MSc Communication Science
Persuasive Communication
Master Thesis
#healthyfood
How Exposure to Instagram Food Posts of Highly Popular
Users Influence Individual’s Intention to Eat Healthy and
Subsequent Food Choice
by
Charlot Lucia Beisswenger
Student ID: 11580496
Supervisor: Dr. Saar Mollen
Submission Date: 29
thJune 2018
Number of Credits: 60 ECTS
Research Period: Spring 2018
Abstract
Social media platforms such as Instagram have developed to become a major part of individuals’ every-day lives. On the photo-sharing platform Instagram, pictures of food are commonly and frequently shared. Observing what others eat through food posts on Instagram may have an impact on individuals’ perception about social norms in regard to eating
behavior. Yet, little is known about the effects of healthy and unhealthy Instagram food posts on individuals’ norm perceptions and subsequent eating behavior. This study examines whether exposure to healthy versus unhealthy food posts by others affects users’ intention to eat healthy, as well as their subsequent healthy or unhealthy food choice. Additionally, the study investigated whether the level of popularity of the poster influences the effect. Results of an online experiment (N = 188) indicate no significant effects of exposure to healthy versus unhealthy Instagram food posts on individuals’ intention to eat healthy and on subsequent food choice. Social norms have not been found to mediate the association between exposure to healthy or unhealthy Instagram food posts and intention to eat healthy and food choice. However, results show that food posts by highly popular users lead to a healthy food choice, irrespective of the content of the posts. Implications of the findings are discussed and
Table of Content
List of Tables ... iii
List of Figures ... iv
Introduction ... 1
Theoretical Framework ... 3
Social modeling of eating and Instagram ... 3
Social norm perception ... 5
Social modeling on Instagram and the role of poster popularity ... 7
Method ... 9
Pre-test ... 9
Results of the pre-test ... 10
Participants and design ... 10
Procedure ... 12 Manipulation ... 13 Measures ... 14 Results ... 17 Randomization check ... 17 Assumptions ... 18
Direct effects and moderation ... 18
Mediation and moderated mediation effects ... 22
Discussion ... 27
Limitations and direction for future research ... 29
Conclusion ... 31
References ... 32
List of Tables
Table 1. Means and Standard Deviation of Intention to Eat Healthy by Condition. ... 19 Table 2. Results of a Univariate Analysis of Variance of Intention to Eat Healthy by Exposure
and Popularity. ... 20
Table 3. Results of a Hierarchical Binary Logistic Regression Analysis Testing the Effect of
Exposure to Healthy or Unhealthy Instagram Food Posts and Level of Popularity on
Subsequent Food Choice. ... 21
Table 4. Results of a Mediation Analyses (Including Moderated Mediation) Predicting
Intention to Eat Healthy. Confidence Intervals and Standard Errors Based on 5000 Bootstrap Samples. ... 24
Table 5. Results of a Mediation Analyses (Including Moderated Mediation) Predicting
Intention to Eat Healthy. Confidence Intervals and Standard Errors Based on 5000 Bootstrap Samples. ... 26
List of Figures
Figure 1. Conceptual Model. ... 9 Figure 2. Results of mediation analysis for descriptive norms. ... 23 Figure 3. Results of mediation analysis for injunctive norms. ... 25
Introduction
Food constitutes an important aspect of our daily lives; it is connected to our identity, culture and lifestyle. The consumption of food often takes place in the presence of others (Cruwys, Bevelander, & Hermans, 2015). Food intake does not only concern nurturing our bodies. In fact, eating constitutes an integral part of our social lives (Sharma & De Choudhury, 2015). Prior research on social modeling and social norms shows that we use the eating behavior of others to guide our own food intake (Herman, Roth, & Polivy, 2003). For instance, it has been shown that what we eat, how much we eat and even how fast we eat is influenced by other people around us (Cruwys et al., 2015). Previous studies revealed that the phenomenon of social modeling of eating can be explained by underlying social and normative influences of such eating behaviors (Herman, & Polivy, 2008; Vartanian, Sokol, Herman, & Polivy, 2013).
The social environment of individuals determines the norms that they have about eating (Cruwys et al., 2015). These influences used to be limited to offline references such as culture, lifestyle, family and friends. However, with the emergence of social networking sites and food posts, people are now exposed to an abundance of social information about the eating behavior of other people online (Hefner, & Vorderer, 2016).
One of these social networking sites is Instagram, a photo-sharing application created in 2010 (Hu, Manikonda, & Kambhampati, 2014). Its photo-driven nature makes Instagram especially appealing for sharing food pictures and the platform has become an integral part of individuals’ daily lives. On Instagram, pictures of both, unhealthy as well as healthy foods, are frequently shared by users. Depending on the content a user is exposed to, one might get the impression that either healthy or unhealthy eating is common among users. A study by Vaterlaus, Patten, Roche, and Young (2015) reveals that social media can have a positive as
well as a negative impact on food choices since food posts on social media can serve as a motivation for healthy as well as unhealthy eating behavior (Vaterlaus, et al., 2015).
While social modeling of eating has been studied extensively in offline contexts in the past, those results need to be validated in today’s online driven world. To date, little is known about the effects of healthy and unhealthy food posts on individual’s eating behavior and the underlying processes, such as norm perceptions, that may explain these effects. In times, when poor eating habits and obesity became a public health issue (Deshpande, Basil, & Basil, 2009), it is crucial to understand how exposure to food related information on social media might influence user’s dietary behavior.
As mentioned before, individuals’ norms about eating behavior are influenced by others in their environment. However, previous studies show that individuals do not conform their behavior equally to others. Rather, individuals are more likely to conform their behavior to individuals whom they perceive as popular (Teunissen, Spijkerman, Prinstein, Cohen, Engels, & Scholte, 2012). Hence, social norms of others are more likely to be internalized if the other person is popular. On social networking platforms such as Instagram, the number of likes and followers provide users with information that help to easily evaluate other’s level of popularity. Despite the fact that prior studies show that popularity can enhance the norm adoption of individuals, it remains unknown whether popularity has similar effects on user’s norm perception and subsequent behavior in social media environments.
Therefore, the current study addresses this question to fill a gap in the scientific literature and to help professionals in health communication to create effective social media-based communication interventions that aim at increasing healthy eating among individuals. The dramatically growing popularity of social media over the last few years (Hu et al., 2014) highlights the importance of the topic. The current paper investigates whether exposure to Instagram food posts influences individual’s intention to eat healthy and subsequent food choice. Additionally, it examines whether the popularity of the poster (the person who shared
a food post) increases or decreases the likelihood of individuals to adopt eating norms. Thus, the following research question is proposed: To what extent can exposure to healthy versus
unhealthy Instagram food posts influence individuals’ intention to eat healthy as well as their food choice? Is the effect influenced by popularity of the poster and explained by social norms?
