Every (Insta)Gram Counts? Applying Cultivation Theory to Explore Instagram Effects on Young Users' Body Image PDF

Psychology of Popular Media...

Psychology of Popular Media © 2019 American Psychological Association 2021, Vol. 10, No. 1, 87–97 ISSN: 2689-6567 http://dx.doi.org/10.1037/ppm0000268 Every (Insta)Gram Counts? Applying Cultivation Theory to Explore the Effects of Instagram on Young Users’ Body Image Jan-Philipp Stein Elena Krause and Peter Ohler University of Würzburg Chemnitz University of Technology Recent research suggests that social networks have replaced traditional media as the main channel by which beauty ideals are conveyed— often resulting in body dissatisfaction and reduced self-esteem This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. among users. Although social comparison theory provides an empirically sound approach to these effects, This document is copyrighted by the American Psychological Association or one of its allied publishers. we argue that additional insight may be offered by cultivation theory and its structured exploration of cognitive, attitudinal, and behavioral outcome variables. Thus, the present study scrutinizes the social network Instagram as a potential cultivation system for young adults’ body image. Recruiting 228 participants aged 18 to 34 years, we systematically explore 3 orders of cultivation, that is, changes in weight-related knowledge, attitudes, and self-reported dietary restraint. As we differentiate between Instagram use quantity and quality, we observe that mere usage time cannot predict the assumed outcomes; instead, only participants’ tendency to browse Instagram’s public content emerges as a relevant predictor, connecting to biased views on the physical appearance of strangers as well as more disordered eating behavior. Considering the fact that Instagram use relates more to other-focused than to self-focused perceptions in our study, we argue that cultivation theory can indeed complement social comparison theory in the current understanding of media-transmitted body images. Public Policy Relevance Statement This article lends both a theoretical foundation as well as empirical support to the argument that highly visual social media constitute a meaningful cultivation system for body-related attitudes and behaviors among young adults. Our research illustrates how the frequent exposure to the virtual self-presentation of others may affect the way people look at strangers’ bodies or indulge in disordered eating— even if their own body esteem remains intact. Keywords: social media, cultivation theory, body image, body dissatisfaction, disordered eating People’s concept of beautiful appearances does not result from on people’s perception and evaluation of physical appearances isolated contemplation but gets shaped— quite literally— by social (Choukas-Bradley, Nesi, Widman, & Higgins, 2019; Holland & interactions and the consumption of media. In regard to the latter, Tiggemann, 2016; Tiggemann, Hayden, Brown, & Veldhuis, a large body of 20th-century media research has explored the role 2018). of TV shows, movies, and printed advertisements as breeding Initially, most scientific interest in the relationship between grounds for a problematic body image among mass audiences social media and people’s body image revolved around Facebook, (Grabe, Ward, & Hyde, 2008). Since the turn of the millennium, the first SNS to achieve global mass-adoption. Following the however, a notable shift of focus has occurred; acknowledging the platform’s swift rise to success, media psychological studies soon evolved media use of the “digital generation,” more and more noted that Facebook’s numerous means of self-presentation ex- studies now address the impact of social networking sites (SNSs) erted a rather problematic influence on the bodily perceptions of its users (Fardouly & Vartanian, 2015; Stronge et al., 2015; Tigge- mann & Slater, 2013)— especially among those who frequently used the website’s photo-related features (Kim & Chock, 2015; This article was published Online First December 12, 2019. Meier & Gray, 2014). In recent years, however, several new SNSs X Jan-Philipp Stein, Psychology of Communication and New Media, have appeared on the global stage, and compared with Facebook, University of Würzburg; Elena Krause and Peter Ohler, Department of many of them put an even stronger emphasis on the sharing of Media Psychology, Chemnitz University of Technology. visual content. In consequence, contemporary body image re- This study was funded by the German Research Foundation (DFG) under Grant 1780 (“CrossWorlds”). searchers have redirected their attention yet again, setting out to Correspondence concerning this article should be addressed to Jan- explore how platforms such as Pinterest (Lewallen & Behm- Philipp Stein, Psychology