On Twitter while watching TV:
A blessing or a curse?
The influence of the valence and source of tweets on the viewers entertainment experience
Master’s Thesis
Kim van Halen – 11856718 University of Amsterdam
Graduate School of Communication
Master’s Program Communication Sciences - Entertainment Communication Supervisor: Susanne Baumgartner
2
Abstract
In this digital age, the way in which we watch television is changing. Instead of only watching a TV show, often social media platforms are used next to it, to see what others think of the show (Social TV). This seems a way to enrich the experience of watching a TV show, but how exactly can social media influence our entertainment experience? Existing studies that investigated the effect of social information on the entertainment experience have shown that valance of information is an important variable in that relationship. However, little research focusses on the possible influence of the source of social information. According to the Celebrity Endorsement Theory, a different effect is to be expected for social information coming from a celebrity compared to social information from an unknown person. The aim of this study was to investigate the possible influence of valance and source of tweets on the entertainment experience (enjoyment and transportation) of a TV show. An online experiment was conducted, with a 2(positive vs. negative) x 2(celebrity vs. unknown) factorial design and one additional control condition. The sample of this study consisted of 174 Dutch speaking participants, ranging from 18 to 39 years old. Results show that negatively framed tweets decrease one’s entertainment experience of a TV show. Furthermore, tweets seem more influential when they are written by a celebrity. Especially for positively framed tweets, the source of the tweet is even a decisive factor for making a significant difference in enjoyment. Positively framed tweets written by a celebrity increase one’s entertainment experience. The current research offers new perspectives regarding the effect of celebrity endorsement in Social TV, limitations of the current research and recommendations for future research are given.
3
Introduction
Watching television in the traditional linear way like we used to do, is not very common any more in this new digital age (Buschow et al., 2014). The emergence of new alternatives to linear TV, like On Demand and YouTube made traditional TV less popular (Buschow et al., 2014). Furthermore, the way in which we watch TV is also changing. Whereas formerly, watching television really was a social act, nowadays we often watch alone on our laptop or smartphone screens (Berry & Schleser, 2014). However, watching television will always keep a component of social interaction and shared experience (Katz & Lazersfeld, 1955). That is where social media comes into play. Social media offers a way to still connect with other people while watching your favorite show through a ‘second screen’. Often, a second screen is used while watching television for multiple purposes. Individuals can use the second screen for searching additional information (Nielsen, 2012), visiting a website mentioned during a TV show, or fact-checking information (Smith & Boyles, 2012). Furthermore, a second screen is used to create a social viewing experience, especially while watching TV alone. Individuals use their second screen to post comments, view what others post and to live-chat with others (Smith & Boyles, 2012).
Using social media while watching TV is often called ‘Social TV’. Social TV is defined as the communicative exchange about linear television content (Buschow et al. 2014). This communicative exchange can include pre-communication, parallel-communication, and follow-up communication, but in most studies only parallel-communication is included in the definition of Social TV (Giglietto & Selva, 2014). Therefore, in this study, we only include parallel-communication in our definition of Social TV, so, watching TV and using social media related to the content of the show at the same time.
4 Especially Twitter is a commonly used medium for engaging in Social TV. Television shows or related topics appear very often in Twitter’s ‘trending topics’ (Deller, 2011). Not only the audience uses social media while watching TV, television producers have also embraced social media. They use it for example to create hashtags or to show extra content (Harrington et al., 2013). Twitter and other social media become a kind of virtual loungeroom, where television producers can connect an active audience with their favorite TV show and enhance their entertainment experience (Harrington et al., 2013).
The current research will examine potential effects of engaging in Social TV on the entertainment experience of individuals, in particular the effect of positively or negatively phrased tweets. Because television producers try to enhance the entertainment experience of their audience online, it is useful to know what effect social media content actually has on the entertainment experience of their audience. Some studies showed a relationship between online social information of others and one’s entertainment experience (e.g. Waddell & Sundar, 2017; Möller & Kühne, in press; Möller et al., 2018), but more research is desired to back up those findings. Furthermore, most research focuses on online videos and comments, whereas the current research will focus on TV shows and tweets.
Moreover, previous research failed to address the possible influence of the source of social information. Especially, the influence of celebrities might be interesting in an entertainment context. Among the 326 million users of Twitter, there are also a lot of celebrities who use this social networking site to express their thoughts. According to the Celebrity Endorsement Theory (Amos et al., 2008), people are more influenced by communication expressions from celebrities compared to communication expressions of unknown people, mainly because of the perceived trustworthiness of celebrities. Especially in a marketing context, celebrity endorsement is proven to be effective (e.g. Atkin & Block, 1983; Petty et al., 1983). However, this has not yet been tested in an entertainment context.