Theoretical Framework
Social modeling of eating and Instagram
In offline environments, the phenomenon of social modeling of eating has been studied extensively, and previous research showed consistently that individuals tend to adjust their eating behavior to the eating behavior they observe among others. Additionally, the modeling of eating behavior has been found to occur under a variety of circumstances. For instance, direct observation of another person’s food intake (e.g., in a dining situation) can affect individuals’ eating behavior, but the eating behavior of another person can also affect one's eating behavior when one is not physically present (e.g., in a video conference). Moreover, it can occur for familiar others such as friends and family members as well as for strangers (Ball et al., 2010). In fact, effects have also been observed when information about another person's food intake was provided in written form or when cues such as empty food wrappers were visible for individuals (Feeney, Polivy, Pliner, & Sullivan, 2011; Higgs, & Thomas, 2016).
Conformity behavior has been found to occur for healthy as well as unhealthy foods, high and low calorie-dense meals and for snacks (Cruwys et al., 2015). Therefore, it can be concluded, that exposure to various kinds of information about other’s dietary behavior influences the eating behavior of individuals and these results appear to be remarkably consistent (Cruwys et al, 2015).
From an evolutionary perspective, observing what others eat has always brought a great advantage for individuals, since people need to know which foods to choose in order to maintain good health (Higgs, 2015). Hence, modeling the eating behavior of others can serve as nutritional information (Boyd, Richerson, & Henrich, 2011). Besides that, humans model eating behavior of others to demonstrate belonginess to the same social group (Deutsch, & Gerard, 1955), or to show sympathy to another person (Robinson, Tobias, Shaw, Freeman, & Higgs, 2011). However, modeling eating behavior appears to occur automatically since studies have shown that most individuals are not aware of their modeling behavior and state that rather hunger or taste are the reason for their food intake (Vartanian, Herman, &
Wansink, 2008).
As mentioned before, prior research found support for effects of social modeling of eating on individual’s food selection, hence the type of food individuals choose to eat. For instance, Prinsen, de Ridder, and de Vet (2013) manipulated the presence or absence of empty chocolate wrappers indicating whether others, who were in the same situation, had eaten chocolate. The bowl, either empty or filled with chocolate wrappers, was placed in a bakery where costumers passed by after ordering their food. Findings show that individuals who passed by the bowl filled with empty chocolate wrappers conformingly chose chocolate over oat biscuits since they assumed that previous others had also chosen chocolate. Similar to this, a study by Burger and colleagues (2010) manipulated the presence of empty wrappers of healthy versus unhealthy snack bars in the context of an ostensible taste test. In line with findings by Prinsen and colleagues (2013), results show that if participants were to believe that prior participants had chosen the healthy snack bar, participants also chose the healthy snack bar whereas when participants were led to believe that prior others had chosen the unhealthy snack bar, they also chose the unhealthy snack bar. Thus, it can be concluded that in situations where individuals choose between healthy and unhealthy foods, they tend to conform their choice to what they think others had chosen.
As social modeling of eating has been found in a variety of circumstances, with a variety of social models (Cruwys et al., 2015), it seems plausible that exposure to Instagram food posts might change individual’s intention to eat healthy and healthy or unhealthy food choice. The majority of prior studies focused on behavioral outcomes such as food choice (Cruwys, et al., 2015). However, according to the theory of planed behavior (Ajzen, 1991), intention is an important predictor for behavior (such as choosing food) which is why it is likely that social modeling first affects intention which in turn, affects behavior. To date, no study examined whether social modeling of eating affects intention but there is a great deal of evidence showing the effects on food choice (e.g. Prinsen et al., 2013; Burger et al., 2010). In addition, while previous studies did not consider social networking platforms, this study takes into account the increased amount of social information (Hefner, & Vorderer, 2016)
accessible on social media sites such as Instagram.
On these grounds, the present study proposes that exposure to healthy or unhealthy Instagram food posts might influence intention to eat healthy as well as the subsequent behavior to either choose a healthy or unhealthy food item. More specifically, it is proposed that exposure to healthy Instagram food posts affects individuals in such a way that they show a greater intention to eat healthy (H1a) and make a healthy food choice (H1b) compared to individuals exposed to unhealthy Instagram food posts.
Social norm perception
Previous research claims that social norms constitute the underlying concept that leads to modeling the eating behavior of others (e.g. Vartanian Sokol, Herman, & Polivy, 2013). Social norms are defined as “implicit codes of conduct that provide a guideline for […]
action” (Higgs, 2016, p. 38). Put differently, social norms are perceived standards that inform individuals about what constitutes correct and appropriate behavior. The theory of normative conduct (Cialdini, Reno, & Kallgren, 1990) distinguishes between two types of norms,
descriptive and injunctive norms. Descriptive norms refer to what individuals observe most others are actually doing. It is perceived as “what everybody does”. As a result, this type of norm acts as a guideline for individual’s own behavior since they assume that “if everybody is doing it – it must be right” (Bodie, Horan, Gannon, Tufano, & Farrell, 2012, p. 36).
On the other hand, injunctive norms point to what individuals think they are supposed to do according to most others. Injunctive norms describe which behaviors are socially approved by societies or cultures (Cialdini, 1990). Thus, injunctive norms add a moral component to the concept of norms since social reward or sanctioning can have possible consequences. Put together, descriptive norms are also called “the norms of is” whereas injunctive norms are referred to as “the norms of ought” (Kallgren, Reno, Cialdini, 2000).
Descriptive and injunctive norms can overlap since the majority of individuals act accordingly to what is socially approved. However, it is crucial to distinguish between both types and to treat them separately since the underlying motivations for individuals differ substantially (Deutsch, & Gerard, 1955). For descriptive norms, it is the desire to act correctly such as choosing foods that provide us with great energy and health. For injunctive norms, it is the desire to act appropriately to be socially accepted by others (Cialdini, & Goldstein, 2004).
Despite the fact, that prior literature claims that social norms act as underlying mechanism that explains the phenomenon of social modeling of eating, (e.g. Higgs, 2015; Higgs & Thomas, 2016), there is only limited empirical evidence to support this claim. To date, solely Vartanian, Sokol, Herman, and Polivy (2013) investigated whether social norms underlie social modeling of eating. In an experimental study, the researchers presented females with a confederate person who either ate a large or small portion of food in front of the participants. Females who were exposed to a low-intake model ate significantly less compared to females who were exposed to a high-intake model. Additionally, these females reported a lower perceived norm about the appropriate amount of food intake. Thus,
injunctive norm perception constituted a significant mediator between exposure to healthy or unhealthy Instagram food posts and subsequent food intake behavior.
The study by Vartanian and colleagues (2013) serves as a basis for the idea that social norms underlie social modeling of eating which is why the current study assumes that
exposure to healthy or unhealthy Instagram food posts will affect social norm perception which in turn, results in a higher or lower intention to eat healthy and a healthy or unhealthy food choice.
Despite the limited number of modelling studies which include norm perception, there is a large body of scientific literature which focuses on investigating the direct effect of norm-based messages on intention to eat healthy and subsequent food choice. For instance, it has been found that descriptive norm messages have the potential to predict intention to eat healthy (Smith, & McLallen, 2008) and healthy food choices (Stok, de Ridder, de Vet, & Wit, 2014). Therefore, the present paper assumes that exposure to healthy Instagram food posts will lead to greater descriptive norm perception and in turn, result in higher intention to eat healthy (H2a) and a healthy food choice (H3a).