of Communication and New Media, University Morawitz, 2016) or Snapchat (Marengo, Longobardi, Fabris, & of Würzburg, Oswald-Külpe-Weg 82, 97074 Würzburg, Germany. E-mail: Settani, 2018) influence young people’s concept of the “ideal” [email protected] body. Most of all, the research field has taken a strong interest in 87  88 STEIN, KRAUSE, AND OHLER Instagram, a subsidiary of Facebook, which currently ranks as the to create an environment of constant upward comparison, prompt- most popular social network among young adults (Anderson & ing dysfunctional self-perceptions among other users (Liu et al., Jiang, 2018). Apart from a basic commenting function, Instagram 2017; Vogel, Rose, Roberts, & Eckles, 2014; Yang, 2016). As disregards any text-based features to strictly emphasize the sharing such, Cohen and Blaszczynski (2015) argued that the social nature of pictures and videos. In its resulting role as the Internet’s leading of platforms such as Facebook and Instagram turns them into a “highly-visual social media” (Marengo et al., 2018), the platform virulent cause for body dissatisfaction, which might even surpass has also emerged as a cornerstone to young people’s conceptual- conventional media in terms of influential power. After all, TV ization of physical beauty, not least due to its strong focus on audiences may consider the displayed perfection as a unique fitness, makeup, fashion, and dietary topics (Carrotte, Vella, & Lim, feature of celebrity life; social networking, on the other hand, 2015; Hu, Manikonda, & Kambhampati, 2014; Pinkerton, Lobin, suggests that beautiful appearances are omnipresent, reaching from Querfurth, Pena, & Wilson, 2017). Along these lines, recent content one’s closest friends to countless “normal” people all over the analyses indicate that Instagram has turned into the main hub for world. Moreover, because social media typically suggest that so-called fitspiration and thinspiration content—portmanteaus from beauty equals likability— kept salient by ubiquitous like and heart This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. the words “fit/thin” and “inspiration”—which glorifies an athletic buttons— users are conditioned to adopt the conveyed beauty This document is copyrighted by the American Psychological Association or one of its allied publishers. and healthy, but sometimes also alarmingly skinny, appearance ideals if they look for approval from the virtual community. As a (Ghaznavi & Taylor, 2015; Ging & Garvey, 2017; Tiggemann & result, a vicious circle of upward social comparisons may ensue. Zaccardo, 2016). Although the creators of the respective uploads usually stress their benevolent intentions, recent studies have dem- Finding Additional Value in Cultivation Theory onstrated that fit- and (especially) thinspiration posts tend to have quite negative effects on young audiences, nudging them toward Although SCT certainly offers a relevant approach to body- unattainable beauty standards and disordered eating behaviors related social media effects, we believe that the theory is somewhat (Lewallen & Behm-Morawitz, 2016; Tiggemann & Zaccardo, limited by its chief focus on self-related attitudes. In other words, 2016). From a media psychological perspective, this suggests that by only addressing how using SNSs shape people’s opinion about certain aspects of social media might invoke cultivation processes their own bodies, the theory fails to acknowledge many other similar to the ones that have been observed in the context of potential consequences of this media practice, such as the forma- traditional mass media. However, we note that previous studies tion of wrong assumptions about society or the adoption of new have not yet pursued this theoretical angle, as scholars usually beliefs about others. Furthermore, we argue that SCT-based re- prefer to conceptualize SNS effects as short-term, social psycho- search is often inhibited by its narrow view on affective outcome logical phenomena. Without wanting to refute the value of this variables (e.g., anxiety and reduced self-esteem), whereas it rarely kind of work, the current article will present readers with argu- mentions, let alone explains, cognitive or behavioral effects. ments— both theoretical and empirical in nature—as to why the To fill in the identified research gaps, the current article pro- research field might benefit from interpreting social media use as poses adding another theoretical framework to the scientific in- a multifaceted cultivation process. vestigation of social media effects— one that has received a lot of attention in the context of traditional media and yet become more The Current Understanding of Social Media Effects on of a side-note in the exploration of SNSs: cultivation theory (CT). Proposed by Gerbner and Gross (1976), CT assumes that the Users’ Body Image repeated and extensive use of mass media will entice audiences to When discussing the potential mechanisms behind the negative absorb the broadcast information into their worldview, succes- impact of social media on their users’ body image, scientific sively homogenizing viewers into a singular mainstream. Once this literature has mainly resorted to social comparison theory (SCT; process has advanced for a notable amount of time, a number of Festinger, 1954) as the decisive explanatory framework. This problematic consequences may ensue, including the loss of polit- theory suggests that people possess an innate tendency to evaluate ical diversity or a desensitization toward violence (Fox & Potocki, their identity, appearance, and beliefs by comparing themselves to 2016). Moreover, according to the work of body image scholars, others around them, leading to various, often-problematic out- CT not only offers an explanation as to why users of traditional comes. Specifically, research suggests that social comparisons media experience body dissatisfaction (Botta, 1999; Grabe et al., seem to be a bidirectional process; whereas downward compari- 2008; Holmstrom, 2010) but also accounts for subsequent behav- sons—focusing on people of lower status—serve as a defensive ioral effects such as the development of eating disorders (Ham- mechanism to reinforce self-esteem (Wills, 1981), upward com- mermeister, Brock, Winterstein, & Page, 2005; Shanahan & Mor- parisons that target more successful or supposedly better-natured gan, 1999). As certain beauty ideals and fitness behaviors are individuals usually result in negative effects, such as a loss of promoted repeatedly by the media of choice, recipients are led to self-worth or reduced psychological well-being (Suls & Wheeler, believe that these depictions constitute the societal norm, thus 2000; Tesser, Millar, & Moore, 1988). assimilating them into their own attitudinal repertoire (thin-ideal With only a few exceptions, social media users tend to manage internalization). Following this process, even small dissimilarities their virtual self-presentation rather elaborately (Krämer & Winter, between one’s own appearance and the cultivated ideal may evoke 2008; Zhao, Grasmuck, & Martin, 2008), for instance, by exclud- strong behavioral efforts to conform to the media-transmitted ing negative experiences from their profiles or actively manipu- beauty norms. lating new uploads via filters and enhancement software (Klee- In the current understanding of CT, each of the described mans, Daalmans, Carbaat, & Anschütz, 2018). In turn, the high changes corresponds to a different level or so-called order of prevalence of curated and edited content on SNS has been shown cultivation. Constituting the ignition spark of the cultivation pro-  EVERY (INSTA)GRAM COUNTS? 89 cess, first-order effects describe changes in people’s estimates of 2018). Thus, it comes as little surprise that extensive use of the real-life frequencies, events, and distributions as they arise from service has been identified as an antecedent of body dissatisfaction mass media consumption (Hawkins & Pingree, 1982). Proceeding and reduced self-esteem (Liu et al., 2017; Vogel et al., 2014). from there, second-order effects encompass transformations of the However, other potential consequences of Instagram’s homoge- audience’s attitudes and value systems, which build on top of these nized, often visually enhanced content remain somewhat neglected changed factual assumptions. Lastly, in a more recent modification in the scientific discourse. To overcome this research gap, the of CT, Nabi and Sullivan (2001) added a third order of cultivation, current study set out to systematically scrutinize the platform capturing changes in media users’ observable behavior that result Instagram as a cultivation system for weight-related cognitions, from the first two orders of the process. Thus, in summary, attitudes, and behavior—ranging from changed assumptions about contemporary CT describes a hierarchical phenomenon following society to the adoption of concrete dietary rules. the prolonged exposure to similar media contents—reaching from During the preparation of our hypotheses, we pondered different cognitive changes to observable conative effects. ways as to how participants’ Instagram use might affect them. Unlike its acclaim in 20th-century media psychological litera- Previous literature suggested to us that the mere duration of social This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. ture, however, CT has been featured rather sparsely in the explo- media use might not suffice to explain hypothesized outcomes in This document is copyrighted by the American Psychological Association or one of its allied publishers. ration of social media effects. For instance, regarding the