What Do People Like to “Share” About Obesity? A Content Analysis of Frequent Retweets About Obesity on Twitter

Abstract

Twitter has been recognized as a useful channel for the sharing and dissemination of health information owing in part to its “retweet” function. This study reports findings from a content analysis of frequently retweeted obesity-related tweets to identify the prevalent beliefs and attitudes about obesity on Twitter, as well as key message features that prompt retweeting behavior conducive to maximizing the reach of health messages on Twitter. The findings show that tweets that are emotionally evocative, humorous, and concern individual-level causes for obesity were more frequently retweeted than their counterparts. Specifically, tweets that evoke amusement were retweeted most frequently, followed by tweets evoking contentment, surprise, and anger. In regards to humor, derogatory jokes were more frequently retweeted than non-derogatory ones and, in terms of specific types of humor, weight-related puns, repartee, and parody were shared frequently. Consistent with extant literature about obesity, the findings demonstrated the predominance of the individual-level (e.g., problematic diet, lack of exercise) over social-level causes for obesity (e.g., availability of cheap and unhealthy food). Implications for designing social-media-based health campaign messages are discussed.
Keywords: Obesity, Twitter, retweets, social sharing, emotion, humor, attributional causes

Full text

OBESITY RETWEETS 1
Running Head: OBESITY RETWEETS
What do people like to “share” about obesity?
A content analysis of frequent retweets about obesity on Twitter
CITATION:
So, J., Prestin, A., Lee, L., Wang, Y., Yen, J., & Chou, W. S. (2016). What do people like to
“share” about obesity? A content analysis of frequent retweets about obesity on Twitter. Health
Communication, 31, 193-206. DOI: 10.1080/10410236.2014.940675

OBESITY RETWEETS 2
Abstract
Twitter has been recognized as a useful channel for the sharing and dissemination of health
information owing in part to its “retweet” function. This study reports findings from a content
analysis of frequently retweeted obesity-related tweets to identify the prevalent beliefs and
attitudes about obesity on Twitter, as well as key message features that prompt retweeting
behavior conducive to maximizing the reach of health messages on Twitter. The findings show
that tweets that are emotionally evocative, humorous, and concern individual-level causes for
obesity were more frequently retweeted than their counterparts. Specifically, tweets that evoke
amusement were retweeted most frequently, followed by tweets evoking contentment, surprise,
and anger. In regards to humor, derogatory jokes were more frequently retweeted than non-
derogatory ones and, in terms of specific types of humor, weight-related puns, repartee, and
parody were shared frequently. Consistent with extant literature about obesity, the findings
demonstrated the predominance of the individual-level (e.g., problematic diet, lack of exercise)
over social-level causes for obesity (e.g., availability of cheap and unhealthy food). Implications
for designing social-media-based health campaign messages are discussed.
Keywords: Obesity, Twitter, retweets, social sharing, emotion, humor, attributional causes

OBESITY RETWEETS 3
What do people like to “share” about obesity?
A content analysis of frequent retweets about obesity on Twitter
“Based on U.S. obesity rates, soon candidates will just walk for president.” Although
this frequently retweeted message may have been intended to be a joke, it nonetheless reflects
the escalating obesity rate. In fact, obesity has become a national epidemic with 34.9 % of adults
being obese (Ogden, Carroll, Kit, & Flegal, 2014). This epidemic must be addressed, as obesity
is a major contributor to numerous leading causes of death in the U.S., including heart disease,
diabetes, and several types of cancer (Flegal, Carroll, Ogden, & Johnson, 2002). Obesity is also
linked to depression and psychological stress (Strauss & Pollack, 2003), in part caused by
prevalent social stigmatization (Puhl & Heuer, 2009).
A considerable body of behavioral science has provided a foundation to understand the
public’s attitudes and beliefs about obesity, which are crucial components in the development of
effective public health interventions. Recently, another fruitful way to learn about public
attitudes towards a health issue has emerged: to unobtrusively examine them through analyses of
social media content (Chou, Prestin, & Kunath, 2014; Neiger et al., 2012). Scholars have begun
to appreciate the utility of social media as an ongoing record of public sentiment about health
and other issues (Salathé & Khandelwal, 2011) and extend their investigation into social media
interactions. For example, Scanfeld and colleagues (2010) performed a content analysis on
tweets about antibiotics and found that the most commonly held misunderstanding concerned the
use of antibiotics as a treatment for viral infections, followed by the belief that sharing leftover
antibiotics with others is acceptable. These types of social media analyses can enhance our
understanding on the public’s genuine attitudes and beliefs about obesity through an unobtrusive
method.

OBESITY RETWEETS 4
The ease of sharing information in social media makes these platforms a potentially
valuable context for research examining the public’s attitudes and beliefs about health issues. In
particular, Twitter has gained scholarly interest due to its unique features including the “retweet”
function, which allows the users to conveniently share tweets with others. Examinations of
widely shared retweets about obesity would allow us to identify the types of obesity-related
messages that people frequently endorse and publicly share, thus shedding light on the attitudes
or beliefs that are widely accepted by the public. Moreover, investigating the obesity-related
messages that prompted social sharing on Twitter has an important implication for designing
health campaign messages that aim at maximizing their reach via Twitter and other social media
outlets. In fact, the latter point gains greater importance when considering the fact that one of
the biggest challenges to the effectiveness of public health campaigns is their limited reach
(Hornik, 2002).
To this end, this research offers a content analysis of obesity-related messages on Twitter
that were frequently shared or “retweeted.” The specific aims of this study are twofold: First,
this paper provides an understanding of frequently shared attitudes and beliefs about obesity
expressed in a naturalistic online setting. Second, by examining the common characteristics or
elements of Twitter messages (or “tweets”) that generate engagement and prompt sharing (or
“retweets”), this research offers insight into designing social media-based health campaign
messages that can maximize their reach to the public.
Twitter possesses a number of distinctive features that allow it to be well-suited for these
inquiries. First, unlike other social media outlets such as Facebook, Twitter users do not need to
post personal information about themselves to find ‘friends’ and make connections with others
(Hughes, Rowe, Batey, & Lee, 2012). This feature offers a potential for anonymity, which may

OBESITY RETWEETS 5
encourage twitter users to be relatively less concerned about the social desirability of their posts
and more honest about what they have to say (Huberman, Romero, & Wu, 2009). Thus, an
unobtrusive analysis of tweets may present a relatively more truthful account of the public’s
attitudes and beliefs about obesity compared to surveys or other self-reported data. Second,
Twitter tends to focus on the sharing of opinions and information rather than on reciprocal social
interaction as Facebook does (Hughes et al., 2012). Thus, it is better suited for instantaneous
sharing among loosely connected individuals (Kwak, Lee, Park, & Moon, 2010), making it an
effective source of news and information (Lee & Oh, 2013) as well as a potential tool for health
promotion and campaigns (Lee & Sundar, 2012). Given the potential utility of Twitter in health
information dissemination, understanding the characteristics of tweets that are frequently shared
will be informative in designing Twitter-based health communication messages.
Thus, with an eye toward understanding features of widely shared tweets about obesity,
the overarching questions that guided this study are as follows: What do people endorse and
share widely on Twitter when it comes to obesity-related topics? What are the prominent
message features of frequently retweeted messages about obesity? What are the implications of
these findings on health campaigns utilizing Twitter?
Retweets and Dissemination of Health Information
Twitter is a popular social media outlet that has been recognized as a promising
communication channel for disseminating health information. It is a microblogging platform
where users can post messages (i.e., “tweets”) within the 140-character limit. Created in 2006,
Twitter has grown in popularity to 200 million active users as of March, 2013, who produce
more than 400 million tweets each day (Twitter, 2013). Twitter allows users to post tweets to
their own profile pages, subscribe to other users’ tweets by “following” them, and share