In regard to injunctive norms, a study by Stok and colleagues (2014) shows that injunctive norm messages affect intention to eat healthy. Additionally, Vartanian and colleagues (2013) provide strong support that injunctive norm messages underlie modeling and subsequent eating behavior. On these grounds, the present paper assumes that exposure to healthy Instagram food posts influences injunctive norm perception and in turn, leads to a higher intention to eat healthy (H2b) and a healthy food choice (H3b).
Social modeling on Instagram and the role of poster popularity
Social networking sites, including Instagram, are characterized by their means of indicating sympathy and social endorsement (Sherman, Payton, Hernandez, Greenfield, & Dapretto, 2016). On social media, having a large number of likes and followers means being popular
and represents high social status (Sherman et al., 2016). The number of likes and followers can easily be observed and thus, provide quick insight into a person's level of popularity. Prior research suggests that popular individuals serve as role models for many others (Valente, Unger, & Johnson, 2004). For instance, a study by Teunissen and colleagues (2012) shows that higher perceived popularity of peers leads to greater social norm adoption in the context of alcohol use among others. Result indicate that individuals’ willingness to follow social norms is substantially greater when they are confronted with norms of highly popular others compared to less popular others. Prior research suggests that following the norms of popular peers is motivated by expected social rewards such as becoming popular oneself (Teunissen et al., 2012).
On Instagram, user popularity can easily be assessed by viewing the number of followers on profiles. Since individuals aim at conforming their behavior to what popular others do (Cialdini, & Goldstein, 2004), it is plausible that user popularity increases the norm perception about healthy or unhealthy eating behavior on Instagram.
Therefore, it is assumed that poster popularity has the potential to increase (or decrease) the effect of exposure to Instagram food posts on social norm perception. It is suggested that individuals who are exposed to Instagram food posts of highly popular users show greater descriptive (H4a) and injunctive norm perception (H4b) compared to individuals exposed to Instagram food posts of less popular users which in turn, leads to a greater
Figure 1. Conceptual Model.
Method
Pre-test
Prior to the main study, a pre-test was conducted to test how participants rate the level of healthfulness of Instagram food posts and the level of popularity of the three Instagram profiles that were chosen as stimulus material. A total of N = 41 participants, who were recruited via Facebook, filled in the online survey. Both male and female participants were included in the sample since there was no reason to assume that males and females would differ in their perceptions about poster popularity and healthfulness of food.
First, participants were informed that the study was part of a pre-test for a master thesis and concerned the perception of Instagram food posts. Then, participants were exposed to three Instagram profiles showing a high number of followers. Subsequently, participants were asked to rate on a 5-point Likert scale ranging from 1 (Not at all popular) to 5
(Extremely popular) how popular they perceived the Instagram users (“How popular do you
think these Instagram users are?”). Afterwards, participants were exposed to the same Instagram profiles, this time showing a low number of followers. Again, participants were asked to rate the popularity of these accounts.
Additionally, participants were asked to rate the healthfulness of the food posts (“How healthy do you think the foods are that you just saw?”) on a 5-point Likert scale ranging from 0 (Not at all healthy) to 5 (Extremely healthy). Subsequently, participants were exposed to a second Instagram timeline, this time showing six unhealthy food posts. Again, participants were asked to rate the level of healthfulness for the food posts. Then, participants were asked to rate how similar they perceive the photographic quality of the food post. Lastly,
participants were thanked for their participation.
Results of the pre-test
Results indicate that participants rated the level of photographic quality between the six food posts of the healthy timeline (M = 85.67, SD = 15.03), as well as for the six posts in the unhealthy timeline (M = 71.00, SD = 22.02) as similar. Numbers above 50 indicated a sufficient level of similarity whereas numbers below 50 indicated an insufficient level of similarity.
Additionally, Instagram users with a low number of followers were perceived as less popular (M = 1.44, SD = .78), compared to the high popular profiles (M = 3.37, SD = .77),
t(40) = 11.17, p < .001, d = 2.5, 95% CI [1.578, 2.275]. Consequently, the manipulation was
successful.
In regard to the healthfulness of the food posts, a paired samples t-test showed a significant difference between the two stimuli t(38) = 40.86, p = <.001, d = . 8.95, 95% CI [3.412, 3.768]. Thus, participants rated the food posts of the healthy Instagram timeline as significantly healthier than the unhealthy food posts and the manipulation was successful.
Participants and design
Data for the main study were collected via an online experiment which was available between 4th and 11th of May 2018. Participants were recruited via Facebook. Participants in the pre-test
were not allowed to participate in the main study. A total of 308 participants started the experiment and 210 finished it, of which 9 were excluded through filters as they were male (n = 3), did not own an Instagram account (n = 5), or were under the age of 18 (n = 1).
Additionally, the data set was checked for extreme outliers by multiplying the difference between the upper and lower quartile by 3 interquartile ranges for the variables descriptive norms, injunctive norms as well as intention to eat healthy (Field, 2013, p.24; Hoaglin, & Iglewicz, 1987). One participant was identified and removed from the data set. On average, participants spent 26.09 seconds (SD = 25.27) looking at the Instagram profiles and 39.38 seconds (SD = 22.87) looking at the presented Instagram timelines before clicking the next button. Due to the fact, that human brains require up to one second, depending on cognitive capacity, to correctly identify an image (Trafton, 2016), it was assumed that at least three seconds were required to fully process the popularity stimulus which consisted of three images and at least six seconds to fully process the timelines which contained six food posts. Hence, twelve participants with unrealistic times less than that were excluded. This resulted in a total number of N = 188 participants1. Participants were only females since prior research suggests that modeling effects are more prevalent among females than males (e.g. Hendy, 2002; Hermans, Herman, Larsen, & Engels, 2010; Vartanian, Spanos, Herman, & Polivy, 2015). Participants were between the age of 18 and 63 years old (M = 23.47, SD = 4.29). The majority of participants were students (81.9%) of which 41% were currently enrolled in a Bachelor’s program and 33% in a Master’s program. Bachelor’s degree was the most
common completed level of education (42.6%), followed by High school diploma (37.2%). In regard to employment status (students were included), 39.4% of participants reported working part-time, 33% were unemployed and 23.9% worked a full-time job. The majority of
participants were German (63.3%), followed by British (9.6%) and Dutch (6.9%)
1
Additionally, 35 participants were identified who uncovered the true purpose of the study. When excluding those participants from the analysis results remain nonsignificant.
nationalities. The majority of participants (70.2%) reported checking Instagram several times a day, 17.6% reported checking Instagram daily and the remaining 11.2 % reported checking it either several times per week (7.4%) or less.
The distribution of participants among the four conditions resulted in the following group sizes: low popularity x healthy: 45 participants, low popularity x unhealthy: 45 participants, high popularity x healthy: 54 participants, high popularity x unhealthy: 44 participants.