5 The research question that is aimed to answer in this research is: To what extent does the valence of tweets about a TV show influence the entertainment experience, and does this influence differ for tweets from celebrities and tweets from unknown people?
Theoretical background
Engaging in Social TVWhy people use Twitter to enrich their media experience can be explained by the Uses and Gratifications Theory of Blumler and Katz (1974). This theory describes how people proactively seek out different types of media with the aim to satisfy a specific need or gratification, and how media enjoyment (obtained gratification) is influenced by many different psychological and social aspects. Watching a television show and using social media can both fulfil different needs or gratifications, and thus, have a different effect on media enjoyment (Whiting & Williams, 2013). With combining social media and television (Social TV), new forms of media enjoyment could arise. Therefore, in this research, Uses and Gratifications Theory can help us understand the possible effects of tweets (social media) on media enjoyment (entertainment experience).
Before understanding and discussing the possible effects of Social TV, some information about why and how people engage in Social TV is needed. People have different motivations for engaging in Social TV. The primary motivation for engaging in Social TV is the sense of community and experiencing connectedness with people with shared interests (e.g. Harrington et al., 2013; Ji & Raney, 2015). Furthermore, individuals engage in Social TV because they have the need to express their opinion about the TV show they are watching (Larsson & Moe, 2011). The content of the tweets is very diverse, but according to a content analysis of Buschow et al. (2014) the main subject of explored tweets about TV shows were
6 communication within the Twitter community and evaluations of the show and actors. In addition, they found that different shows evoke different kinds of tweets. During a talent show, people tweet about fandom and critiques of the candidates, during live events, critical debates about the show arised, and political talk shows stimulated public discourses on Twitter.
Altogether, people engage in Social TV for different reasons and in different forms, but how does Social TV actually affect us? With the Uses and Gratifications Theory in mind, Twitter is a suitable venue to view how people express their uses and gratifications of their television viewing behavior (Wohn & Na, 2011). Tweets show the immediate reaction of an audience, which may reflect the nature of their uses and perhaps affect the their short-term gratifications and entertainment experience (Wohn & Na, 2011).
Peer influence and social conformity
Another point of view on the effects of Social TV can be found in peer influence and social conformity. The opinion of others can influence how we think about certain topics. Of course, this is something that TV broadcasters also know. That is why they often integrate the responses of others in their own show, for example by showing the audience in the studio (Nabi & Hendriks, 2003). Several studies show that people tend to conform to responses of other audience members. Fein et al. (2007) found that judgements of candidate performance in presidential debates are influenced by the mere knowledge of others’ reactions. The results show the power of social context on influencing individual’s opinions.
Another way for broadcasters to use responses of the public in their show is through continuous response measures (CRMs), which are often displayed as graphs or charts. CRMs measure the real-time opinions of individuals during the consumption of a media message (Davis et al., 2011). Several studies show that showing others’ reactions through CRM can
7 influence the opinion of individuals. A study by Weaver et al. (2009) found that showing a line graph during a music video with positive, negative or neutral evaluations, resulted in a predictable response. When the CRM showed a positive evaluation of the video, respondents had a more positive opinion about the video, and when the CRM showed negative evaluation, respondents were more negative. Furthermore, Davis et al. (2011) showed a manipulated CRM in the form of a line graph during a live political debate. The perceptions of participants about who won the debate, preferred candidate and voting intentions conformed to the manipulated CRM.
Social media can be seen as a new form of the traditional CRMs in representing the public opinion (Cameron & Geidner, 2014). So, instead of a graphical representation of people’s opinion, with social media one can show or look for individual comments during a TV show. Just as studies showed for CRMs, also for individual comments, viewers tend to adjust their judgements to what others say on social media. Cameron and Geidner (2014) found that the viewers’ perception was influenced by a manipulated Twitter-based CRM. They showed the respondents a Twitter feed with either mostly negative or positive tweets about the performance of the TV presenter. For almost all respondents in their study, the viewers judgement of the performance of the TV presenter was the same as the majority opinion represented in the showed Twitter feed.
