective’ ownward appearance comparisons (i.e., comparing their appear- nce to less attractive friends). These predictions are concerning, ince high body dissatisfaction among women is a primary risk fac- or for the development of eating disorders and is correlated with ow self-esteem and depression (Meier & Gray, 2014; Tiggemann &

iller, 2010). Therefore, it is important for researchers to under- tand the causal effects that social media and self-presentation trategies have on young women by using experimental research ethods.

.2. Self-presentation and impression management

Past research on the psychological effects of social media has ainly focused on the implications of social media use for body sat-

sfaction in general. However, there is a lack of empirical research hat evaluates the effects of the specific self-presentation strategies hat social media users rely on. According to Toma and Hancock 2010), self-presentation involves “adjusting and editing the self uring social interactions to create a desired impression on the udience.” The motivation to selectively self-present also relates o impression management, whereby individuals carefully present hemselves in order to make specific impressions on their viewers Pounders, Kowalczyk, & Stowers, 2016). As a result, social media sers are driven to present the most attractive versions of them- elves to others in order to make a favorable impression (Toma

Hancock, 2010). These photos, however, often do not portray n accurate depiction of one’s true physical appearance (Toma & ancock, 2010). The most common way that users selectively self- resent on social media is through the taking and uploading of selfies” (photos taken by and of oneself). Users tend to capture elfies from flattering angles and using bright lighting, and may also dit their photos using colour correction, skin-retouching, and even hoto-shopping to make body parts appear thinner (Anderson, agan, Woodnutt, & Chamorro-Premuzic, 2012). In this way, social edia users are able to manage the impressions they have on oth-

rs by presenting only the most flattering images of themselves nd minimizing perceived flaws or imperfections (Anderson et al., 012; Bell, Cassarly, & Dunbar, 2018; Pounders et al., 2016). It

as also been found that individuals who desire to boost their elf-esteem upload selfies more frequently, and that women of 6–25 years of age spend up to 5 h per week taking selfies and haring them on social media (Pounders et al., 2016). Research on

e 27 (2018) 86–92 87

gender differences in Internet activities has found that, compared to men, women tend to be more motivated to create a positive self-presentation on their social media profiles, and as a result, they engage in more photo-enhancement behaviours (Haferkamp, Eimler, Papadakis, & Kruck, 2012; Toma & Hancock, 2010). Over- all, research has suggested that the taking and retouching of selfies may be a particularly risky behaviour in terms of its potential to negatively impact the body image and self-esteem of young girls and women.

1.3. The current study

In summary, previous research demonstrates that social media use is positively correlated with appearance concern. Furthermore, the literature suggests that selfie-taking and photo-retouching, which are very common social media behaviours, are associated with poorer self-esteem and body image among young women. It has been suggested that editing and uploading selfies may worsen appearance concerns (de Vries et al., 2014), but it is not yet known whether a causal relationship exists.

To fill this gap in the literature, the current study tested the effects of selfie taking on body image and mood in women. It was hypothesized that updating one’s social media profile with a selfie photo would result in lowered mood and increased body concerns as compared to a control group. To answer a secondary research question, we also tested the effects of having control over self- presentation on social media, by retaking and retouching a selfie photo, on women’s body image and mood. It was hypothesized that participants who were allowed to retake and retouch their selfie would experience better mood and body image compared to women who were not allowed to modify their selfie before posting it on social media. This is because women typically react to seeing a photo of themselves by feeling dissatisfied with their appearance (Mills, Shikatani, Tiggemann, & Hollitt, 2014) and photo modifica- tion allows a person to present an idealized version of themselves to others (Tiggemann & Miller, 2010).

2. Method

2.1. Participants

Participants were 113 psychology undergraduate students recruited through an online experiment management system at York University in Toronto, Canada. Inclusion criteria included being female, being between 16 and 29 years old (M = 19.00, SD = 1.66), and having an active account on Facebook or Insta- gram. In exchange for their participation in a single, hour-long lab session, participants received partial course credit toward their Introduction to Psychology course. The self-reported eth- nic distribution of the sample was 24.8% South Asian, 20.2% European/Caucasian, 12.8% Black/African-American, 10.1% Middle Eastern, 9.2 Caribbean, 6.4% Pacific Islands American, 5.5% East Asian, 2.8% Latino/ Hispanic, and 8.2% other ethnic identification. Body mass index (BMI = kg/m2) scores ranged from 15.84 to 36.23 (M = 23.71, SD = 4.03) across the sample, with the mode, median, and mean all falling within the “normal” weight range (18.5 < BMI <24.9) (Centers for Disease Control & Prevention, 2015). One partic-

ipant who mistakenly signed up for the study was excluded because he self-identified as male. Two participants declined to partici- pate after reading the informed consent form because they were uncomfortable taking a photo of themselves for religious reasons.