Procedure
Participants were randomly assigned to one of four conditions in a 2 (food post: healthy vs. unhealthy) x 2 (user popularity: high vs. low) between-subject factorial design (post-test-only). By clicking on the link, participants entered the online experiment where they were informed that the research concerned the perceived quality and aesthetic of Instagram food posts to cover the true purpose. Two filter questions assessed sex and whether participants owned an Instagram account; males and people who did not own an Instagram account were removed from the experiment. Demographics and Instagram use intensity were measured. Subsequently, participants were assigned to one of the four conditions. Firstly, participants were asked to take a look at the three presented Instagram accounts of young women, which either showed a high or low number of followers. Next, participants were told that they will be presented with an Instagram time line which includes food posts of the previously shown Instagram users. Participants were asked to imagine that this was their own Instagram time line. After exposure to either healthy or unhealthy Instagram timeline, the outcome variables food choice and intention to eat healthy were assessed, as well as the mediating variables descriptive and injunctive norm perception. Filler items were included which asked participants to rate the level of photographic quality and to rate how aesthetically pleasant they perceived the posts to support the cover story.
Finally, participants were asked what they think the goal of the research was and could voluntarily enter their email address for a chance to win their previously made food choice as a prize. Then, participants were debriefed and thanked for their contribution to scientific research.
Manipulation
User popularity. Participants were exposed to three Instagram profiles which differed
in the numbers of followers (high vs. low). According to marketing experts, macro
influencers are Instagram users who have more than 100,000 followers and hence, are seen to be highly popular individuals on social media platforms such as Instagram. Users who have more than one million followers are more likely to be considered celebrities (Wolfson, January, 2017), which is why numbers in the range of 100 to 200 thousand were chosen for the high popularity condition. On that basis, the three Instagram profiles in the high
popularity condition had the following numbers of followers: Profile 1 showed 146 thousand followers, profile 2 showed 155 thousand followers, and profile 3 showed 108 thousand followers.
For the low popularity condition, statistics about the average number of followers on Instagram served as an indication. Statistics show that most Instagram users (25%) have between 0 and 100 followers, 11% have between 100 and 200 followers, 6% have between 201 and 300 followers and 18% have more than 300 followers (Statista, 2018). Thus, numbers for the low popularity condition were chosen accordingly: Profile 1 showed 31 followers, profile 2 showed 147 followers and profile 3 showed 333 followers.
The profile pictures of each account showed young women and the same pictures were used for low and high popularity conditions. However, no face or other personal indication were visible and names were made incognizable to avoid allowing to draw conclusions about the true identity of these users.
Food posts. Twelve food posts (six in each condition) from various international
Instagram food bloggers were retrieved between 20th and 26th of April. Four of the six healthy food posts showed colorful salad bowls with fresh vegetables and leafy greens, the remaining two showed a breakfast bowl with bananas and strawberries and wraps filled with leafy greens and vegetables. Inclusion criterion for the food posts were a similar level of photographic quality, as well as displaying mainly fruits and vegetables. The unhealthy condition included six posts of which three displayed savory foods (curly fries, fried onions rings, and fries). The remaining three posts displayed sweet foods, namely, donuts, waffles with ice cream and chocolate milkshakes. The goal was to find foods which are commonly regarded as unhealthy.
To enhance external validity and to make the experience for participants as true to reality as possible, the Instagram layout was copied. Participants had to scroll up and down in the same way how Instagram users do when checking their own timeline. However, likes, comments, and names were not visible, to ensure that the manipulated number of followers was the only metric which indicated the level of popularity.
Measures
Instagram Usage Intensity. This variable was assessed using the following item:
“How often do you check Instagram?”. Participants indicated their Instagram use intensity on scale ranging from 1 (Less than once a month) to 7 (Several times per day). Participants indicated an average usage of M = 4.85 (SD = 94).
Perceived descriptive norms. Similar to Ball and colleagues (2010), descriptive norm
perception of healthy and unhealthy eating was assessed by asking participants whether they perceive that other females their age eat healthy. The original item “Lots of women I know eat healthy food when they are out” (Ball et al., 2010, p. 3) was transformed and extended to the following four items: “Most women my age eat healthy”, “The majority of women my age
try to maintain a healthy diet”, “Most women my age frequently eat healthy foods such as fruits and vegetables”, “The majority of women my age frequently eat candy such as chocolate and fried foods” (reversed). Participants were asked to rate their level of agreement on a 7-point Likert scale ranging from 1 (Strongly disagree) to 7 (Strongly agree).
Before computing a scale, an exploratory factor analysis using principal axis factoring was conducted since the scale was created by the researcher and has not been tested before. The Eigenvalue criterion showed one underlying dimensions (factor 1: 2.02). The factor explains 50.49% of the variance in descriptive norms. Assumptions of KMO and Bartlett’s test were met (.653, p < .001). Item 4 loaded very low (.396) and impaired the reliability score, which is why it was not included in the final scale. The scale with the remaining three items showed acceptable internal consistency (Cronbach’s ⍺ = .73; M = 3.80, SD = .96).
Perceived Injunctive Norms. Participants’ perception of injunctive norms about
appropriate eating behavior was assessed by asking participants to indicate their level of agreement on a 7-point Likert scale ranging from 1 (I strongly disagree) to 7 (I strongly
agree) with the following statements: “Most women my age think that eating healthy is
important”, “Most women my age think one should eat healthy”, "The majority of women my age approve of maintaining a healthy diet”, “Most women my age think that eating unhealthily is inappropriate”, “Most women my age approve of frequently eating fruits and vegetables”, and one reversed item: “The majority of women my age approve of eating unhealthy foods such as chocolate and fried foods”.
Again, an exploratory factor analysis using principal axis factoring was conducted since the items have not been previously tested. Assumptions of KMO and Bartlett’s test were met (.678, p < .001). The Eigenvalue criterion showed two underlying dimensions (factor 1: 2.66, factor 2: 1.01). Together, the two factors explain 61.27% of the variance in the items. Item 6 loaded separately and impaired the reliability score which is why it was not included in
the scale. The scale including the remaining five items showed acceptable internal consistency (Cronbach’s ⍺ = .731; M = 5.23, SD =.82).
Intention to eat healthy. Based on Connor, Norman, and Bell (2002) as well as Rhodes
and colleagues (2007), five items, asking participants for their level of agreement on a 7-point Likert scale, were used. Items included the following statements: “I intend to eat a healthy diet in the coming days”, “I want to eat a healthy diet in the coming days”, “I try to eat a healthy diet in the coming days”, “I expect to eat healthily in the coming days”, “In the coming days, it is very likely that I will eat healthy”. Items of Connor and colleagues (2002) as well as Rhodes and colleagues (2007) were adapted to fit the current study.
An exploratory factor analysis using principal axis factoring was conducted since the scale has not been previously tested. Assumptions of KMO and Bartlett’s test were met (.872,
p < .001). The Eigenvalue criterion showed one underlying dimensions (factor 1: 3.62). This
factor explains 72.34% of the variance in the four items. The scale showed excellent internal consistency (Cronbach’s ⍺ = .902, M = 5.23, SD = 1.19).