popular a meaningful way (Frison & Eggermont, 2017; Yang, 2016). This platform Instagram, only a handful of studies—most of them with also seemed to apply to our specific research focus: People might extremely limited sample sizes (Goldstraw & Keegan, 2016; spend a lot of time on SNSs but engage only with their own O’Brien, 2015)— even mentions CT. In our interpretation, two profiles or the content of friends, which would shield them from reasons might account for this recline in academic interest. First, the proposed cultivation effects. At the same time, we did not want experimental studies indicate that social media effects often occur to disregard the time factor altogether, keeping in mind that the quite quickly, sometimes even after a single, brief exposure (Far- whole idea of cultivation revolves around repeated and lengthy douly, Diedrichs, Vartanian, & Halliwell, 2015; Haferkamp & media exposure. Moreover, we believed that a longer net time Krämer, 2011; Kleemans et al., 2018), which arguably undercuts spent on the platform was likely to increase users’ exposure to the need for CT’s overarching approach. Second, and maybe more Instagram’s fitness and dieting content, even if they could not importantly, it might be argued that cultivation processes require explicitly recall it in a qualitative self-report measure. As such, we invariable, standardized stimuli (such as a predetermined TV decided to include both (a) the weekly time people spent on schedule) to come into effect, a requirement that modern SNSs and Instagram as well as (b) their specific interest in browsing the their dynamically adjusted content feeds do not seem to meet. platforms’ public contents as two separate predictors in our re- However, we believe that this logic is flawed. In fact, the con- search propositions. sumption of conventional media does not necessarily constitute an Starting with the first order of cultivation, our first hypothesis equalized stimulus either— quite the opposite, media such as TV assumed changes in participants’ factual assumptions about the or newspapers typically offer a vast range of content as well, with general population, as we asked them to estimate the prevalence of hundreds of channels or magazines available to customers. Yet, both overweight and body satisfaction in their native country. despite this range of potential sources, previous cultivation re- Considering that the highly selective uploads on Instagram convey search has still managed to uncover notable effects, which suggests a distorted impression of real life (Kleemans et al., 2018), we that people ultimately tend to choose similar media offerings, thus supposed that active users of the SNS would tend to underestimate prompting media producers to create more of the same (Bourdieu, the prevalence of overweight yet overestimate the number of 2001; Waisbord, 2004). people satisfied with their own body. In a similar vein, literature suggests that social media experi- ences may be much more homogenous than their often-claimed Hypothesis 1: Young adults’ (a) Instagram-use duration and customizability suggests. On most SNSs, the divide between pro- (b) Instagram browsing activity negatively predict their esti- ducers and consumers has blurred, and most new uploads are mation of the overweight prevalence in the general population. created by layperson users (user-generated content). According to recent analyses, this has only fostered the homogenization of Hypothesis 2: Young adults’ (a) Instagram-use duration and popular platforms, as many people try to align their posts with (b) Instagram browsing activity positively predict their esti- “trending” topics or the aesthetics they attribute to a specific site mation of the body satisfaction prevalence in the general (Ong, 2018; Shamsian, 2018; Yau & Reich, 2019). Moreover, population. studies show that social media users typically show very high agreement on the gratifications they expect from different SNSs Proceeding to second-order cultivation effects—that is, the for- (Lee, Lee, Moon, & Sung, 2015; Phua, Jin, & Kim, 2017), further mation of attitudes and norms via media exposure—we explicitly invoking the homogenization of the respective platforms. Follow- differentiated between views on one’s own body and views on the ing these arguments, we believe that it actually makes a lot of bodies of strangers. In terms of self-assessment, we expected sense to explore highly visual social media as cultivation nexuses active Instagram users to hold a less favorable view on their own that influence many users in a similar way. body, as the repeated exposure to flattering and manipulated im- agery should equip them with unrealistic beauty ideals. Arguably, this specific hypothesis marked the one instance where our CT- Hypotheses guided assumptions blended rather well with SCT, mirroring pre- The social network Instagram provides a popular backdrop for vious findings from social comparison studies (Liu et al., 2017; fitness, fashion, and other body-related media content (Ehrhardt, Vogel et al., 2014).  