8 y Imag






m t s c T 1 e o w V fi a s i p p c i T p


p o v


P v b t a d w p a l fi s p p S t ( d t a b C r n f T

8 J.S. Mills et al. / Bod

.2. Apparatus

.2.1. iPad Participants used the Internet browser, camera, and photo mod-

fication app (“You-Cam Now”), if applicable, installed on an iPad.

.3. Measures

.3.1. Mood and body image A series of visual analogue scales (VAS) was used to measure

ood and body image at baseline as well as after the experimen- al manipulation (described below). This commonly used set of cales was designed to assess pre-post fluctuations in psychologi- al states, typically in experimental research designs (Heinberg & hompson, 1995). The measure consisted of six VAS, each with a 0-centimeter horizontal line labeled with a specific attitude or motional state. Participants are asked to place an X on the point n the line that most accurately depicts the degree to which they ere experiencing that feeling at the moment, from Not at all to ery much. The mood items included anxiety, depression, and con- dence. The body image items included feelings of fatness, physical ttractiveness, and body size satisfaction. Rather than collapsing cores into global affect or appearance concerns, we separated the tems so that we could examine specific affective changes among articipants. VAS format is recommended over Likert scales for pre- ost research designs since it reduces recall bias (i.e., participants annot recall their previous response), can be completed quickly, s sensitive to emotional changes (Hargreaves & Tiggemann, 2003). he measure used in the current study is the same one used in other ublished studies.

.3.2. Demographics Age and race/ethnicity demographics were collected from each

articipant. Filler items not of interest to the study were included n the questionnaire (e.g., living arrangements, year of study, uni- ersity program, and media consumption).

.4. Procedure

Ethics approval was received from York University’s Human articipants Review Committee. Female undergraduate students olunteered for an advertised study examining “the relationship etween personality and social media use.” Participants were ested individually behind a partition wall from the experimenter nd were asked to leave their bags and any personal electronic evices (including phones) outside of the testing area. Participants ere randomly assigned to one of three experimental conditions rior to arriving at the lab. Upon arrival to the lab, participants read nd signed a written informed consent form, were given a base- ine VAS, and then the demographics questionnaire with additional ller items to distract from the purpose of the study. For ethical rea- ons, the informed consent form contained the information that articipants may be asked to post a selfie to their own social media rofile. For the experimental task, participants in the Untouched elfie condition were asked to take a single photo (a headshot) on he lab’s iPad and upload it to their preferred social media profile Facebook or Instagram). Participants in the Retouched Selfie con- ition were asked to take one or more photos of themselves on he lab’s iPad and were told that they could use the photo editing pp installed on the iPad to retouch the photo to their satisfaction efore uploading it to their social media profile. Participants in the ontrol condition were also given the lab’s iPad but were asked to

ead a short article from a social media news website chosen for eutral, non-appearance related content (i.e., popular travel ideas

or university students) and to answer questions about the article. his task was chosen to maintain the cover study of social media

e 27 (2018) 86–92

use and to control for using an iPad, and for the amount of time elapsed between pre-post measures. It was intentional that Con- trol condition participants not engage on Facebook or Instagram (theirs or other people’s profiles, since we could not be certain that they were not exposed to appearance-related content, which could affect mood and/or body image).

The assigned tasks in the Untouched Selfie and Control condi- tions were timed (5 min each). The Retouched Selfie condition was not timed so that participants could retake and retouch their selfie to their satisfaction. However, time to completion was recorded by the experimenter and participants in the Retouched Selfie condi- tion took a similar amount of time to complete their task (mean time to completion = 4.5 min). Instructions and set up in all three conditions took approximately 1–2 min.

As manipulation checks, Control condition participants were asked to answer written questions about their article to ensure that they read the article. Selfie condition participants were asked verbally by the experimenter whether they completed the tasks as instructed. In addition, at the end of the study the experi- menter checked the photo and browser histories, and any deleted files on the iPad to ensure that participants in all conditions adhered to the instructions and did not open any other web- sites or social media profiles. All participants confirmed that they followed the instructions and there was no evidence of non-adherence.

Upon completion of the experimental tasks, all participants completed the post-manipulation VAS. Participants were asked to complete the scales based on how they were feeling at that particular moment. The elapsed time between the baseline and post-manipulation VAS measure was approximately 10 min. Fur- thermore, the format of the VAS scale is such that participants cannot recall their previous answer; thus, recall bias is minimized. Participants were then debriefed and probed as to what they believed to be the purpose of the study. Lastly, height and weight were measured by the experimenter on a balance beam scale.