Food choice. The majority of previous studies measured food choice as behavioral
outcome variable in offline experiments and let participants make a selection of various healthy and unhealthy food products such as chocolate and oat bars (e.g. Prinsen et al., 2013, Burger et al., 2010). To fit the outcome variable to an online experiment, participants were given the option of choosing between a fruit basket from an organic farm in Germany and a Lindt chocolate gift basket, which they could win in a raffle. Both prizes had the same value (44.99 Euro) and both were claimed to be vegan and vegetarian friendly and would be delivered to participants’ address. In total, 117 participants (62.2%) chose the fruit and vegetable basket whereas 71 participants (37.8%) decided for the chocolate gift basket.
Analyses
Plan of analysis
First, results of randomization checks are presented. Secondly, assumptions for analysis of variance (ANOVA) and logistic regression are explicated. Then, an ANOVA is conducted to test whether exposure to healthy or unhealthy Instagram food posts affects individuals’ intention to eat healthy (H1a). Subsequently, a logistic regression is performed to test the effect of exposure to healthy or unhealthy Instagram food posts on food choice. (H1b). For the sake of completeness, it was tested whether popularity has an effect on the outcome variables even though there were no explicit hypotheses regarding it.
Finally, a mediation analysis using PROCESS by Hayes (2013) is performed to test whether exposure to healthy or unhealthy Instagram food posts influences descriptive (H2a) or injunctive norm perception (H3a) which in turn, affects intention to eat healthy.
Additionally, it is examined whether level of popularity influences descriptive (H4a) or injunctive norm perception (H4b) which in turn, affects intention to eat healthy.
Lastly, to test whether exposure to healthy or unhealthy Instagram food posts
influences descriptive (H2b) and injunctive norm perception (H3b) which in turn affects food choice, the approach by Kenny (2018) is chosen, since food choice constitutes a nominal variable and can therefore not be included in PROCESS by Hayes (2013).
Results
Randomization check
A randomization check was conducted to assess if all variables were equally distributed among the four conditions. Results showed no significant difference between the groups for the variables age (F(3, 184) = .87, p = .455), Instagram use intensity χ2 (18, n = 188) = 19.46,
p = .364, level of completed education χ2 (12, n = 188) = 10.43, p = .578), being a student χ2
of origin χ2 (63, n = 188) = 81.73, p = .056) were equally distributed among the four conditions. Hence, there was no need to include control variables (Gruijters, 2016).
Assumptions
Prior to the main analysis, all assumptions for ANOVA and binary logistic regression were examined. For ANOVA, the requirement of normally distributed sample was not fully met since the outcome variable intention showed a slightly negatively skewed distribution. However, due to the central limit theorem which implicates that normalized sums tend towards normal distribution, and due to the large sample size of N = 188, a normal distribution was assumed (Field, 2009, p. 134, Lewis-Beck, 1980, p. 30; Weber & Fuller, 2011, p. 169). Measures for heteroscedasticity are reported separately for each analysis where it applies to indicate whether the variance of the residuals of the independent variables is constant. Furthermore, the assumptions of independent cases, meaning that each observation only occurs in one condition, as well as a binary dependent variable, for logistic regression were met.
Direct effects and moderation
In the following, the direct effects of exposure to healthy or unhealthy Instagram food posts on intention to eat healthy (H1a) and food choice (H1b), as well as the role of popularity are analyzed.
Intention. H1a assumed that exposure to healthy or unhealthy Instagram food posts
affects individuals’ intention to eat healthy. In greater detail, it was suggested that exposure to healthy Instagram food posts leads to a greater intention to eat healthy compared to exposure to unhealthy Instagram food posts.
The two independent variables exposure to healthy or unhealthy Instagram food posts and level of popularity have a nominal level of measurement whereas the dependent variable
intention has an interval/ratio level of measurement. An ANOVA was conducted and Levene’s test remained nonsignificant (F(3,184) = .751, p = .523) which is why equal variances were assumed.
Against the prediction of this study, descriptive statistics indicate that the healthy conditions did not report a higher intention to eat healthy compared to the unhealthy
conditions. Results show that exposure to healthy or unhealthy Instagram food posts did not result in different intentions to eat healthy (F(1, 184) = .097, p = .756). Consequently, H1a was rejected. For means and standard deviations see table 1. For sums of square see table 2.
In regard to popularity, results show neither a significant main effect of popularity on intention to eat healthy (F(1, 184) = 0.97, p = .756), nor a significant interaction effect with exposure to Instagram food posts (F(1, 184) = .41, p = .522). Hence, results indicate no effect of popularity on intention to eat healthy.
Table 1. Means and Standard Deviation of Intention to Eat Healthy by Condition.
Condition N M SD
Con. 1: Healthy – Low Popularity
45 5.28 1.13
Con. 2: Unhealthy – Low Popularity
45 5.12 1.32
Con. 3: Healthy – High Popularity
54 5.22 1.18
Con. 4: Unhealthy – High Popularity
44 5.29 1.14
Table 2. Results of a Univariate Analysis of Variance of Intention to Eat Healthy by Exposure
and Popularity.
Source SS df MS F p
Exposure to healthy or unhealthy
Instagram food posts
.08 1 .08 .06 .804
Level of Popularity .14 1 .14 .09 .756
Interaction .59 1 .59 .41 .522
Error 262.79 184 1.43 – –
Total 263.57 188 – – –
Note. N = 188. Minimum score 1, maximum score 7.
Food Choice. Despite the fact, that no effect of exposure to healthy or unhealthy
Instagram food posts on intention was found, the effects on food choice were tested as well. H1b of this paper assumed that exposure to healthy or unhealthy Instagram food posts influences participants’ subsequent food choice. In greater detail, it was suggested that exposure to healthy Instagram food posts leads to a healthy food choice, whereas exposure to unhealthy Instagram food posts leads to an unhealthy food choice. Additionally, the role of popularity was examined.
To assess the effect of exposure to healthy or unhealthy Instagram food posts and level of popularity on food choice (H1b), a binary hierarchical logistic regression was performed. Variables were entered stepwise into the model. The independent variables exposure to healthy or unhealthy Instagram food posts and level of popularity have a nominal level of measurement. Food choice served as the outcome variable and has a nominal level of measurement as well. An additional interaction term between the two independent variables
was built. In total, 117 participants chose the fruit and vegetable basket whereas 71
participants decided for the chocolate gift basket. Results reveal that exposure to healthy or unhealthy Instagram food posts does not significantly predict subsequent food choice (b = .12,
p = .905, SE = .97, OR = .01, 95% CI [.16, 7.6]). Hence, no support for H1b was found.
In a second step, level of popularity was added to the model and results show a significant main effect of popularity on food choice (b = .64, p = .036, SE = .30, OR = 4.41, 95% CI [1.04, 3.45]). As a third step, the interaction term was added but showed no
significant interaction between popularity and exposure to healthy or unhealthy Instagram food posts (b = –.01, p = .983, SE = .61, OR = .00, 95% CI [.29, 3.26]). Hence, it was found that level of popularity had a significant and positive direct effect on food choice indicating that exposure to high levels of popularity lead to a healthy food choice.
Table 3. Results of a Hierarchical Binary Logistic Regression Analysis Testing the Effect of
Exposure to Healthy or Unhealthy Instagram Food Posts and Level of Popularity on Subsequent Food Choice.