OBESITY RETWEETS 6
messages from other users with their own followers by “retweeting” them. The retweet function
makes content sharing easy and efficient, and also gives Twitter the potential to magnify the
reach of health-related messages. In fact, Twitter users are already producing and sharing health
information. For example, Paul and Dredze (2011) found that, out of 2 billion messages posted
in a 17-month period, 1.5 million tweets were health-related messages. Reflecting such
prevalence, 18% of Americans reported that they rely on Twitter to receive health-related
information (National Research Corporation, 2011). Recognizing the potential of Twitter as a
convenient and cost-effective way to reach large audiences (Neiger et al., 2012), health
organizations such as the National Institutes of Health (NIH), the Centers for Disease Control
and Prevention (CDC), and the World Health Organization (WHO) utilize Twitter as a
dissemination tool for health information (Park, Rogers, & Stemmle, 2013).
Taken together, health organizations and health communication researchers alike
recognize the utility of Twitter to engage in health communication. Now the question is how to
utilize Twitter to its fullest capacity and connect health communication efforts to the widest
range of people. As an initial step in understanding the diffusion of health information on
Twitter, we discuss factors that may motivate people to retweet messages and how the extant
literature can help us predict what type of obesity-related tweets might reach a wider audience.
Why Do We Retweet?
Retweeting is a key mechanism for information diffusion that allows tweets to reach a
new set of audiences beyond their initial reach (boyd, Golder, & Lotan, 2010). As such, this
feature has spurred research in a number of areas, including reasons for retweeting. According to
boyd and her colleagues (2010), for example, some of the major motivations for retweeting
include the desire to entertain or inform followers as an act of curation, to publicly agree with or

OBESITY RETWEETS 7
validate someone, and to comment on a tweet by retweeting with new information added.
Narrowing down the investigation to specific message features that motivate retweeting, Naveed
and his colleagues (2011) found that frequently retweeted messages were more likely to concern
broader public interests such as the economy and public events rather than more narrow topics or
personal tweets. This pattern was replicated in a content analysis of retweets about H1N1 flu
(Chew & Eysenbach, 2010). It can be inferred from these studies that enthusiasm for utilizing
Twitter as an effective communication channel for disseminating information about public health
concerns is well grounded. Despite the growing interest in retweeting behavior, however, extant
research aimed at understanding motivations for retweeting and features of retweeted content is
still limited. In addition, a vast majority of extant studies on this topic are not grounded in theory
and are exploratory in nature. Thus, we turn to a larger body of research that delves into a more
fundamental psychology of social sharing to shed more light on this “social sharing”
phenomenon on Twitter.
Emotion and Social Sharing
A useful theoretical framework that can help us understand the fundamental psychology
behind retweeting behavior is the social sharing of emotions (Rimé, 1995). According to Rimé
(1995), emotion is an important motivator of social sharing. A considerable body of literature on
the social sharing of emotions has widely documented the instinctive need people have to
disclose to others when they experience emotionally-charged events (e.g., Christophe & Rimé,
1997). Generally, the more intense the emotional experience, the more likely it is to be socially
shared (Rimé, Mesquita, Philippot, & Boca, 1991). For example, a review of eight studies, in
which participants were asked to recall a recent experience that evoked a specified emotion (e.g.,
joy, fear, or sadness) and then describe the extent to which they shared this experience (e.g.,

OBESITY RETWEETS 8
when and how often), found that over 90% of the emotional episodes were shared on average
(Rimé, Philippot, Boca, & Mesquita, 1992). As predicted, the extent of sharing (i.e., number of
repetitions and recipients) was positively correlated with the intensity of the emotional arousal.
There is also evidence that emotionally evocative events shared with one set of
individuals can be further shared by those individuals, a process called secondary social sharing.
Evidence of this phenomenon emerged from two studies by Christophe and Rimé (1997) in
which participants were asked to reflect on a time when someone had shared an emotional
experience with them and indicate whether they had shared the story they heard with anyone.
These shared episodes were able to evoke emotional responses in the listeners, and, on average
across two studies, over 70% of the stories were secondarily shared. Again, the intensity of the
emotional response was an important predictor of secondary social sharing. Secondary social
sharing of emotions is a particularly relevant phenomenon for the present study, as it is very
similar with retweeting behaviors in that the individuals are motivated to share the information
they initially received from someone else. Taken together, if we can consider retweeting
behavior as a form of social sharing, then research on social sharing of emotions suggests that
emotionally arousing tweets will provoke social sharing, specifically retweeting behavior.
In line with research on social sharing of emotions, research on viral marketing (Dobele
et al., 2007; Lindgreen &Vanhamme, 2005) and online social sharing (Shamma, Yew, Kennedy,
& Churchill, 2011) also positions emotion as an important ingredient in the success of online ads
that have gone “viral,” or gained heightened prominence and viewership from social
transmission. Indeed, empirical research shows that online content that sparks strong emotional
responses is more likely to be passed along to others. For instance, Bardzell and colleagues
(2009) found that greater emotional arousal measured by elevated heart rate was a significant

OBESITY RETWEETS 9
predictor of greater intention to share Internet videos with others. Drawing from the foregoing
discussion on the centrality of emotion in social sharing, the following hypothesis is advanced:
H1: Tweets drawing emotional responses will be more frequently retweeted than
non-emotional messages.
Given the influence of emotionality on the likelihood of sharing social media content,
scholars have begun to investigate the types of emotion that would have the greatest potential to
motivate sharing. In terms of valence of emotions, the literature is mixed: Some studies found
that messages provoking positive emotions are more widely shared (e.g., Eckler & Bolls, 2011),
while other studies found evidence substantiating the power of negative emotions (e.g., Heath,
Bell, & Sternberg, 2001). Given these mixed findings, Berger and Milkman (2012) investigated
the role of extent of arousal (or intensity) as opposed to valence of emotions in social
transmission as suggested by the literature on social sharing of emotion (Rimé et al., 1991).
Specifically, they proposed that high-arousal emotions (e.g., awe, surprise, and anger) would
increase the likelihood of social sharing compared to low-arousal emotions (e.g., sadness)
regardless of valence. As predicted, they found that online New York Times articles evoking
high-arousal emotions were more frequently forwarded to others via email than articles evoking
low-arousal emotions.
Though the literature investigating the influence of specific types of emotions on social
sharing is growing, it is not yet mature enough to allow us to make specific predictions about the
role of discrete emotions in social sharing. However, identification of discrete emotions that
likely induce retweeting would be informative in guiding the design of campaign messages
utilizing emotional appeals. Thus, we propose this follow-up research question:
RQ1: Which discrete emotions are conveyed frequently in retweets about obesity?

OBESITY RETWEETS 10
Humor
Humor is one of the most widely recognized message features of online content that
facilitates social transmission and sharing (Masland, 2001). For example, Phelps and his
colleagues (2004) found that almost half of all pass-along emails participants received were
jokes. Similarly, Dobele and her colleagues (2007) found that humor and amusement were key
mechanisms for the success of online viral marketing campaign messages. This argument was
substantiated in an experiment, which demonstrated that humorous advertisements that are also
high in violence were significantly more likely to be forwarded to others (Brown, Bhadury, &
Pope, 2010). Moreover, a content analysis comparing traditional media advertisements and
Internet advertisement that has “gone viral” gave humor the position of a universal appeal that
was used almost unanimously in viral advertisements (Porter & Golan, 2006). Given the
emphasis numerous scholars have placed on the role of humor in facilitating social transmission
of online content, it is not surprising that empirical research shows amusement or exhilaration, a
typical emotional reaction to humorous stimuli (Ruch, 1993), is one of the most frequently
experienced emotions individuals expressed when consuming viral online content.
Popularity of humorous content is also evident in social media research. For instance,
Holton and Lewis (2011) content analyzed tweets generated by the 430 most-followed journalists
active in Twitter and found that humorous tweets were significantly more retweeted. More
relevant to the current study, Yoo and Kim (2012) found that about 20% of the sampled YouTube
videos about obesity portrayed an obese individual as an object of humor (i.e., weight-based
teasing theme). More interestingly, they found that these videos with weight-based humor were
viewed more than 6 times more frequently than videos without weight-based teasing. Thus, we