90 STEIN, KRAUSE, AND OHLER Hypothesis 3: Young adults’ (a) Instagram-use duration and nymity and protection of their data, as well as the goal of the (b) Instagram browsing activity negatively predict their own current research project. Only after they gave their informed con- body esteem. sent, participants were allowed to proceed to the actual question- naire. As a reward for their contribution, every person taking part To explore other-related attitudes, we asked participants to rate in this study received a ticket for a €25 gift card raffle; students various full-body photographs in terms of the weight of the de- from the local university could alternatively choose partial course picted people (ranging from “clearly underweight” to “clearly credits. overweight”). According to CT, we assumed that longer Instagram use as well as a stronger engagement with the platforms’ public content would lead to stricter weight norms—manifested as a Stimulus Materials systematic shift toward higher weight ratings in the provided task. For the exploration of participants’ attitudes toward the bodies Hypothesis 4: Young adults’ (a) Instagram-use duration and of strangers— one aspect of the proposed second-order cultivation (b) Instagram browsing activity positively predict the weight effect—we prepared visual stimuli by assembling our own set of This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. level they ascribe to the bodies of strangers. 36 full-body photographs (taken from public Instagram profiles). This document is copyrighted by the American Psychological Association or one of its allied publishers. Expecting a systematic distortion in Instagram users’ view on Addressing the third order of cultivation, we presumed that physical appearances, we decided to depict a wide spectrum of the spending a lot of time on—and actively browsing through— human physique, collecting photos of many different body types Instagram with its numerous “fitspiring” and nutrition-related (e.g., muscular, slender, chubby, and obese). Doing so, we also posts would affect users’ behavioral intentions regarding meals made sure to portray a perfectly balanced female-to-male ratio and and diets. Therefore, we hypothesized: to include different ethnicities in our collection. To avoid con- founding effects that might have arisen from the recognition of Hypothesis 5: Young adults’ (a) Instagram-use duration and celebrities, we further decided to focus on unknown pictures with (b) Instagram browsing activity positively predict their dietary little social media visibility— even though research suggests that restraint. the uploads of laypersons may actually exert quite similar effects Despite the ongoing alleviation of societal norms, research on users’ body image than the photos of famous celebrities (Brown shows that young women are still subjected to much stricter beauty & Tiggemann, 2016). ideals than men (Carlson Jones, 2004; Hargreaves & Tiggemann, To confirm if our final set of pictures (depicting 18 women and 18 2004)—a fact that has also been used to explain the skewed gender men) indeed resembled the intended broad spectrum of body shapes, distribution of clinical eating disorders (Udo & Grilo, 2018). At we conducted a pretest with N ⫽ 15 independent student raters the same time, scientific evidence cautions against trivializing (including both Instagram users and nonusers), requesting them to sort male beauty standards, as a growing number of studies indicates the selected pictures into five ascending categories (“clearly under- that young boys are also severely affected by medially transmitted weight,” “slightly underweight,” “average,” “slightly overweight,” body ideals (Agliata & Tantleff-Dunn, 2004; Blashill & Wilhelm, and “clearly overweight”). By calculating the mean score from the 2014). In consequence, we decided to include both female and obtained 15 ratings, we found that our 36 pictures were indeed male participants in our study, which allowed us to explore scattered rather equally across the five weight categories, indicating whether the assumed cultivation effects would be moderated by the appropriateness of our materials as a depiction of very different participants’ gender. Based on the reviewed evidence on negative body shapes. outcomes for both men and women, an open-ended research ques- tion was formulated instead of a directional hypothesis: Measures Research Question 1: Are the proposed negative outcomes of Instagram use. Previous research has cautioned against using Instagram use moderated by participants’ gender? only the duration of social