Predictor B SE OR p Exposure to healthy or unhealthy Instagram food posts .126 .301 .175 .676 Level of Popularity .636 .305 4.34 .037* Interaction –.013 .611 .000 .983 Constant –1.623 1.513 1.151 –
Mediation and moderated mediation effects
Furthermore, H2a, H2b, H3a and H3b proposed that descriptive as well as injunctive norm perception mediates the relationship between exposure to healthy or unhealthy Instagram food posts and the two outcome variables intention to eat healthy and subsequent food choice. Additionally, it was proposed that popularity moderates the mediation and that a high level of popularity leads to a greater descriptive (H4a) and injunctive norm perception (H4b).
Intention. H2a and H3a assumed that descriptive social norms respectively injunctive
social norms mediate the relationship between exposure to healthy or unhealthy Instagram food posts and intention to eat healthy. The independent variable exposure to healthy or unhealthy Instagram food posts has a nominal level of measurement whereas the mediating variables descriptive and injunctive norm perception as well as the outcome variable intention to eat healthy have an interval/ratio level of measurement. To test the assumptions of H2a, H3a, H4a, and H4b the SPSS macro PROCESS by Hayes (2013) with model 7 and a 5000 bootstrap sample was used.
In regard to descriptive norm perception (H2a), results show that exposure to healthy/unhealthy Instagram food posts does not significantly affect descriptive norm perception (b = –.05, t = –. 13, p = .900, SE = .45, 95% CI [–.93, 82]). Furthermore, there is no significant effect of descriptive norm perception on intention to eat healthy (b = .07, t = – .77, p = .440, SE = .09, 95% CI [–.11, .25]). In sum, descriptive norms have not been found to mediate the relationship between exposure to healthy/unhealthy Instagram food posts and intention to eat healthy and H2a was rejected.
Moreover, results of the moderated mediation show that popularity has no significant main effect on descriptive norms (b = –.20, t = –. 45 p = .651, SE = .44, 95% CI [–1.07, .67]) nor does popularity significantly moderate the association between exposure to healthy/ unhealthy Instagram food posts and descriptive norm perception (b = .12, t = –.42, p = .672,
SE = .29, 95% CI [–.44, .68]). Thus, level of popularity does not increase or decrease
descriptive norm perception and H4a was rejected. For more details see figure 2 and table 4.
Table 4. Results of a Mediation Analyses (Including Moderated Mediation) Predicting
Intention to Eat Healthy. Confidence Intervals and Standard Errors Based on 5000 Bootstrap Samples.
Predictor B SE t p BC 5000 Boot
LL95 UL95 Descriptive norm perception (Mediator 1)
Constant 3.92 .686 5.711 .000 2.566 5.275,
Exposure to Instagram food
posts –. 056 .445 –.126 .900 –.934 .822
Popularity –.201 .443 –.453 .651 –1.076 .674
Interaction .121 .285 .424 .672 –.441 .682
Intention to eat healthy (DV)
Constant 5.037 .425 11.857 .000 4.199 5.875
Exposure to Instagram food
posts –.050 .174 –.288 .774 –.394 .294
Descriptive Norm Perception .070 .090 .773 .440 –.108 .248
Conditional indirect effect on intention to eat healthy if popularity is high or low
Popularity Effect BootSE BootLLCI BootULCI
High .005 .020 –.037 .049
Low
.013 .027 –.034 .079
Note. N = 188. The unstandardized regression coefficients (B) have been reported.
In regard to injunctive norm perception (H3a), results show that exposure to healthy/unhealthy Instagram food posts does not significantly affect injunctive norm
perception (b = –.51, t = 1.36, p = .174, SE = .38, 95% CI [–.23, 1.25]). Furthermore, there is no significant effect of injunctive norm perception on intention to eat healthy (b = .16, t = 1.46, p = .145, SE = .11, 95% CI [–.05, 37]). In sum, injunctive norm perception has not been
found to mediate the relationship between exposure to healthy/unhealthy Instagram food posts and intention to eat healthy and H3a was rejected.
Additionally, results of the moderated mediation show that popularity has no
significant direct effect on injunctive norm perception (b = .49, t = –1.33, p = .186, SE = .37, 95% CI [–.24, 1.23]) nor does popularity significantly moderate the association between exposure to healthy/unhealthy Instagram food posts and injunctive norm perception (b = –.29,
t = –1.19, p = .233, SE = .24, 95% CI [–.76, .19]). Hence, level of popularity does not increase
or decrease injunctive norm perception and H4b was rejected. For more detail see figure 3 and table 5.
Table 5. Results of a Mediation Analyses (Including Moderated Mediation) Predicting
Intention to Eat Healthy. Confidence Intervals and Standard Errors Based on 5000 Bootstrap Samples.
Predictor B SE t p BC 5000 BOOT
LL95 UL95 Injunctive norm perception (Mediator 2)
Constant 4.368 .578 7.553 .000 3.227 5.509
Exposure to Instagram food
posts .512 .375 1.356 .174 –.288 1.251
Popularity .496 .374 1.327 .186 –.241 1.233
Interaction –.287 .240 –1.197 .233 –.760 .186
Intention to eat healthy (DV)
Constant 4.498 .605 7.440 .000 3.305 5.691
Exposure to Instagram food
posts –.056 .174 –.321 .748 –.398 .287
Injunctive Norm Perception .155 .106 1.463 .145 –.054 .365
Conditional indirect effect on intention to eat healthy if popularity is high or low
Popularity Effect BootSE BootLLCI BootULCI
High .035 .037 –.015 .128
Low –.010 .035 –.077 .074
Note. N = 188. The unstandardized regression coefficients (B) have been reported.
Food Choice. Furthermore, H2b and H3b assumed that the relationship between
exposure to healthy or unhealthy Instagram food posts on subsequent food choice is mediated by descriptive norms respectively injunctive norms. To test these assumption, the approach of Kenny (2018) was used which consists of the following four steps. First, the independent variable has to be a significant predictor of the outcome variable. Second, the independent
variable needs to be correlated with the mediator. Third, the mediator has to affect the outcome variable. Lastly, when controlling for the mediator, the association between the independent and dependent variable becomes zero. Only if all four steps are fulfilled, a mediation process is supported. Applying these steps to the current study, it was first tested whether exposure to healthy or unhealthy Instagram food posts affects subsequent food. This effect reflects the concept of H1b and has already been tested. Results showed previously that there is no significant effect of exposure to healthy or unhealthy Instagram food posts on subsequent food choice as well as no significant interaction with popularity (see p. 21 in this paper).
Since the first step of Kenny’s approach (2013) showed that there is no effect of exposure to healthy or unhealthy Instagram food posts on subsequent food choice, the first criterion for a mediation process is not fulfilled. To conclude, there was no mediation process of descriptive norms between exposure to healthy or unhealthy Instagram food posts and food choice, and H2b and H3b were rejected.