OBESITY RETWEETS 11
predict that tweets about obesity that contain humorous elements will be more frequently
retweeted and spread more widely than non-humorous tweets.
H2: Twitter messages containing humor will be more frequently retweeted than
those without humor
Though much research has considered humor as a key ingredient in motivating sharing behavior,
little is known about the specific types of humor that individuals like to share online. There is
some evidence regarding the thematic types of humor that are frequently shared. For instance, of
the humorous pass-along emails participants received in a month, over 20% were general (or
topic-diffuse) jokes, followed by sexual humor (14.5%), gender issues, and work or computer-
related jokes (Phelps et al., 2004). In this study, we utilized a theory-grounded humor typology
developed by Buijzen and Valkenburg (2004) to examine the types of humor that are frequently
retweeted. Thus, we ask:
RQ2: What types of humor are used frequently in retweets about obesity?
Attributions of Causal Claim in Obesity
Obesity is a complex health condition caused by behavioral, genetic, environmental, and
psychosocial factors (Agurs-Collins & Bouchard, 2008). Despite the multifaceted nature of the
causes of obesity, the society has placed a much greater emphasis on individual factors such as
excessive food intake, lack of physical exercise, and, more recently, genetics, than on societal
factors such as the marketing of low-cost unhealthy food (Kim & Willis, 2007). Consistent with
this pervasive attitude, a survey of a nationally representative sample of U.S. adults showed that
respondents viewed obesity as a consequence of individual-level factors more so than of
environmental factors (Oliver & Lee, 2005). Attribution of cause for obesity is an important
issue to address as it accompanies behavioral implications that likely influence the prevalence of

OBESITY RETWEETS 12
obesity. For instance, publics must recognize the existence of systematic, societal-level
influence on obesity in order for policy changes to reduce obesity rates to take place (Blendon,
Hunt, & Benson, 2006). Similarly, if individuals believe that obesity is primarily caused by
genetic factors, which are out of one’s control, they will not engage in behavioral changes that
are conducive to maintaining a healthy weight (Wang & Coups, 2010).
The unbalanced emphasis on individual factors also prevails in the mass media.
Numerous content analytic studies of news articles have documented the tilted perspective (e.g.,
Boero, 2007; Jeong & Hwang, 2007). For example, Kim and Willis (2007) found that, for a 10-
yearperiod from 1995 and 2004, six major national and local newspapers mentioned individual-
level causes for obesity significantly more frequently than societal-level causes. More recently,
Yoo and Kim (2012) extended this line of research to the social media context and found that,
similar to traditional mass media, a significantly greater volume of YouTube videos endorsed
individual-level causes than systematic or societal-level causes for obesity.
Thus, it is speculated that a pattern similar to Yoo and Kim’s (2012) study will be
observed in obesity-related tweets as well. Since this study focuses on frequently retweeted
tweets as opposed to original tweets, however, a slightly different approach using the cognitive
dissonance theory (Festinger, 1957) is employed to make a prediction. In essence, the cognitive
dissonance theory posits that we experience psychological discomfort when we are exposed to
beliefs or attitudes that are inconsistent with our own. The theory further predicts that, in an
effort to reduce this discomfort, we try to selectively expose ourselves to information that helps
us avoid or resolve the cognitive conflict. For instance, Knobloch-Westerwick and Meng (2009)
found that subjects in an experiment spent significantly more time reading attitude-consistent
news. Drawing from this line of research, we anticipate that Twitter users will be more

OBESITY RETWEETS 13
motivated to expose themselves to, endorse, and retweet messages that are congruent with their
preexisting beliefs. Given that individual-level factors are much more widely recognized as the
primary cause of obesity, we predict that tweets endorsing individual-level causes will be
retweeted more frequently than those concerning societal-level causes. A follow-up research
question examining the specific types of causes that are frequently retweeted is advanced as well.
H3: Tweets endorsing individual-level causes for obesity will be more frequently
retweeted than those endorsing societal-level causes for obesity.
RQ3: What are the types of specific causes that are most frequently retweeted?
Method
Data Collection and Sample
Twitter data were collected as a part of a larger-scale research project concerning obesity-
related contents in social media in general (see Chou et al., 2014 for broad-level findings from
the larger corpus). Social media data were initially extracted through a commercially available
web-crawling search service, which offers a data-monitoring product to help marketing
companies track and analyze social media conversations about their brands. A profile, a
collection of pre-determined keywords used for data mining, was established, including the
following four search terms: obese, obesity, overweight, and fat. Data were downloaded from
the server at 12-hour intervals in a two-month period between January 23, 2012 and March 23,
2012, each time extracting the first 20,000 pieces of data. Approximately 200,000 posts
containing at least one of the keywords were collected on a given day. Once the initial data were
collected, a machine-learning, decision-tree classifier based on the human-coded training data
was constructed and used to automatically exclude irrelevant posts.1 Of the data collected across
different social media platforms, 1.25 million pieces (about 91%) were from Twitter. With the

OBESITY RETWEETS 14
cleaned dataset, an additional procedure took place in order to select the most frequently
retweeted tweets in all four keywords. Because Twitter users often make minor modifications to
the tweets before retweeting, an algorithm that utilizes a similarity metric was designed.
Specifically, the designed algorithm considered tweets with 65% or greater similarity to be
retweets of the original message and counted the frequency at which the message was shared.
The algorithm generated a rank-ordered list of tweets with the frequency at which each message
was retweeted.
Purposive sampling was used to select the most frequently retweeted tweets.
Specifically, as this study aimed at examining the common features of frequent retweets, the 30
most retweeted tweets for each keyword were sampled, resulting in a total of 120 tweets. The
decision to examine the 30 most frequent retweets was based on the fact that the retweet
frequency dropped significantly after the 30th most retweeted messages for most keywords.2 In
sum, the total retweet frequency of the sampled 120 tweets was accumulated to be 121,268,
which represents approximately 12.13 % of all twitter data collected.
Coding Procedure
Each tweet served as the unit of analysis for coding.3 There were a total of seven coding
tasks that needed to be performed to test the hypotheses and answer research questions advanced
in this study. For each task, two independent coders discussed the coding rules and agreed on the
conceptual and operational differences among the coding categories before initiating the tasks.
After establishing coding rules and specifying them in a codebook, the two coders independently
coded approximately 20% of the tweets and reconvened to examine whether the coding rules
were adequate and sufficient to code the complete sample of tweets. Instances that were difficult
or ambiguous to code with the previously agreed-upon coding rules were discussed. After the

OBESITY RETWEETS 15
discussion, the codebook was modified by the inclusion of additional rules and the further
specification of existing rules. Then, the two coders independently coded the entire sample of
tweets using the modified codebook.
Upon the completion of coding, the two coders reconvened and intercoder reliability was
computed. This study utilized Krippendorff’s alpha as a reliability coefficient. As suggested by
Krippendorff (2004), a reliability coefficient greater than or equal to .80 was considered ideal
and a coefficient between .667 and .80 was deemed acceptable for drawing tentative conclusions.
Disagreement was resolved through discussion between the two coders.
Content Analysis Variables
General theme. The lead author studied the entire sample of tweets to identify
prevalent themes conveyed in the tweets. A total of 8 major thematic categories emerged from
the preliminary examination: derogatory jokes, non-derogatory jokes, advocating societal
change, causal factors for obesity, factoids, sarcastic comments, obesity prevalence, and non-
scientific weight loss tips. Prior to coding, the coders discussed the operational definitions of
each theme category. For instance, utilizing the definitions offered by Merriam-Webster
dictionary (2014), “derogatory jokes” were operationalized as messages containing humor
“expressing a low opinion of someone or something; showing a lack of respect for someone or
something” (e.g., Just saw a fat ginger girl buying a rape whistle... gotta admire her optimism)
and “factoids” were operationalized as “a briefly stated and usually trivial fact” related to
obesity (e.g., Researchers in Australia have completed a study that shows weight loss may
result in better sex for overweight diabetic). When tweets contained information about factors
and/or behaviors that contribute to obesity, they were categorized as “causal factors for
obesity”(e.g., One soda a day increases kid’s risk of being obese by 60%). Tweets containing

OBESITY RETWEETS 16
critical comments about obesity issues with arguments for a change or action in the society
were categorized as “advocating societal change” (e.g., Obesity figures are troubling: new
trends need to be countered in some way). All categories were mutually exclusive to allow for
statistical inference (Krippendorff’s α = .87; see Table 1 for other examples of each category).
When a given tweet did not fall into one of the 8 themes in the codebook, the coders were
instructed to mark them as miscellaneous and were asked to identify possible coherent themes
among these cases. Of the 10 Tweets that were classified as miscellaneous, 6 were categorized
as forming a coherent category of “personal experience or anecdote” after discussion (e.g., I hate
when someone skinnier than you says ‘I'm fat’ and it makes you feel obese).
Emotion. Since H1 was based on research on the social sharing of emotion (Rimé,
1995), which assumes that the emotional state of the sharer motivates social sharing, the emotion
coding focused on the coders’ interpretation of what the message sharer likely felt when
retweeting. First, the coders identified whether or not each tweet was deemed to have been
retweeted due to an emotional response to reading the tweet (Krippendorff’s α = .84). The
tweets that are coded as being retweeted due to an emotional response were again coded for
discrete emotion. In other words, the coders put themselves in the perspective of the “social
sharer” and identified the emotion that would have motivated social sharing. The list of emotion
words offered by Shaver, Schwartz, Kirson and O’Connor (1987) was utilized to identify the
emotion. Consistent with the previous content analysis studies (e.g., Freimuth, Hammond,
Edgar, &Monahan, 1990), the tweets were coded for one predominant emotion (Krippendorff’s α
= .77). For example, when the tweets were thought to generate the feeling of being pleased after
encountering a statement that one strongly agrees with or approves, the coders coded them as
evoking “contentment,” a low-arousal, positive emotion (Fredrickson, 1998) evoked by