media use to explain different outcomes (Yang, 2016). Therefore, we decided to differentiate between Method quantitative and qualitative aspects of participants’ Instagram use, adding both as distinct predictors to our exploration of cultivation Participants effects. To measure the quantity of participants’ Instagram use, we Following a recruitment process via university mailing lists and simply asked them for the average number of hours they spent on social media groups, the study’s initial sample consisted of N ⫽ the platform in a normal week—which had to be entered as a 234 participants (age M ⫽ 22.8 years, SD ⫽ 4.13; 176 women, 57 numerical value into a text field. Concerning their use quality, on men, one unspecified). To be included in the study, participants the other hand, participants were requested to express how much had to indicate German as their main language and fall into the age they typically browsed through the public content of the Instagram range between 18 and 34 years, as we were specifically interested community. For this purpose, we adapted four items from a pre- in body image issues among young adults. However, six obtained viously published list of SNS activities (Hendrickse, Arpan, Clay- data sets exceeded the specified age bracket, so that we only used ton, & Ridgway, 2017; Meier & Gray, 2014), namely, “I look at the data of n ⫽ 228 participants in our final analyses (age M ⫽ the profiles of other users,” “I read the comments on photos and 22.5 years, SD ⫽ 3.04; 171 women, 56 men, one unspecified). videos of other users,” “I comment on photos and videos of other Immediately after accessing the prepared online questionnaire, users,” and “I use the like function on photos and videos of other participants received comprehensive information about the ano- users”. All items were presented as 5-point Likert scales (1 ⫽  EVERY (INSTA)GRAM COUNTS? 91 never, 5 ⫽ very often). To further emphasize our focus on the result of this subtraction in our later analyses. This specific pro- content of strangers, we juxtaposed the described items with two cedure was chosen to account for the possibility that both under- irrelevant questions about the content of friends (e.g., “On Insta- and overweight could have had notable effects on our dependent gram, I look at the photos and videos of friends.”). The resulting variables—a finding that would have been clouded by simply Instagram Browsing Index showed very good internal reliability, using the BMI in our linear regression procedure. Cronbach’s ␣ ⫽.86. Although it could be argued that two of our chosen items referred more to the passive exploration of content while the other two addressed active interactions, an exploratory Results factor analysis clearly demonstrated the measure’s unidimension- Table 1 gives a descriptive overview of all measured variables, ality, with a single factor explaining 72.2% of the total observed whereas Table 2 summarizes the zero-order correlations between variance. them. Corresponding to our five hypotheses, we conducted a total Weight-related estimations. To assess potential changes in of five hierarchical linear regression analyses with IBM SPSS 20 participants’ assumptions about real-life frequencies, we asked (IBM Corp, 2011) software— exploring the prediction of partici- This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. them to estimate the percentage of overweight people—that is, pants’ factual estimations (Hypotheses 1 and 2), self- and other- This document is copyrighted by the American Psychological Association or one of its allied publishers. those with a body mass index (BMI) above 25 kg/m2—in the total related attitudes (Hypotheses 3 and 4), and self-reported disordered population of Germany.1 As a second, potentially skewed factual eating behavior (Hypothesis 5). In each of these analyses, partic- assumption, we explored participants’ estimations of the country- ipants’ age, gender, and deviation from a perfect BMI were entered wide body satisfaction (“What percentage of the German popula- in the first step of the hierarchical procedure. Then, in the second tion do you believe to be happy with their own body?”).2 step, we added the obtained measures of Instagram use: use dura- Body image attitudes. Participants’ view on their own body tion (in hours) and browsing activity (as assessed by our self- was measured using the Body Esteem Scale for Adolescents and developed browsing index). If either of these two variables Adults (Mendelson, Mendelson, & White, 2001), which contains emerged as a significant predictor, we further calculated a third, the three subscales Appearance Satisfaction (10 items, e.g., “I like moderated regression model, which included interaction terms what I look like in pictures”), Weight Satisfaction (eight items, between the significant Instagram