An explanation for why other people’s opinions can influence your own opinion can be found in the Social Conformity Theory from Asch (1951). This theory states that people tend to form their opinions, attitudes and behaviours based on those of others. Basing your opinion on what others say is a way to fit into a group, a desire that emanates from the nature of human being (Asch, 1951). The Social Conformity Theory (Asch, 1951) also tells us that peer pressure and group pressure are important influencers of one’s opinion. Influence and power of groups is seen as a significant factor in human development (Rodkin, 2004), and is
8 therefore always from influence in an individual’s life. Asch (1951) founded this theory after an experiment in which participants had to make judgements on a line-matching task in a room with others. Every member in the room identified which lines were the same length, and the participant had to answer lastly. As Asch (1951) predicted, most participants conformed to the answer of the other people in the room, even if the others answered wrongly.
The theory of Asch (1951) is also in line with the argumentation of Waddell and Sundar (2017), they suggest that people conform to social comments because they assume that the comments of a few reflect the opinion of a majority. This is based on the MAIN (Modality-Agency-Interactivity-Navigability) model, which states that user comments affect the perception of others by activating the bandwagon heuristic. This heuristic entails the rule of thumb that ‘if others like it, I should like it too’.
In this digital age, social conformity might often take place through electronic word of mouth (EWoM). There is particularly much research about the EWoM effect in a marketing context. For example, Chevalier and Mayzlin (2006) found that online book reviews strongly influence people’s purchase behavior. When a book has positive book evaluations, it is more likely that someone else will buy it. Furthermore, it appeared that hotel reviews can influence booking intentions (Tsao, et al., 2015). Respondents were more likely to book a hotel when reviews were positive.
As discussed above, previous research shows that responses of others can have an influence on one’s opinions, but what would be the effect on the entertainment experience of people? The kind of conformity effects describe above could also occur during entertainment experiences. It has been found that the evaluations of others tend to have an effect on how someone experiences media content (Ramanatha & McGill, 2007).
9 Entertainment experience refers to “the immediate phenomenal awareness that users of acknowledged media entertainment productions typically have” (Tan, 2008, p. 29). In this study, entertainment experience is divided into two components. The first component is enjoyment, which can be seen as a pleasurable response to media use (Raney, 2003), that lies at the heart of the entertainment experience (Vorderer et al., 2004). The second component of entertainment experience is transportation, often also described as engagement. Engagement indicates the extent to which one becomes engaged, transported or immersed in a narrative (Busselle & Bilandzic, 2009). In this study it is expected that the presence and valance of tweets will have an influence on both enjoyment and transportation.
Few research has been done about the influence of online evaluations on one’s entertainment experience of a TV show. However, Shedlosky-Shoemaker et al. (2011) studied this effect in another context, they investigated the influence of peer evaluations on the enjoyment and transportation of written narratives. In this study, participants had to read either positive or negative online peer evaluations before reading a narrative. Results show that social influence guides reader’s enjoyment and feelings of transportation. Thus, negative evaluations decreased respondent’s enjoyment and transportation, and positive evaluations increased respondent’s enjoyment and transportation. Moreover, this effect was found to be stronger for participants who were given negative evaluations of the stimulus.
The results of the study of Shedlosky-Shoemaker et al. (2011) are related to the results of Ramanathan and McGill (2007), who investigated how enjoyment of a movie can be affected by watching it together with a friend. Results showed that joint consumption leads to similar patterns and evaluations. If an individual has someone sitting next to them expressing signs of enjoyment, that individual will enjoy the movie more.
Whereas previous studies focus on either the effect of online evaluations in other subject areas or the effect of ‘offline’ evaluations in entertainment media, a recent study of
10 Möller and Kühne (2018) combined both and focused on the effect of user comments on the entertainment experience of online videos. They found that positive user comments increased the hedonic entertainment experience of new viewers. This is also in line with another study of Möller et al. (in press), in which results show that negative social information had a decreasing effect on the enjoyment of online videos, and positive social information had an increasing effect.
Another study that supports the effect of online evaluations on entertainment experience is a study of Waddell and Sundar (2017). They investigated whether comments on social media can affect an individual’s TV program enjoyment, and whether the valence (negative vs. positive) and placement (beginning vs. end of program) of the comments had different influences. Results of this study show that negative social media comments have an indirect effect on viewer’s enjoyment, and led to less enjoyment. For positive comments, no effect was found.