OBESITY RETWEETS 17
satisfaction experienced after the fulfillment of one’s need (Berenbaum, 2002). As
“disappointment” refers to an emotion experienced by non-fulfilment of desired expectation
(Frijda, 1986), a tweet was coded as reflecting “disappointment” when the expectation of what
the society should be primarily concerned about (e.g., “Syrian massacre” as opposed to “pet
obesity on the rise”) was violated or unmet. Using these operational definitions, the following 9
discrete emotions were identified: amusement, contentment, surprise, anger, sympathy,
disappointment, sadness, worry, and hope (see Table 2).
Humor. The coders determined whether each tweet contained humorous content and
could therefore be classified as jokes (Krippendorff’s α = .80). Once the humorous tweets were
identified, the coders classified them into different types of humor utilizing the categorization
scheme offered by Buijzen and Valkenburg (2004). As this humor taxonomy was originally
specific to audio-visual humor content, only the types that could be applied to textual humor
were utilized as coding categories in this study (Krippendorff’s α = .80). Of the original 27
humor types from Buijzen and Valkenburg’s (2004) taxonomy that are relevant to textual humor,
a total of 16 humor types were identified (see Table 3).
Attribution of causes of obesity. The coders determined whether or not each tweet
contained information about causes of obesity. Tweets that did not explicitly state but rather
implied causal factors of obesity were also categorized as conveying such information (e.g.,
“You know why America is obese? Because the only running they do is on temple run”;
Krippendorff’s α = .78). Coders then engaged in a secondary level of coding this data for
specific causal factors related to obesity. Drawing from extant research on obesity, tweets
implying a cause for obesity were specified into one of the four categories: societal-level cause,
(individual-level) dietary cause, (individual-level) exercise cause, and (individual-level) genetic

OBESITY RETWEETS 18
cause (Krippendorff’s α = .90). Following the coding method employed for thematic coding, if
the tweet did not fall into one of the four categories, the coders were instructed to mark it as
miscellaneous and asked to construct a cohesive category among these miscellaneous cases, if
possible. There were two tweets that were marked as miscellaneous: One concerned a
personality factor (“People who remain calm in stressful situations have higher rates of
depression and obesity, a study finds”) and the other concerned lack of sleep as a cause (“Lack of
sleep increases your risk for heart disease, diabetes and obesity”). Since they did not form a
cohesive category together, they remained in the category of their own (see Table 4).
Results
Overview of the Data: Two Possible Approaches to Data Analysis
Since the data contained both the rank-ordered, 30 most frequently retweeted tweets for
each of the four keywords, as well as the actual frequency with which each tweet was retweeted,
there were two ways to analyze the data. First, quite simply, each of the 120 tweets in the data
without reflecting the actual frequency of sharing could serve as a unit of analysis. In this type
of analysis, each sampled tweet is given the same weight regardless of whether it was ranked as
the 1stor the 30th most frequently shared tweet. This approach is useful in gaining understanding
of the general landscape of the tweets sampled in this study. The second approach involves an
analysis of weighted data, in which the actual number of retweets, or retweeted frequency of
each tweet, is reflected in the analysis. As this study was primarily motivated by the question
“What causes an obesity-related tweet to be retweeted?” the second approach that takes retweet
frequency into account aligns more closely with the research questions and hypotheses advanced
in this paper. Thus, the primary analysis testing the hypotheses was conducted using the

OBESITY RETWEETS 19
weighted dataset. Nonetheless, the findings from the first analytic approach are also included
briefly for information purposes.4
Overall Thematic Analysis
Prior to discussing the findings related to the aforementioned hypotheses and research
questions, the prevalent themes of the frequently retweeted messages about obesity will be
presented to provide an overall idea of the thematic landscape of this content. The results of
descriptive analysis of the themes of each tweet indicate that the majority (55.8%) of the
collected tweets were jokes about obesity or weight (see Table 1). The jokes were further
categorized as either derogatory or non-derogatory in tone. A slightly greater number of jokes
were non-derogatory (30.8%) and the rest were derogatory (25%). Causal factors for obesity
was the third most frequently appearing theme among the top retweets (13.3%), followed by
advocating societal change (6.7%) and factoids (6.7%).
A slightly different pattern emerged when retweet frequency was taken into
consideration. Descriptive analysis of data weighted by retweet frequency revealed that
derogatory jokes were by far most frequently retweeted (49.7%), followed by non-derogatory
jokes (32.8%). These two joke categories together made up the vast majority of retweets
(82.5%). In sum, jokes were not only the prevalent theme among the 120 tweets but they were
also very frequently retweeted. Tweets concerning personal experience or anecdotes about
obesity (6.5%) were also retweeted quite frequently and were followed by tweets advocating
societal change (4.5%), tweets concerning causal factors for obesity (3.0%), and factoids (2.5%;
see Table 1).
Emotion

OBESITY RETWEETS 20
H1 predicted that messages drawing emotional responses will be retweeted more
frequently than non-emotional messages. A chi-square test of goodness-of-fit indicated that a
significantly greater number of tweets drew emotional responses (85%), χ²(1, N = 120) = 58.80,
p< .001. The finding was replicated with the analyses involving the data weighted by retweet
frequency: Messages drawing emotions (99%) were significantly more frequently retweeted than
those not drawing emotions (1%), χ²(1, N = 121,268) = 116,425.34, p< .001. Thus, H1 was
supported. In addition, a cross tabulation analysis showed that the predicted tendency was
maintained across the four obesity keywords (see Table 5).
RQ1 asked what the prevalent discrete emotions identified in the frequent retweets were.
Perhaps due to the fact that the majority of retweets were humor-based, amusement was
identified as the most prevalent discrete emotion (78.8%), followed by contentment (11.5%),
surprise (4.9%), and anger (3.7%; see Table 2).
Humor
H2 predicted that humorous tweets would be retweeted more frequently than the ones
without humorous content. The results of a chi-square test of goodness-of-fit showed that a
slightly greater number of tweets contained humor (n = 66) than not (n = 54) but the difference
was not statistically significant, χ²(1, N = 120) = 1.20, p = ns. When using weighted data,
however, humorous tweets (82.2%) were significantly more frequently retweeted than non-
humorous tweets (17.8%), χ²(1, N = 121,268) = 50,319.21, p< .001, as predicted by H2 (see
Table 5). Thus, H2 was supported. In addition, a cross tabulation analysis showed that the
prediction held for obese, obesity, and fat keywords but not for overweight. With the overweight
keyword, non-humorous tweets were retweeted more frequently (see Table 5).