use variable(s) and participants’ e.g., “I’m proud of my body”), and External Attribution (five gender (i.e., “Gender ⫻ Instagram-Use Duration” or “Gender ⫻ items, e.g., “Other people consider me good-looking”). However, Instagram Browsing Activity”).3 This third step served to give an because we were not interested in factors such as clothing or external informed answer to our additional research question on the mod- validation, we focused our analysis on the second subscale, which erating role of gender for the proposed cultivation effects. For strictly addresses attitudes regarding the own body shape and weight. readers’ convenience, Tables 3 through 7 present an overview of All items had to be rated on a 5-point Likert scale (1 ⫽ not at all, 5 ⫽ all calculated regression models. completely), with internal consistency of the Weight Satisfaction scale turning out excellent, Cronbach’s ␣ ⫽.93. For the exploration of participants’ other-related body image, First-Order Cultivation: Estimations they were required to rate the collected 36 full-body portraits on Our first hierarchical regression focused on participants’ esti- 5-point scales, ranging from 1 (clearly underweight) to 5 (clearly mations of the national overweight prevalence as a potentially overweight). By averaging all 36 evaluations, we acquired a com- cultivated real-world assumption (Hypothesis 1). Using partici- posite weight rating score spanning the full spectrum of presented pants’ estimations as a criterion and entering the three sociodemo- body types. graphic predictors, the first step of our model accounted for ap- Dietary behaviors. The Eating Attitudes Test (Garner & Gar- proximately 11% of the observable variance, F(3, 223) ⫽ 8.81, finkel, 1979) ranks among the most well-established instruments p ⬍.01. Specifically, we found that all three demographic vari- for a reliable assessment of diet-related motivations and activities. ables significantly predicted participants’ estimations of the over- Although the questionnaire has been originally developed for the weight prevalence: age, ␤ ⫽.25, t(223) ⫽ 3.67, p ⬍.01; gender, diagnosis of clinical eating disorders (and still retains its popularity ␤ ⫽ ⫺.24, t(223) ⫽ 3.48, p ⫽.03; and deviation from a perfect in this field), it has also been used in nonclinical settings (Rogoza, BMI, ␤ ⫽.19, t(223) ⫽ 2.98, p ⬍.01. Considering the orientation Brytek-Matera, & Garner, 2016). In our study, we used the Ger- of the beta values and our dummy coding of the gender variable, man short version of the scale (Berger et al., 2012), which has been this means that both higher age and female gender were related to described as a risk screening of problematic dietary behavior by its higher estimations of the national overweight prevalence. In terms authors. All 13 items (e.g., “I like my stomach to be empty” and “I display self-control around food”) had to be answered in a 5-point format (1 ⫽ never, 5 ⫽ always). We observed excellent internal 1 The actual value provided by the Robert Koch institute, a subordinate consistency for the German short version of Eating Attitudes Test, of the German Federal Ministry of Health, is 54% (Schienkiewitz, Cronbach’s ␣ ⫽.92. Mensink, Kuhnert, & Lange, 2017). Accordingly, any number below this value indicated a participant’s underestimation of the national obesity Control variables. Apart from participants’ age and gender, prevalence, whereas higher percentages indicated an overestimation. we kindly requested them to disclose their body height and weight 2 Using the empirical value provided by a large-scale survey of 1,058 so that we could calculate individual BMIs (defined as body mass Germans (myMarktforschung, 2016), any answer below 51% indicated an divided by the square of the body height). Next, we subtracted underestimation of the body satisfaction in the broader public, whereas higher values expressed an overestimation. participants’ actual BMI from the value 21.7—which marks the 3 As required for this kind of procedure, nonbinary variables such as age “perfect” midpoint of the healthy weight range suggested by the or Instagram use duration were mean-centered before calculating the World Health Organization (2008)—and only used the absolute interaction terms for the moderated regression.  