2.5. Data analysis

Statistical analyses were conducted using SPSS version 24. An alpha level of .05 was used for significance testing. A power analysis was conducted using G*Power (Faul, Erdfelder, Lang, & Buchner, 2007); an alpha of .05, medium effect size, and power estimate of .80 resulted in a recommended sample size of 110, which was obtained. Repeated measures analysis (Time 1 – Time 2) was chosen to analyze the effects of experimental condi- tion instead of VAS change scores to maximize power and use within-subject error estimates. To control for Type I error, an initial repeated measures multivariate analysis of variance (RM- MANOVA) was performed with time (Time 1 – Time 2) and test (VAS item) as the within-subject factors, and experimental condition (Untouched Selfie, Retouched Selfie, and Control) as the between-subjects factor. Any significant multivariate 3-way interaction (time × test × condition) on the combined dependent measures was followed by univariate repeated measures ANOVAs, with time (Time 1 – Time 2) as the within-subject factor and experimental condition as the between-subjects factor. Any signif- icant within-subjects contrasts (time × condition) were followed by post-hoc t-tests to examine which conditions differed. For

ease of interpretation, change scores (Time 1 – Time 2) were used only for these post hoc t-tests to examine the direction and magnitude of change to psychological states as a function of condition.

J.S. Mills et al. / Body Image 27 (2018) 86–92 89

F s



a t n o r


i m T t

o T t b

a . t 1 g T H H b


l a

c n T

Fig. 2. Mean change in confidence as a function of condition. Means with different superscripts are significantly different from one another at p < .05.

ig. 1. Mean change in anxiety as a function of condition. Means with different uperscripts are significantly different from one another at p < .05.

. Results

.1. Preliminary analyses

Inspection of histograms, skewness, and kurtosis suggested that ll of the variables were normally distributed. There were no statis- ical outliers (± 3.0 SD) among the dependent variables; therefore, o adjustments were made. Groups did not differ significantly n baseline levels of any variable, suggesting that randomization esulted in equivalent groups.

.2. Multivariate effects of experimental condition

Means and standard deviations for all dependent variables of nterest (pre- and post-manipulation) as a function of the experi-

ental condition are shown in Table 1. For ease of interpretation, able 1 also shows the change in participants’ self-ratings across he psychological states.

A significant 3-way (test × time × condition) multivariate effect n the combined dependent variables was found, Hotelling’s race = .21, F(10, 201) = 2.14, p = .02, partial �2 = .10, meaning that he experimental groups differed with respect to how mood and ody image ratings changed between Time 1 and Time 2.

Significant 2-way (time × condition) interactions were found for nxiety, Hotelling’s Trace = .06, F(2, 107) = 3.32, p = .04, partial �2 =

06, confidence, Hotelling’s Trace = .07, F(2, 107) = 3.69, p = .03, par- ial �2 = .07, and physical attractiveness, Hotelling’s Trace = .07, F(2, 07) = 3.59, p = .03, partial �2 = .06, meaning that the experimental roups were not equal with respect to changes on those items from ime 1 to Time 2. Interactions were not significant for depression, otelling’s Trace = .01, F(2, 107) = 0.48, p = .62, feelings of fatness, otelling’s Trace = .02, F(2, 107) = 0.97, p = .38, or satisfaction with ody size, Hotelling’s Trace = .01, F(2, 107) = 0.75, p = .47.

.3. Changes to psychological states as a function of condition

The significant 2-way interactions reported above were fol- owed up with t-tests to compare changes to psychological states cross experimental groups.

As can be seen in Fig. 1, participants in the Untouched Selfie ondition experienced an increase in anxiety and this was sig- ificantly greater than the Control condition t(71) = 2.35, p = .02. he Retouched Selfie condition also experienced an increase in

anxiety but was not significantly different from the Control con- dition, t(71) = 1.80, p = .08. The Untouched and the Retouched Selfie conditions did not differ with respect to changes in anxiety, t(71) = 0.79, p = .43.

Fig. 2 shows that participants in the Untouched Selfie condition experienced a decrease in confidence and this was significantly greater than the Control condition, t(71) = 2.48, p = .01, and marginally greater than that experienced in the Retouched Selfie conditions t(72) = 1.92, p = .06. There was no difference in changes to feelings of confidence between the Retouched Selfie and Control conditions, t(71) = 0.60, p = .55.