OBESITY RETWEETS 21
RQ2 asked what types of humor were most frequently retweeted. Tweets with puns, or
humorous use of words that are alike or nearly alike in sound but different in meaning, were
most frequently retweeted (22.2%). Next, repartee (18.1%), or verbal banter usually in a witty
dialogue, was found to be frequently shared, followed by parody (12.1%), and sarcasm (9.1%).
As expected with the high volume of derogatory jokes, ridicule (8.9%) was another frequently
shared humor type, followed by outwitting (7.5%), outsmarting someone by a retort, response, or
punch line makes the former statement seem inferior. Absurdity (6.6%) and conceptual surprise
(6.4%), a humor tactic that misleads the reader by means of a sudden unexpected change of
concept, were also found (see Table 3).
Attribution of Causes of Obesity
Prior to examining the specific types of causes of obesity mentioned or implied in the
frequently retweeted tweets, we examined how many of the messages were, in fact, either
containing or implying any types of cause of obesity at all. The results show that exactly half of
the tweets (50%) either contained information about causes of obesity or implied it in a more
subtle way, χ²(1, N = 120) = 0.00, p = ns. The same pattern was observed with the weighted
data: 50.6% of retweets contained information about causes of obesity, χ²(1, N = 121,268) =
17.92, p < .001.5
H3 predicted that tweets emphasizing individual-level causes for obesity would be
retweeted more frequently than those implying societal-level causes for obesity. Of the 60
tweets that either mentioned or implied causal factors, a vast majority (91.7%) stated or implied
individual-level causes, and the rest concerned societal-level causes (8.3%), χ²(1, N = 60) =
41.67, p < .001. The same pattern was observed with the weighted data: Tweets concerning
individual-level causes were more frequently retweeted (86.4%) than those concerning societal-

OBESITY RETWEETS 22
level causes (13.6%), χ²(1, N = 61,371) = 32,602.73, p < .001. Thus, H3 was supported. A cross
tabulation analysis indicated that the predicted pattern was observed in data for the overweight,
obesity, and fat keywords but not in data for the obese keyword. Tweets containing obese
keyword were more frequently retweeted when they concerned societal level causes (75.8%)
than individual level causes (24.2%), which was a direct opposite to the patterns observed in the
overall data and data for the other three keywords (see Table 5).
RQ3 asked what types of specific causes were most frequently retweeted. Of the
retweets concerning individual-level causes, problematic diet was the most frequently retweeted
cause (60.8%), followed by lack of exercise (22%), personality issues (3.2%), genetic
predisposition to obesity (0.3%), and sleep deprivation (0.2%). Of the retweets concerning
societal-level causes, the availability of cheap and unhealthy food was the most prevalently
retweeted cause (11.94%), followed by the current school system (1.05%), portion sized offered
at restaurants (0.48%), and a broken food system (0.08%; see Table 4).
Discussion
This study presents findings from a content analysis of frequently retweeted obesity-
related messages on Twitter, a popular social media outlet often utilized for sharing health-
related information. As predicted, the results of the content analysis indicate that obesity-related
tweets that are emotionally evocative, humorous, and concern individual-level cause for obesity
were more frequently retweeted and shared than their counterparts.
Consistent with research on the social sharing of emotions and viral online content that
emphasize the role of emotion in social transmission, emotionally arousing tweets were much
more likely to be retweeted than non-emotional ones. Specifically, a vast majority of tweets
were found to draw positive emotions such as amusement (78.8%) and contentment (11.5%).

OBESITY RETWEETS 23
This finding resonates with research on viral online media content, which demonstrates that
people are more inclined to share media content that is expected to evoke positive emotions in
the recipients. From a self-presentation perspective, Berger and Milkman (2012) explain that
this is because people prefer to share messages that will generate positive emotion in the
receivers rather than upset them, thus potentially reflecting positively on themselves.
In contrast, in health communication campaigns, the most widely utilized emotional
appeal involves fear and its family emotions such as worry and anxiety (Freimuth et al., 1990;
Job, 1988). However, only 0.1% of the frequent retweets about obesity reflected such emotions
in this study. The stark contrast between this finding and the type of emotional appeal health
campaigns utilize the most presents an important implication for designing social-media based
obesity-related messages: If the message focuses on evoking negative emotions such as fear and
anxiety as in a traditional health campaign context, it may not be shared as widely as those
utilizing other positive emotional appeals, thus possibly defeating the purpose of utilizing such
social networking sites. In other words, if social media were to be utilized to maximize its reach
to public, the message may need to be different from traditional health campaign messages that
very often employ negative emotional appeals (e.g., fear, guilt).
The third most pervasive emotion in the analysis was surprise. This finding lines up well
with the finding that tweets containing health information, interesting facts, and statistical
information that likely have surprise or novelty value were retweeted quite frequently. This
finding also resonates with extant research on risk perception that suggests novelty value as an
arousing aspect of perceived risk (Fischhoff, Slovic, Lichtenstein, Read, & Combs, 1978). Thus,
surprise may be another viable candidate for emotional appeals when using social media to
transmit health-related information.

OBESITY RETWEETS 24
Although on the whole, tweets that evoked negative emotions were less likely to be
shared, anger (3.7 %) was found to be shared quite frequently. Tweets that potentially made
users angry were far more likely to be retweeted than those that concerned other negative
emotions such as disappointment, sadness, and worry (all falling below 1.0%). A closer look
into the data reveals that a vast majority of the tweets that seemed to have been shared due to
anger concerned issues related to the widely-spread stigmatization against overweight and obese
individuals. Thus, the observed prevalence of anger may be specific to health issues that are
linked to social stigmatization. It also suggests that publics are, to some degree, not only aware
of the social injustice issues associated with weight-related stigmatization but also motivated to
share their “anger” with others on social media.
As can be predicted from the finding that highlights the dominant role of amusement in
motivating retweeting behaviors, humorous tweets were significantly more likely to be retweeted
than non-humorous ones. In terms of tone, derogatory jokes about obesity were more likely to
be retweeted (60.64%) than non-derogatory ones (39.54%; % within humorous tweets). This
finding is similar to what Yoo and Kim (2012) found in their content analysis of obesity-related
YouTube videos: It provides yet another piece of empirical evidence demonstrating the
omnipresence of weight stigma in social media. Moreover, the fact that these derogatory jokes
were widely shared on Twitter shows that a great number of Twitter users do not see a problem in
publicly discriminating against overweight people. However, it should also be noted that
roughly half of the humorous tweets about obesity studied here were not derogatory in tone.
Investigation into the different types of humor appeals shows that Twitter users found a
variety of humor appeals, including those that could be used in a health communication context,
to be amusing enough to share with others. Specifically, tweets using “puns,” humorous use of

OBESITY RETWEETS 25
words that sound alike but have different meanings, were retweeted most frequently, closely
followed by “repartee,” a verbal banter in a witty dialogue. As these types of humor can be more
easily delivered in brief statements compared to other types of humor (e.g., conceptual surprise),
they seem to be particularly well-suited for tweets. For obvious reasons, some of the humor
techniques identified in this study that are inherently derogating in tone such as ridicule,
stereotype, irreverent behavior, and malicious pleasure, cannot and should not be considered in
public health campaigns aiming at reducing obesity. This is particularly important to note since
some health communication practitioners and designers of anti-obesity PSAs seem to hold an
unfounded assumption that provoking shame in overweight and obese individuals will motivate
them to manage their weight (Puhl & Brownell, 2003).
One implication of this finding is to incorporate humor in social-media based health
messages to maximize exposure. The use of humor in a public health communication context is
not new. For instance, owing to humor’s potential to ameliorate denial and resistance towards
serious health issues, humor-based HIV/AIDS prevention messages have been shown to be quite
effective (Peterson, 1992). The use of humor is also prevalent in anti-smoking PSAs. In a
content analysis of televised anti-smoking PSAs, humor was one of the most prevalent appeals
used along with informational appeals (Cohen, Shumate, & Gold, 2007). Despite its popularity,
however, some studies report adverse effects of humor (e.g., Fishbein, Hall-Jamieson, Zimmer,
Haeften, & Nabi, 2002). Given the mixed findings, the utility of humor in health communication
needs to be examined further. In sum, though the use of some of the humor techniques observed
in this study (e.g., puns) is feasible and likely to generate wider dissemination than others,
caution should be practiced when using humor in communicating about obesity.