92 STEIN, KRAUSE, AND OHLER Table 1 Table 3 Means and Standard Deviations for the Measured Variables Hierarchical Regression With Participants’ Estimations of the National Overweight Prevalence as a Criterion Measured variable M SD Minimum Maximum Predictor ␤ t F R2 Age 22.46 3.05 18 34 Body mass index 22.56 3.60 15.9 37.9 1 Difference from perfect body mass Age.25ⴱⴱ 3.67 8.81ⴱⴱ.11 index 2.65 3.60 0.01 16.21 Gendera ⫺.24ⴱ 3.48 Weekly Instagram use in hours 5.10 7.16 0 55 Difference from perfect a Instagram browsing index 2.48 1.00 1.00 4.25 body mass index.19ⴱⴱ 2.98 2 First-order cultivation measures ⴱⴱ Age.25 3.57 5.34ⴱⴱ.11 Estimation of overweight Gendera ⫺.25ⴱ 3.39 prevalence 33.4% 15.2% 5% 73% Difference from perfect Estimation of body satisfaction body mass index.19ⴱⴱ 2.96 This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. prevalence 46.9% 15.1% 13% 87.5% Weekly Instagram use.04 0.55 This document is copyrighted by the American Psychological Association or one of its allied publishers. Second-order cultivation measures Instagram browsing ⫺.05 0.63 BESAA weight satisfaction sub-scale1 3.26 0.99 1.00 5.00 a Gender coded with 0 ⫽ female, 1 ⫽ male. ⴱ Average weight ascribed to p ⬍.05. ⴱⴱ p ⬍.01. strangers’ bodiesa 3.25 0.19 2.72 3.75 Third-order cultivation measure EAT-13D scale for disordered believed a higher percentage of Germans to be satisfied with their eatinga 2.40 0.88 1.00 4.77 own body. Again, adding our Instagram variables to the regression Note. BESAA ⫽ Body Esteem Scale for Adolescents and Adults; EAT- model did not lead to a significantly higher R2(⌬R2 ⫽.02; p ⫽ 13D ⫽ German short version of Eating Attitudes Test. N ⫽ 228..09), suggesting that more intense usage of Instagram did not a Scale range: 1–5. connect to changed estimations of prevalence in a statistically noteworthy way (Table 4). In conclusion, we have to reject both Hypotheses 1 and 2. of BMI, the positive coefficient indicates that participants with an increased deviation from a healthy body weight expected a higher Second-Order Cultivation: Attitudes percentage of people in the general population to be obese. Pro- ceeding to the second step of our regression model, however, no Our next hierarchical regression addressed participants’ atti- significant increase in R2 could be observed, ⌬R2 ⬍.01, p ⫽.79. tudes about their own body as measured by the Body Esteem Scale As such, we note that participants’ Instagram use remained mostly for Adolescents and Adults Weight Satisfaction Index (Hypothesis unrelated to their perceptions of the weight distribution in the 3). An initial model using only the three demographic predictors general population (Table 3). resulted in a significant equation, F(3, 223) ⫽ 13.24, p ⫽.00, To explore Hypothesis 2, we conducted a hierarchical linear R2 ⫽.15. More precisely, both age, ␤ ⫽ ⫺.17, t(223) ⫽ 2.49, p ⫽ regression using participants’ assumed prevalence of body satis-.01, and gender, ␤ ⫽.22, t(223) ⫽ 3.28, p ⬍.01, as well as faction in Germany as the criterion. Our first step led to a signif- deviation from a perfect BMI, ␤ ⫽ ⫺.34, t(223) ⫽ 5.43, p ⬍.01, icant regression equation, F(3, 223) ⫽ 2.93, p ⫽.03, albeit with a could be linked to the outcome variable in a significant way. rather low coefficient of determination, R2 ⫽.04. Among the Whereas the difference between participants’ BMI and the perfect entered predictors, only participants’ gender became significant, value of 21.7 actually emerged as the strongest predictor—a less ␤ ⫽.21, t(223) ⫽ 2.95, p ⬍.01, indicating that male respondents ideal BMI went along with higher body dissatisfaction—we also Table 2 Zero-Order Correlations Between Measured Variables Measured variable 1 2 3 4 5 6 7 8 9 10 1. Age — 2. Gendera.37ⴱⴱ — 3. Body mass index.14ⴱ.26ⴱⴱ — 4. Weekly Instagram use ⫺.12 ⫺.22ⴱⴱ.07 — 5. Instagram browsing ⫺.24ⴱⴱ ⫺.39ⴱⴱ.05.53ⴱⴱ — 6. Estimation: National overweight prevalence.18ⴱⴱ ⫺.12.08.05.01 — 7. Estimation: National body satisfaction prevalence ⫺.01.17ⴱⴱ ⫺.04.09.04 ⫺.23ⴱⴱ — 8. Personal weight satisfaction ⫺.11.12 ⫺.43ⴱⴱ ⫺.13 ⫺.12 ⫺.09.17ⴱ — 9. Weight ascribed to strangers’ photos.09 ⫺.05 ⫺.18ⴱⴱ.00.16ⴱ.19ⴱⴱ.03 ⫺.02 — 10. EAT-13D disordered eating index.11 ⫺.17ⴱ.28ⴱⴱ.18ⴱⴱ.23ⴱⴱ.10 ⫺.08 ⫺.67ⴱⴱ.14ⴱ — Note. EAT-13D ⫽ German short version of Eating Attitudes Test. a Gender coded with 0 ⫽ female, 1 ⫽ male. ⴱ p ⬍.05. ⴱⴱ p ⬍.01.  EVERY (INSTA)GRAM COUNTS? 93 Table 4 Table 6 Hierarchical Regression With Participants’ Estimations of the Hierarchical Regression With Participants’ Average Rating of National Body Satisfaction Prevalence as a Criterion Strangers’ Weight as a Criterion Predictor ␤ t F R2 Predictor ␤ t F R2 1 1 Age ⫺.09 1.23 2.93ⴱ.04 Age.12 1.68 1.48.02 Gendera.21ⴱⴱ 2.95 Gendera ⫺.09 1.21 Difference from perfect Difference from perfect body mass index ⫺.04 0.61 body mass index ⫺.07 1.04 2 2 ⴱ Age ⫺.07 1.03 2.76.06 Age.14 2.00 2.59ⴱ.06ⴱ Gendera.26ⴱⴱ 3.43 Gendera ⫺.03 0.37 Difference from perfect Difference from perfect body body mass index ⫺.05 0.71 mass index ⫺.07 1.08 This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Weekly Instagram use.10 1.32 Weekly Instagram use ⫺.11 1.43 This document is copyrighted by the American Psychological Association or one of its allied publishers. Instagram browsing.07 0.91 Instagram browsing.24ⴱⴱ 2.89 3 a Gender coded with 0 ⫽ female, 1 ⫽ male. Age.13 1.87 3.05ⴱⴱ.08ⴱ ⴱ p ⬍.05. ⴱⴱ p ⬍.01. Gendera