Fig. 3 shows that participants experienced decreases in feel- ings of physical attractiveness that were significantly greater than in the Control condition in both the Untouched Selfie condition, t(71) = 2.43, p = .02, and the Retouched Selfie condition, t(71) = 2.32, p = .02. These decreases were equivalent between the two selfie conditions t(72) = 0.12, p = .90.

Fig. 3. Mean change in feelings of physical attractiveness as a function of condition. Means with different superscripts are significantly different from one another at p < .05.

90 J.S. Mills et al. / Body Image 27 (2018) 86–92

Table 1 Dependent Variables of Interest (Pre and Post) as a Function of Experimental Condition.

Experimental Condition

Untouched Selfie (n = 37) Retouched Selfie (n = 37) Control (n = 36) M (SD) M (SD) M (SD)

Anxiety - Pre Anxiety - Post Change

27.15 (26.91) 37.14 (30.17) 9.99 (26.87)

26.92 (24.82) 32.97 (26.74) 5.50 (21.26)

32.11 (27.48) 27.22 (28.35) −4.89 (27.29)

Depression - Pre Depression - Post Change

13.61 (20.79) 16.30 (21.63) 2.69 (15.76)

19.81 (20.97) 19.24 (24.02) −0.57 (14.54)

18.56 (19.96) 20.00 (22.96) 2.00 (14.90)

Confidence - Pre Confidence - Post Change

54.93 (24.37) 39.05 (27.83) −15.88 (21.38)

55.86 (23.77) 48.51 (22.44) −7.35 (16.61)

60.17 (23.14) 55.89 (23.84) −5.11 (15.06)

Feelings of Fatness - Pre Feelings of Fatness - Post Change

34.78 (28.11) 35.58 (29.23) −0.17 (19.83)

36.81 (27.37) 34.95 (28.17) −1.86 (17.15)

37.63 (30.44) 31.92 (28.32) −6.26 (20.93)

Physical Attractiveness - Pre Physical Attractiveness - Post Change

48.35 (22.58) 41.89 (25.10) −6.46 (17.07)

49.57 (19.81) 43.59 (21.45) −5.97 (16.66)

51.63 (22.11) 53.53 (22.67) 1.90 (11.85)

Satisfaction Body Size - Pre 47.89 (32.11) 46.89 (25.75) 50.49 (29.34)


i s m a i d o

w a a c a t w fi e t s t t i s T w a

a f t h t c I p b o i

Satisfaction Body Size - Post Change

46.51 (28.99) −1.38 (13.21)

. Discussion

This is the first experimental study of the causal effects of post- ng selfies to social media on young women. The findings generally upported our hypothesis that taking and posting a selfie on social edia would result in lowered mood and worsened self-image. We

lso found that women who had the opportunity to retake and mod- fy their selfie before posting it to social media still experienced ecreases to mood and anxiety that were similar to the reactions f those who could not retouch their photo.

Participants who took and uploaded a selfie onto social media, ithout the option to retouch or take multiple photos, felt more

nxious, less confident, and less physically attractive afterward, nd these differences were significantly greater than the control ondition (i.e., reading a neutral news article online). These results ll yielded medium effect sizes. These findings are consistent with he previous suggestion that appearance concerns are heightened hen women interact with and construct their social media pro- les, manifesting in poorer body image and mood (e.g., de Vries t al., 2014). However, we did not find significant effects of selfie- aking on all of the dependent variables of interest in the current tudy; we found null effects on state feelings of fatness, satisfac- ion with one’s body, and depression. We interpret these findings o suggest that the psychological states affected by taking and post- ng selfies to social media are specifically related to feelings of elf-consciousness and/or fear of negative evaluation by others. his interpretation seems likely given that participants in the study ere sharing their selfie photos on their own social media profiles

nd for other people they know to see. It is interesting that feelings of physical attractiveness were neg-

tively affected by selfie taking and posting, but not feelings of atness or satisfaction with one’s body size. However, it is important o note that the current study involved taking a photo only of one’s ead and face. In other words, it may not be surprising that effects of aking a selfie on body-related constructs were not found, since the urrent study looked only at the effects of taking selfies of one’s face. f the current study had examined the effect of taking and posting

hotos that showed the participant’s body the results might have een different. Celebrities, but probably many social media users, ften post body-conscious selfies on their social media (e.g., wear- ng bathing suits, lingerie, or no clothing at all). Posting selfies of

46.14 (24.06) −0.76 (21.15)

53.56 (28.58) 3.07 (14.81)

one’s body (and not just the face), even when clothed, could …