Discussion
The aim of this study was to shed light on the question whether the phenomenon of social modeling of eating can be applied to a social media context. To be more concrete, it was tested whether information about healthful versus unhealthful eating behavior, in the form of Instagram food posts, affects other users’ intention to eat healthy and their subsequent healthy or unhealthy food choice. It was expected that exposure to healthy Instagram food posts leads to a higher intention to eat healthy as well as to a healthy food choice compared to exposure to unhealthy Instagram food posts. Furthermore, it was examined whether social norms act as possible underlying mechanism to explains the effects. Additionally, the moderating role of popularity of the poster was explored. To the best of the author’s knowledge, the current
study is the first to examine the role of (poster) popularity in the light of social modeling of eating. It was proposed that popularity acts as a moderator and that high user popularity increases social norm perception which in turn increases intention to eat healthy and a healthy or unhealthy subsequent food choice. However, results were not consistent with the posed hypotheses.
The nonsignificant results might be due to insufficient identification of the participants with the three presented Instagram accounts. Literature about social modeling of eating argues that identification is a key component which triggers modeling behavior. In a first step,
individuals compare themselves to the norm referent to decide whether the norm is relevant to them (Higgs, & Thomas, 2016). Only if identification is strong enough, individuals consider matching their behavior to the referent person. The current study exposed participants to three Instagram accounts showing profile pictures of young females. Hence, age, gender and
Instagram usage served as identification characteristics. However, it might be that these characteristics have not been sufficient to trigger identification. Especially in online contexts where an abundance of social information is accessible, identification could become more important since individuals have to select corresponding others.
Furthermore, the study examined whether social norms explain the modelling effects of eating behavior. Since it was unclear from previous research which type of norm underlies the effect of social modeling of eating, descriptive and injunctive norms were tested. Neither type of norm was found to mediate the process between exposure to healthy or unhealthy Instagram food posts and intention to eat healthy as well as food choice. This contradicts findings by Vartanian and colleagues (2013) who found indeed that injunctive norms act as a mediator between modeling food intake and subsequent food consumption. As mentioned above, a lack of identification might serve as an explanation.
Support for this explanation is provided by Åstr⊘sm and Rise (2001), who show that
moderator between norm perception and intention to healthy. Identification has the power to determine whether individuals perceive a norm as relevant (Åstr⊘sm, & Rise, 2001). Only if
social norms appear applicable, individuals are motivated to follow them. Without identification, individuals might acknowledge social norms about eating, but lack a motivation to match their behavior.
Unexpectedly, results showed that the perception of how popular an Instagram user is influences individual’s food selection. Results suggest that individuals who were exposed to highly popular users were more likely to make a healthy food choice compared to individuals exposed to less popular others. Interestingly, this effect is not depending on the content they were exposed to (healthy versus unhealthy food posts). A possible explanation for this unexpected finding might be that individuals tend to associate popular others (having a high number of friends and being attractive and successful) with healthy eating habits (König, Giese, Stok, &Renner, 2017). Being exposed to popular others on Instagram might increase individual’s motivation to eat healthy which is why participants were more likely to choose healthy over unhealthy foods. Why popularity is linked to healthy eating might be explained by social and cultural images of foods and social status. Nevertheless, more research is needed to clarify this. Results of the present study can be used as a basis for professionals in health communication who aim at increasing healthy eating. For instance, health interventions could expose young women to narratives in form of short video clips featuring popular
females making healthy food choices.
Limitations and direction for future research
There are several limitations to this study which will be discussed in the following. The sample consisted of intense Instagram users since the majority of participants checked Instagram several times a day (70.2%). This repeated behavior serves as an indication for a developed habit among users. Habits are usually performed automatically,
hence, without much attention or consciousness (Peng, Kim, & Larose, 2010). Thus, it could be possible, that due to a lack of awareness and attention Instagram food posts had no
influence on intention and food choice in this study.
Moreover, the fact that it was not participant’s own Instagram timeline to which they were exposed to, individuals might not have felt a social connection to the content or the poster. Usually, Instagram users choose carefully, whom they follow and whose content they want to see. Hence, being exposed to random food posts might not be as effective as seeing food posts published by friends, family or at least people who they have chosen to follow. Again, the concept of identification might play a role here. Prior studies used “in-group” norms such as belonging to a certain university to trigger a sense of connection and identity and in turn, make the perceived norms relevant to individuals (e.g. Cruwys et al., 2015).
Furthermore, it is plausible that testing the variable food choice in an online
experiment makes participants especially aware of the fact that this variable is in the interest of the researcher. While previous (offline) experiments manipulated food choice by letting participants choose between a healthy and unhealthy snack as a reward at the end of the experiment or included it in a mock test, it was less obvious for participants that this was part of the experiment. Hence, participant’s behavior was less biased and more natural when choosing the food.
Future research can place more focus on the identification process to ensure a greater likelihood for a change in the perception of eating norms. More research is needed to identify the attributes that are required to elicit identification processes on social networking sites. Future studies can include a measure which assesses the extent to which participants identified with the reference persons to have more insight in whether identification took place or vary the level of identification.
Moreover, researchers can apply long-term methods and observe whether norms become more salient if individuals are exposed to healthy or unhealthy Instagram food posts
on a regular basis. Thereby, participants’ own Instagram could be used to make the content more relevant.
Conclusion
This study was the first to apply the phenomenon of social modeling of eating to the social media platform Instagram, where users frequently share healthy as well as unhealthy food posts. In a time when Instagram becomes increasingly dominant in peoples’ lives (Lup, Trub, & Rosenthal, 2015), the present study provides valuable insights. However, exposure to healthy or unhealthy Instagram food posts has not been found to influence individual’s intention to eat healthy or subsequent food choice. In addition, poster popularity has not been found to increase or decrease this relationship. Lastly, social norms did not constitute an underlying mechanism that could explain the effect of exposure to healthy or unhealthy Instagram food posts on intention to eat healthy and food choice.
Unexpectedly, popularity was found to directly influence food choice in such a way that exposure to highly popular users leads to a healthy food choice. Furthermore, the study showed that identification requires more attention in the context of social modeling of eating. Especially with the abundance of social information on platforms such as Instagram,
identification constitutes an important determinant for social norm perception. Despite the fact that no support for the posed hypotheses was found, it is still plausible that observing what others eat on social media influences individuals’ norm perception and future research can dig deeper into this urgent field of research.
References
Ajzen, I. (1991). The theory of planned behavior. Organizational behavior and human
decision processes, 50(2), 179–211.
Åstr⊘ sm, A. N., & Rise, J. (2001). Young adults' intention to eat healthy food: Extending the theory of planned behaviour. Psychology and Health, 16(2), 223–237.
Ball, K., Jeffery, R. W., Abbott, G., McNaughton, S. A., & Crawford, D. (2010). Is healthy behavior contagious: associations of social norms with physical activity and healthy eating. International Journal of Behavioral Nutrition and Physical Activity, 7(1), 86 – 95.
Bodie, Z., Horan, S., Gannon, J., Tufano, P., & Farrell, C. (2012). Panel Discussion on Consumer Finance.
Boyd, R., Richerson, P. J., & Henrich, J. (2011). The cultural niche: Why social learning is essential for human adaptation. Proceedings of the National Academy of Sciences,
108(2), 10918–10925.