OBESITY RETWEETS 26
The final hypothesis concerned the prevalence of retweets concerning individual- versus
societal-level causes of obesity. As predicted, the majority of popular retweets concerned a
variety of individual-causes of obesity, including individual dietary practices and lack of physical
activity. This finding resembles the results of previous content analytic studies that showed
predominance of the individual frame in the traditional mass media. In other words, the data
shows that, as much as public health agencies and health communication researchers
acknowledge that the issue is complex and multi-level solutions are necessary, the public’s day-
to-day conversations on Twitter indicate predominant discussions about individual factors. As
obesity is a multi-factorial health issue that is also profoundly affected by societal- or
environmental-factors, efforts to promote a more balanced view of attribution are warranted.
Although it was not a central component of the study, this study also showed that the
predictions were held across the four different obesity keywords, with two exceptions: When
tweets contained the word “overweight,” non-humorous tweets were more frequently retweeted
than humorous ones. In addition, when tweets contained the word “obese,” tweets concerning
societal-level causes were significantly more frequently retweeted than those concerning
individual-level causes. We speculate that these exceptions were in part caused by technical and
scientific connotations attached to the words “overweight” and “obese.” However, due to the
paucity of research in this area, this finding awaits more scholarly attention from future research.
This study has several limitations. First, though this study involved an examination of a
large set of data, the data were collected over a two-months period. Moreover, based on the fact
that retweet frequency dropped significantly after the top 30 retweets (see Footnote 2 for details),
the 30 most frequently shared tweets for each of the four keywords were sampled in this study,
resulting in 120 pieces of tweets. Thus, the findings are, by no means, representative of the

OBESITY RETWEETS 27
obesity-related content on Twitter as a whole. In terms of coding, the method we used to code
for a predominant emotion that likely motivated social sharing warrants a note. Due to the
nature of the dataset that only allowed for a secondary analysis, we cannot ascertain that these
are the emotional responses that actually motivated the twitter users to retweet the messages.
However, it should be also noted that this type of interpretive coding is often employed in
secondary analyses (e.g., Berger & Milkman, 2012), particularly those studying the contents of
social media. Future survey-, diary-based, and qualitative studies employing interviews would
provide invaluable insight into understanding the actual emotional processes that take place
when retweeting and effectively complement the findings from the current study. Lastly,
although this study has identified some active ingredients in tweets that likely lead to social
sharing, it is important to note that the message itself is not the only factor influencing social
sharing. In particular, in the context of social media, whether or not you are connected to an
individual disseminating the health message is of utmost importance because it determines
whether or not you are exposed to the message at all. However, examination of the networking
aspect of Twitter is beyond the scope of the current study. Therefore, in order to fully understand
the social transmission of health-related information on Twitter, future research should also
investigate what influences one’s decision to be connected to other Twitter users, thus presenting
a more complete picture of the process of social transmission and potentially aiding public health
agencies to build a widespread social network in social media.

OBESITY RETWEETS 28
References
Agurs-Collins, T., & Bouchard, C. (2008). Gene-nutrition and gene-physical activity interactions
in the etiology of obesity. Obesity, 16, S2–S4. doi: 10.1038/oby.2008.510
Alhabash, S., McAlister, A. R., Hagerstrom, A., et al. (2013). Between likes and shares: Effects
of emotional appeal and virality of anti-cyberbullying messages on Facebook.
Cyberpsychology, Behavior, and Social Networking, 16, 175–182. doi:
10.1089/cyber.2012.0265
Bardzell, J., Bardzell, S., & Pace, T. (2009). Understanding affective interaction: Emotion,
engagement, and internet videos. In Proceedings of 2009 IEEE International Conference
on Affective Computing and Intelligent Interaction. doi: 10.1109/ACII.2009.5349551
Berenbaum, H. (2002). Varieties of joy-related pleasurable activities and feelings. Cognition &
Emotion, 16, 473–494.
Berger, J., & Milkman, K. L. (2012). What makes online content viral? Journal of Marketing
Research, 49, 192–205. doi: http://dx.doi.org/10.1509/jmr.10.0353
Blendon, R. J., Hunt, K., & Benson, J. M. (2006). Understanding the American public’s health
priorities: A 2006 perspective. Health Affairs, 25, w508–w515.
Boero, N. (2007). All the news that’s fat to print: The American “obesity epidemic” and the
media. Qualitative Sociology, 30, 41–60. doi: 10.1007/s11133-006-9010-4
boyd, d., Golder, S., & Lotan, G. (2010). Tweet, tweet, retweet: Conversational aspects of
retweeting on Twitter. In Proceedings of HICSS’10. New York, NY: ACM Press. doi:
10.1109/HICSS.2010.412
Brown, M. R., Bhadury, R. K., & Pope, N. K. (2010). The impact of comedic violence on viral
advertising effectiveness. Journal of Advertising, 39, 49-65. doi: 0.2753/JOA0091-

OBESITY RETWEETS 29
3367390104
Buijzen, M., & Valkenburg, P. (2004). Developing a typology of humor in audiovisual media.
Media Psychology, 6, 147–167. doi: 10.1207/s1532785xmep0602_2
Chew, C., & Eysenbach, G. (2010). Pandemics in the age of Twitter: Content analysis of Tweets
during the 2009 H1N1 outbreak. PLoS ONE, 11, 1-13. doi:
10.1371/journal.pone.0014118.
Chou, W. S., Prestin, A., & Kunath, S. (2014). Obesity in social media: A mixed methods
analysis. Translational Behavioral Medicine, 1-10.doi: 10.1007/s13142-014-0256-1
Cohen, E. L., Shumate, M. D., & Gold, A. (2007). Anti-smoking media campaign messages:
Theory and practice. Health Communication, 22, 91-102. doi:
10.1080/10410230701453884
Christophe, V., & Rimé, B. (1997). Exposure to the social sharing of emotion: Emotional impact,
listener response, and secondary social sharing. European Journal of Social Psychology,
27, 37–54. doi:10.1002/(SICI)1099-0992(199701)27:1.
Dobele, A., Lindgreen, A., Beverland, M., Vanhamme, J., & van Wijk, R. (2007). Why pass on
viral messages? Because they connect emotionally. Business Horizons, 50, 291-304. doi:
http://dx.doi.org/10.1016/j.bushor.2007.01.004
Eckler, P., & Bolls, P. (2011). Spreading the virus: Emotional tone of viral advertising and its
effect on forwarding intentions and attitudes. Journal of Interactive Advertising, 11, 1-11.
doi:10.1080/15252019.2011.10722180
Festinger, L. (1957). A theory of cognitive dissonance. Stanford, CA: Stanford University Press.
Fischhoff, B., Slovic, P., Lichtenstein, S., Read, S., & Combs, B. (1978). How safe is safe
enough? A psychometric study of attitudes towards technological risks and benefits.

OBESITY RETWEETS 30
Policy Sciences, 9, 127–152. doi: 10.1007/BF00143739
Fishbein, M., Hall-Jamieson, K., Zimmer, E., Haeften, I., & Nabi, R. (2002). Avoiding the
boomerang: Testing the relative effectiveness of antidrug public service announcements
before a national campaign. American Journal of Public Health, 92, 238–245. doi:
10.2105/AJPH.92.2.238
Flegal, K. M., Carroll, M. D., Ogden, C. L., & Johnson, C. L. (2002). Prevalence and trends in
obesity among US adults, 1999–2000. Journal of the American Medical Association,
288, 1723–1727. doi: 10.1001/jama.288.14.1723.
Fredrickson, B. L. (1998). What good are positive emotions? Review of General Psychology, 2,
300-319. doi:10.1037/1089-2680.2.3.300
Freimuth, V., Hammond, S. L., Edgar, T., & Monahan, J. L. (1990). Reaching those at risk: A
content-analytic study of AIDS PSAs. Communication Research, 17, 775-791. doi:
10.1177/009365029001700604
Frijda, N. H. (I986). The emotions. Cambridge, England: Cambridge University Press.
Heath, C., Bell, C., & Sternberg, E. (2001). Emotional selection in memes: The case of urban
legends. Journal of Personality and Social Psychology, 81, 1028-1041. doi:
10.1037/0022-3514.81.6.1028
Holton, A., & Lewis, S. (2011). Journalists, social media, and the use of humor on Twitter.
Electronic Journal of Communication, 21, 1-21.
Hornik, R. (2002). Public health communication: Making sense of contradictory evidence. In R.
Hornik (Ed.), Public health communication: Evidence for behavior change (3rd ed., pp.
1-19). Mahwah, MJ: Lawrence Earlbaum.
Huberman, B., Romero, D. M., & Wu, F. (2009). Social networks that matter: Twitter under the

OBESITY RETWEETS 31
microscope. First Monday, 14, 1-5. doi: arXiv:0812.1045v1
Hughes, D. J., Rowe, M., Batey, M., & Lee, A. (2012). A tale of two sites: Twitter vs. Facebook
and the personality predictors of social media usage. Computers in Human Behavior, 28,
561–569. doi: 10.1016/j.chb.2011.11.001
Jeong, S., & Hwang, Y. (2007, November). News about obesity: A content analysis of news
coverage of the issue of obesity from 1985 to 2005. Chicago: National Communication
Association.
Job, R. F. S. (1988). Effective and ineffective use of fear in health promotion campaigns.
American Journal of Public Health, 78, 163–167. doi: 10.2105/AJPH.78.2.163
Kim, S. H., & Willis, L. A. (2007). Talking about obesity: News framing of who is responsible
for causing and fixing the problem. Journal of Health Communication, 12, 359–376. doi:
10.1080/10810730701326051.
Knobloch-Westerwick, S., & Meng, J. (2009). Looking the other way: Selective exposure to
attitude-consistent and counterattitudinal political information. Communication Research,
36, 426-448. doi: 10.1177/0093650209333030
Krippendorff, K. (2004). Reliability in content analysis: Some common misconceptions and
recommendations. Human Communication Research, 30, 411-433. doi: 10.1111/j.1468-
2958.2004.tb00738.x
Kwak, H., Lee, C., Park, H., & Moon, S. (2010). What is Twitter: A social network or a news
media. In Proceedings of the 19th International Conference on World Wide Web (pp. 591–
600). New York, NY: ACM. doi: 10.1145/1772690.1772751
Lee, E., & Oh, S. Y. (2013). Seek and you shall find? How need for orientation moderates
knowledge gain from Twitter use. Journal of Communication, 63, 745-765. doi:

OBESITY RETWEETS 32
10.1111/jcom.12041
Lee, J. Y., & Sundar, S. S. (2012). To tweet or to retweet? That is the question for health
professional on Twitter. Health Communication, 28, 509–524.
dx.doi.org/10.1080/10410236.2012.700391.
Lindgreen, A., & Vanhamme, J. (2005). Viral marketing: The use of surprise. In I.C. Clarke & T.
B. Flaherty (Eds.), Advances in electronic marketing (pp. 122-138). Hershey, PA: Idea
Group.
Masland, E. (2001). Viral marketing word of mouth comes of age. Websolvers Inc. Retrieved
from http://www.websolvers.com/portfolio/papers/viral.asp
Merriam-Webster Online (2014). Definition of “derogatory.” Retrieved from
http://www.merriam-webster.com/dictionary/derogatory
National Research Corporation (2011). 1 in 5 Americans use social media for health care
information. Retrieved from http://hcmg.nationalresearch.com/public/News.aspx?ID=9
Naveed, N., Gottron, T., Kunegis, J., & Alhadi, A. C. (2011). Bad news travel fast: A content-
based analysis of interestingness on Twitter. In Proceedings of the 3rd International Web
Science Conference (pp. 1 -7). Koblenx, Germany. doi: 10.1145/2527031.2527052
Neiger, B., Thackeray, R., Van Wagenen, S., Hanson, C., West, J., Barnes, M., & Fagen, M.
(2012). Use of social media in health promotion: Purposes, key performance indicators,
and evaluation metrics. Health Promotion Practice, 13, 159-164. doi:
10.1177/1524839911433467
Ogden, C. L., Carroll, M. D., Kit, B. K., & Flegal, K. M. (2014). Prevalence of childhood and
adult obesity in the United States, 2011-2012. Journal of American Medical Association,
311, 806-814. doi:10.1001/jama.2014.732.

OBESITY RETWEETS 33
Oliver, J. E., & Lee, T. (2005). Public opinion and the politics of obesity in America. Journal of
Health Politics, Policy and Law, 30, 923–954. doi: 10.1215/03616878-30-5-923
Park, H., Rodgers, S., & Stemmle, J. (2013). Analyzing health organizations’ use of Twitter for
promoting health literacy. Journal of Health Communication, 18, 410-425.
doi:10.1080/10810730.2012.727956
Paul, M. J., & Dredze, M. (2011). You are what you tweet: Analyzing Twitter for public health.
In Proceedings of the Fifth International AAAI Conference on Weblogs and Social Media
(pp. 265-272). Barcelona, Spain.
Peterson, K. E. (1992). The use of humor in AIDS prevention, in the treatment of HIV-positive
persons, and in the remediation of caregiver burnout. In M. R. Seilson & K. E. Peterson
(Eds.), AIDS prevention and treatment: Hope, humor, and healing (pp. 37-57). New York:
Hemisphere.
Phelps, J. E., Lewis, R., Mobilio, L., Perry, D., & Raman, N. (2004). Viral marketing or
electronic word-of-mouth advertising: Examining consumer responses and motivations to
pass along email. Journal of Advertising Research, 44, 333-348. doi:
10.1017/S0021849904040371
Porter, L. & Golan, G. J. (2006). From subservient chickens to brawny men: A comparison of
viral advertising to television advertising. Journal of Interactive Advertising, 6, 4-33.
doi:10.1080/15252019.2006.10722116
Puhl, R. M., & Heuer, C. A. (2009). The stigma of obesity: A review and update. Obesity, 17,
941–964. doi: 10.1038/oby.2008.636
Puhl, R. M., & Brownell, K. D. (2003). Ways of coping with obesity stigma: Review and
conceptual analysis. Eating Behaviors, 4, 53–78. http://dx.doi.org/10.1016/S1471-

OBESITY RETWEETS 34
0153(02)00096-X
Rimé, B. (1995). The social sharing of emotion as a source for the social knowledge of emotion.
In J. A. Russell et al. (Eds.), Everyday conceptions of emotion: An introduction to the
psychology, anthropology, and linguistics of emotion. (pp. 475–489). Dordrecht,
Netherlands: Kluwer Academic Publishers.
Rimé, B., Mesquita, B., Philippot, P. & Boca, S. (1991). Beyond the emotional event: Six studies
on the social sharing of emotion. Cognition and Emotion, 5, 435–465. doi:
10.1080/02699939108411052.
Rimé, B., Philippot, P., Boca, S., & Mesquita, B. (1992). Long-lasting cognitive and social
consequences of emotion: Social sharing and rumination. European Review of Social
Psychology, 3, 225-258, doi: 10.1080/14792779243000078
Ruch, W. (1993). Exhilaration and humor. In M. Lewis & J. M. Haviland (Eds.), Handbook of
emotions (pp. 605-616). New York: Guilford Press.
Salathé, M., & Khandelwal, S. (2011). Assessing vaccination sentiments with online social
media: Implications for infectious disease dynamics and control. PLoS Computational
Biology, 7, 1–7. doi:10.1371/journal.pcbi.1002199
Scanfeld, D., Scanfeld, V., & Larson, E. L. (2010). Dissemination of health information through
social networks: Twitter and antibiotics. American Journal of Infection Control, 38, 182-
188. doi: http://dx.doi.org/10.1016/j.ajic.2009.11.004
Shamma, D., Yew, J., Kennedy, L., & Churchill, E. F. (2011). Viral actions: Predicting video
view counts using synchronous sharing behaviors. In Proceedings of the Fifth
International AAAI Conference on Weblogs and Social Media. New York, NY: ACM
Press. doi:10.1145/2527031.2527052

OBESITY RETWEETS 35
Shaver, P., Schwart z, J., Kirson, D., & O’ Connor, C. (1987). Emotion knowledge: Further
exploration of a prototype approach. Journal of Personality and Social Psychology, 52,
1061-1086. doi: 10.1037/0022-3514.52.6.1061
Strauss, R. S., & Pollack, H. A. (2003). Social marginalization of overweight children. Archives
of Pediatrics & Adolescent Medicine, 157, 746–752. doi:10.1001/archpedi.157.8.746.
Twitter (2013). Celebrating #Twitter7. Retrieved from https://blog.twitter.com/2013/celebrating-
twitter7
Wang, C., & Coups, E. J. (2010). Causal beliefs about obesity and associated health behaviors:
Results from a population-based survey. International Journal of Behavioral Nutrition
and Physical Activity, 7, 19. doi:10.1186/1479-5868-7-19
Yoo, J. H., & Kim, J. (2012). Obesity in the new media: A content analysis of obesity videos on
YouTube. Health Communication, 27, 86-97. doi: 10.1080/10410236.2011.569003.