Burger, J. M., Bell, H., Harvey, K., Johnson, J., Stewart, C., Dorian, K., & Swedroe, M. (2010). Nutritious or delicious? The effect of descriptive norm information on food choice. Journal of Social and Clinical Psychology, 29(2), 228–242.
Cialdini, R. B., & Goldstein, N. J. (2004). Social influence: Compliance and conformity.
Annu. Rev. Psychol., 55, 591–621.
Cialdini, R. B., Reno, R. R., & Kallgren, C. A. (1990). A focus theory of normative conduct: Recycling the concept of norms to reduce littering in public places. Journal of
personality and social psychology, 58(6), 1015–1026.
Conner, M., Norman, P., & Bell, R. (2002). The theory of planned behavior and healthy eating. Health psychology, 21(2), 194–224.
Cruwys, T., Bevelander, K. E., & Hermans, R. C. (2015). Social modeling of eating: A
review of when and why social influence affects food intake and choice. Appetite, 86, 3–18.
Deshpande, S., Basil, M. D., & Basil, D. Z. (2009). Factors influencing healthy eating habits among college students: An application of the health belief model. Health marketing
quarterly, 26(2), 145–164.
Deutsch, M., & Gerard, H. B. (1955). A study of normative and informational social
influences upon individual judgment. The journal of abnormal and social psychology,
51(3), 629–636.
Feeney J., Polivy J., Pliner P., Sullivan M. D. (2011). Comparing live and remote models in eating conformity research. Eat Behav, 12, 75–77.
Field, A. (2009). Discovering statistics using SPSS. Sage Publications.
Gruijters, S. L. (2016). Baseline comparisons and covariate fishing: Bad statistical habits we should have broken yesterday. The European Health Psychologist, 18(5), 205–209.
Hayes, A. F. (2009). Beyond Baron and Kenny: Statistical mediation analysis in the new millennium. Communication monographs, 76(4), 408–420.
Hefner, D., & Vorderer, P. (2016). Permanent Connectedness and Multitasking. In The
Routledge Handbook of Media Use and Well-Being: International Perspectives on Theory and Research on Positive Media Effects, 237–249. New York, NY: Routledge.
Hendy, H. M. (2002). Effectiveness of trained peer models to encourage food acceptance in preschool children. Appetite, 39(3), 217–225.
Herman, C. P., & Polivy, J. (2008). External cues in the control of food intake in humans: the sensory-normative distinction. Physiology & Behavior, 94(5), 722–728.
Handbook of gender research in psychology (pp. 455-469). Springer, New York, NY.
Hermans, R. C., Herman, C. P., Larsen, J. K., & Engels, R. C. (2010). Social modeling effects on snack intake among young men. The role of hunger. Appetite, 54(2), 378-383.
Herman, C. P., Roth, D. A., & Polivy, J. (2003). Effects of the presence of others on food intake: a normative interpretation. Psychological bulletin, 129(6), 873–886.
Higgs, S. (2015). Social norms and their influence on eating behaviours. Appetite, 86, 38–44.
Higgs, S., & Thomas, J. (2016). Social influences on eating. Current Opinion in Behavioral
Sciences, 9, 1–6.
Hoaglin, D. C., & Iglewicz, B. (1987). Fine-tuning some resistant rules for outlier labeling.
Journal of the American Statistical Association, 82(400), 1147–1149.
Hu, Y., Manikonda, L., & Kambhampati, S. (2014, June). What We Instagram: A First Analysis of Instagram Photo Content and User Types. In ICWS.
Kallgren, C. A., Reno, R. R., & Cialdini, R. B. (2000). A focus theory of normative conduct: When norms do and do not affect behavior. Personality and social psychology
bulletin, 26(8), 1002–1012.
Kenny, A. D. (January, 26, 2018). MEDIATION. Retrieved 28th May, 2018 from: http://davidakenny.net/cm/mediate.htm
König, L. M., Giese, H., Stok, F. M., & Renner, B. (2017). The social image of food Associations between popularity and eating behavior. Appetite, 114, 248–258.
Lewis-Beck, M. S. (1980). Applied Regression. An Introduction. Newbury Park: Sage.
Lup, K., Trub, L., & Rosenthal, L. (2015). Instagram# instasad?: exploring associations among instagram use, depressive symptoms, negative social comparison, and
252.
Mayer, R. E. (Ed.). (2005). The Cambridge handbook of multimedia learning. Cambridge university press.
Peng, W., Kim, J., & Larose, R. (2010). Social networking: Addictive, compulsive,
problematic, or just another media habit?. In A networked self (pp. 67-89). Routledge.
Prinsen, S., de Ridder, D. T., & de Vet, E. (2013). Eating by example. Effects of environmental cues on dietary decisions. Appetite, 70, 1–5.
Robinson, E., Thomas, J., Aveyard, P., & Higgs, S. (2014). What everyone else is eating: a systematic review and meta-analysis of the effect of informational eating norms on eating behavior. Journal of the Academy of Nutrition and Dietetics, 114(3), 414–429.
Sharma, S. S., & De Choudhury, M. (2015, May). Measuring and characterizing nutritional information of food and ingestion content in instagram. In Proceedings of the 24th
International Conference on World Wide Web, 115–116, ACM.
Sherman, L. E., Payton, A. A., Hernandez, L. M., Greenfield, P. M., & Dapretto, M. (2016). The power of the like in adolescence: effects of peer influence on neural and
behavioral responses to social media. Psychological science, 27(7), 1027–1035.
Smith-McLallen, A., & Fishbein, M. (2008). Predictors of intentions to perform six cancer- related behaviours: roles for injunctive and descriptive norms. Psychology, Health and
Medicine, 13(4), 389– 401.
Statista (2018b). Average number of Instagram followers of teenage users in the United States as of March 2015. Retrieved 22, March, 2018 from
https://www.statista.com/statistics/419326/us-teen-instagram-followers-number/
Trofton, A. (Januray, 16, 2014). In the blink of an eye. MIT neuroscientists find the brain can identify images seen for as little as 13 milliseconds. Retrieved June, 16, 2018 from http://news.mit.edu/2014/in-the-blink-of-an-eye-0116
Valente, T. W., Unger, J. B., & Johnson, C. A. (2005). Do popular students smoke? The association between popularity and smoking among middle school students. Journal
of Adolescent Health, 37(4), 32–329.
Vaterlaus, J. M., Patten, E. V., Roche, C., & Young, J. A. (2015). # Gettinghealthy: The perceived influence of social media on young adult health behaviors. Computers in
Human Behavior, 45, 151–157.
Vartanian, L. R., Herman, C. P., & Wansink, B. (2008). Are we aware of the external factors that influence our food intake?. Health Psychology, 27(5), 533–538.
Vartanian, L. R., Sokol, N., Herman, C. P., & Polivy, J. (2013). Social models provide a norm of appropriate food intake for young women. PLoS One, 8(11), e79268.
Vartanian, L. R., Spanos, S., Herman, C. P., & Polivy, J. (2015). Modeling of food intake: a meta-analytic review. Social Influence, 10(3), 119–136.
Weber, R., & Fuller, R. (2013). Statistical Methods for Communication Researchers and