OBESITY RETWEETS 36
Footnotes
1Two independent coders evaluated the relevance of each piece of data in a randomly
selected sample of data to create human-coded training data for a machine-learning classifier.
2For example, the frequency of the 31st most frequently retweeted message for “obese”
dropped to 24 times while the 1st ranked message for the same keyword was retweeted 7,188
times. “Fat” keyword was an exception to this pattern as it maintained a high retweet frequency
after the first 30 most retweeted messages: The 31st most retweeted tweet under the keyword
“fat” was retweeted 1,523 times.
3In a few instances where the tweet contained a link, the two coders discussed and
decided on whether the tweet itself contained enough information to be coded and to assist the
Twitter users’ decision to retweet the message. When the message was deemed to provide
insufficient information, it was marked and coders were instructed to visit the link contained in
the tweet and code for the dominant theme of the information provided in the link. When doing
so, the coders were reminded to put themselves in the position of a typical Twitter user and read
the content casually without overanalyzing the content.
4We anticipated some degree of inconsistency between the results yielded by the two
approaches since it was evident that there were significant differences in the retweet frequency
between not only the top retweets and the 30th retweets but also across the four keywords. For
instance, the keyword “fat” generated most retweets (79.93%), followed by “obese” (9.5%),
“obesity” (7.6%), and “overweight” (3.01%). The difference in the retweet frequency across the
four keywords was significant, χ²(3, N = 121,268) = 196,221.36, p< .001. Thus, if the study only
employed the first analytic approach, the findings may be misleading because they do not take
into consideration the actual frequency with which the tweets were shared.

OBESITY RETWEETS 37
5Although the chi-square statistic showed a statistically significant difference, it is likely to be
due to the large sample size, not an actual difference as indicated by descriptive statistics (50.6%
versus 49.4%).

OBESITY RETWEETS 38
Table 1
Common Themes of the Frequent Retweets about Obesity
Example tweet Top
Tweets
Retweet
frequency
Derogatory jokes Apparently, "I can't believe it's not butter", is not an appropriate comment
to make when your obese neighbors show you their newborn baby.
25% 49.7%
Non-derogatory jokes Dear Food, either stop being delicious or stop making me fat. 30.8% 32.8%
Personal experience or anecdote If it wasn't for softball, I would be obese. #softballprobz 5.8% 6.5%
Advocating societal change 77% of girls think they're ugly. 52% of girls think they're fat. 100% of
society should stop insulting girls for their appearance.
6.7% 4.5%
Causal factors for obesity Not eating breakfast increases your risk of becoming obese by 450%,
according to a UMass study.
13.3% 3.0%
Factoids When you hear eating disorder you automatically think of skinny instead of
obese
6.7% 2.5%
Sarcastic comments Americans spend $77billion a year treating obesity solving the worlds
water crisis would cost just $30 billion
4.2% .4%
Obesity prevalence Obesity kills 300000 Americans a year 3.3% .4%
Non-scientific weight loss tip Tired of being overweight? This program will turn your body into a fat
buring machine! [URL OMITTED]
1.7% .1%
Miscellaneous Today we wish a happy second anniversary and an early #FF to
@LetsMove, @MichelleObama’s initiative to fight childhood obesity.
2.5% 6.7%
* Note. The actual URL included in the tweets are omitted to preserve anonymity of the posters.

OBESITY ON TWITTER39
Table 2
Discrete Emotions Evoked by Retweets
Emotion Example tweet Top
Tweets
Retweet
frequency
Amusement Based on US obesity rates, soon candidates will just walk for president. 56.9% 78.8%
Contentment The only "overweight" thing about Adele is her paycheck 7.8% 11.5%
Surprise The difference between overweight & normal-weight Americans? Only 100 calories a day!
Burn it off: Go for a brisk 23-minute walk, vacuum for 30 minutes, or swim for 17 minutes.
15.7% 4.9%
Anger Camera Crew Discreetly Trails Overweight Woman For Obesity Segment [URL
OMITTED]
9.8% 3.7%
Sympathy Overweight guy asks for help [URL OMITTED] via @youtube 2.0% 0.4%

OBESITY ON TWITTER40
Disappointment I turn on @CNN looking for an update on the Syrian massacre. I get a story about "Pet
Obesity on the Rise". Yay, America!
2.0% 0.3%
Sadness 60 percent of all Americans are either overweight or obese while 30,000 people starve to
death each day.
2.0% 0.2%
Worry Learn About Health Problems Associated With Being Overweight [URL OMITTED] 2.0% 0.1%
Hope Lettuce. Carrots. I don't care what you say, these boys are bringing an end to teenage
obesity.
1.0% 0.1%

OBESITY ON TWITTER41
* Note. The actual URL included in the tweets are omitted to preserve anonymity of the posters.Table 3
Frequently Retweeted Humor Types
Humor types Example tweet Top
Tweets
Retweet
frequency
Puns Do drug dealers sell 'Diet Coke' to their overweight customers?? 25.8% 22.2%
Repartee "what do we want?!" ... "a cure for obesity" ... "when do we want it?" ... "after dinner!". 12.1% 18.1%
Parody Life is like a box of chocolates. It doesn't last long if you’re morbidly obese. 4.5% 12.1%
Sarcasm Happy Chocolate Day. But in these obese United States, every day is Chocolate Day 9.1% 9.1%
Ridicule Apparently clumsy people are more likely to be obese. That's because they keep walking
into things...like McDonald's.
4.5% 8.9%
Outwitting "Does this dress make me look fat?" No, I'm pretty sure your fat makes you look fat. 6.1% 7.5%
Absurdity I love my six pack so much, I protect it with a layer of fat. 10.6% 6.6%
Conceptual surprise Everybody thinks a Girl's Dream is to find the Perfect Guy. Lol no, our dream is to eat
without getting fat.
12.1% 6.4%

OBESITY ON TWITTER42
Anthropomorphism Dear Food, Either stop being delicious or stop making me fat. 1.5% 3.3%
Stereotype We all have that one skinny friend that eats more than fat person :P :D 1.5% 2.7%
Irony Dear Chubby kids chasing me, this is my way of helping cure Obesity...Sincerely, the Ice
Cream Truck Driver. =) #TeamGuilty
4.5% 1.1%
Embarrassment An obese guy was in the elevator with me. He caught me staring at the weight limit sign.
Awkward.
1.5% 1.0%
Irreverent behavior Dear obese gym teacher, We will run the mile once we see you do it first. Sincerely, your
students.
1.5% .5%
Disappointment I turn on @CNN looking for an update on the Syrian massacre. I get a story about "Pet
Obesity on the Rise". Yay, America!
1.5% .2%
Ignorance It's not fair that all those kids in Africa and Asia get to play hunger games while
American kids are stuck playing childhood obesity games.
1.5% .1%
Malicious pleasure 90% of the BAD females you went to high school with are now either A: overweight B:
have kids or C: both A & B
1.5% .1%

OBESITY ON TWITTER43
* Note. The actual URL included in the tweets are omitted to preserve anonymity of the posters.Table 4
Attribution of Causes of Obesity
Attribution of
causes
Specific causes Example tweet Top
Tweets
Retweet
frequency
Societal level Cheap and
unhealthy food
Here is why we have an obesity problem in America: Because Burgers
are $.99, & Salads are $4.99.
3.28% 11.94%
School system Excess homework has been linked as a cause to childhood obesity. 1.64% 1.0%
Portion size Very important #infographic on portion sizes and the #obesity epidemic
[URL OMITTED] #health #diet #food
1.64% 0.48%
Broken food system For the sake of our country &economy, it's time to see obesity/ diabetes/
allergies/ cancer for what they are: symptoms of a broken food system
1.64% 0.08%
Individual level Diet Don't pick on fat people. They have enough on their plates. 57.4% 60.8%
Exercise Relationships are like fat people, most of them don't work out. 26.2% 22%

OBESITY ON TWITTER44
Personality People who remain calm in stressful situations have higher rates of
depression and obesity, a study finds.
1.6% 3.2%
Genetic Is being overweight in the genes? [URL OMITTED] 4.9% 0.3%
Lack of sleep Lack of sleep increases your risk for heart disease, diabetes and obesity. 1.6% 0.2%
* Note. The actual URL included in the tweets are omitted to preserve anonymity of the posters.

OBESITY ON TWITTER45
Table 5
Test of Hypotheses through Comparison
Top
Tweets
Retweet
frequency
Retweet frequency by keywords
Overweight Obese Obesity Fat
H1
Emotion 85%a99%a 74.3% 98.3% 99.1% 100%
Non-emotion 15%b1%b 25.7% 1.7% 0.9% 0%
H2
Humor 55% 82.2%a 35.2% 88.3% 59.1% 85.4%
Non-humor 45% 17.8%b 64.8% 11.7% 40.9% 14.6%
H3
Individual cause 91.7%a86.5%a 100% 24.2% 83.8% 100%
Societal cause 8.3%b13.5%b 0 % 75.8% 16.2% 0%
* Note. Within the test of each hypothesis, percentages in the same column that do not share
subscripts differ at p < .001.