However, studies of Möller and Kühne (2018), Möller et al. (in press), and Waddell and Sundar (2017) are one of the few studies that focus on the effect of online evaluations on entertainment experience and more research is needed to support this effect. In addition, previous studies mostly focused on the effect on enjoyment of online videos, whereas the current study focuses on enjoyment of a TV show. Therefore, the aim of the current study is to examine the effect of tweets (negative vs. positive) on the entertainment experience (enjoyment and transportation) of a TV show. Based on previous literature, for the current study, the following hypothesis will be tested:
H1 – Positively framed tweets will increase one’s entertainment experience of a TV show
11 Furthermore, the study of Möller et al. (2018) showed that - next to the effect of social information on enjoyment of online videos - a negativity bias occurred. This study was the first to examine how much attention is paid to social information alongside an online video. It appeared that respondent paid more attention to negative social information compared to positive social information. In addition, the effect of social information on enjoyment was stronger for negative social information. As discussed before, also in the study of Waddell and Sundar (2017), only an effect on enjoyment was found for negatively phrased comments. This is also in line with literature about the negativity bias in general. According to Baumeister et al. (2001), the bias for negative information derives from the beginning of humanity, it is a human instinct to quickly detect negative events in order to stay save and survive threats. Several studies in other context found a bias for negative information: people are better in detecting negative words on a screen than positive words (Dijksterhuis & Aarts, 2003); people spend more attention to negative online product reviews than positive or neutral ones (Daugherty & Hoffman, 2014); and people are more likely to select negative news articles than positive ones (Trussler & Soroka, 2014). Therefore, in the current study, a third hypothesis will be tested:
H3 - The effect of negatively framed tweets on one's entertainment experience will be stronger compared to the effect of positively framed tweets
Celebrity endorsement
This study focuses not only on the valance (positive vs. negative) of a tweet, but also on the source of the tweet. In particular, this research will investigate the potential difference between tweets from celebrities compared to tweets from unknown people. Taking into account the Celebrity Endorsement Theory (Amos et al., 2008), we could expect a difference in the effect of tweets from these two different sources.
12 Especially from a marketing position, celebrity endorsement is an often-used instrument. Using celebrities in marketing is not a new phenomenon, but already exists since the late nineteenth century (Erdogan, 1999). Advertisers use celebrities in their marketing strategy to promote their products. According to several studies, celebrity endorsement in the marketing field results in more positive attitudes towards advertising and increases an individual’s purchase intentions compared to non-celebrity endorsement (e.g. Atkin & Block, 1983; Petty et al., 1983).
An explanation for this effect can be found in the Source Credibility Model (Hovland & Weiss, 1951). This model states that the effectiveness of a communication message depends on the level of trustworthiness and expertise the endorser is perceived to have (Hovland & Weiss, 1951). Celebrities seem in general more trustworthy and expert than unknown people (e.g. Amos et al., 2008; Ohanian, 1990), and therefore the statements of a celebrity could influence one’s beliefs, attitudes, behavior or opinion more strongly. This seems to be especially the case for young people, most likely because of the significant connection between the lives of young people and celebrity culture (Turner, 2004). On Twitter in particular, celebrities might be seen as fellow social media users, and therefore their electronic word of mouth seems to be more credible and trustworthy compared to celebrities appearing on TV or in print (Russell, 2012).
Besides the often successful use of celebrity endorsement in marketing, celebrity endorsement is also commonly used in other sectors, like the political environment (Garthwaite & Moore, 2013; Austin et al., 2008). For example, the endorsement of Oprah Winfrey for Barack Obama resulted in an increase of votes and campaign contributions for Obama during the 2008 presidential election (Garthwaite & More, 2013), and celebrities can have a positive impact on motivating young people’s voting behavior (Austin et al., 2008). In addition, celebrity endorsement has also proven to be successful in health care, where for
13 example celebrity endorsement for breast cancer screening resulted in respondents being more likely to undergo mammography (Larson et al., 2005).
Concluding from the studies discussed above, the opinion of a celebrity is usually more influencing than the opinion of an unknown person. Although research about celebrity endorsement mostly focuses on marketing, politics and health care environments, it could provide a foundation of the effect of celebrity endorsement in the entertainment environment. Based on existing literature, we now know that in the field of marketing, politics and health care, the opinion of a celebrity could be more influencing than the opinion of an unknown person. In the field of entertainment experience, we only know that online opinions (social information, comments) can have an effect on our media enjoyment, but there is no research that focuses on the possible effects of the source of social information. Based on research about celebrity endorsement in different areas, it would be reasonable to think that also in an entertainment context, the opinion of a celebrity would be more influencing than the opinion of an unknow person. More specifically, this research will focus on the effect of celebrity word of mouth on the entertainment experience of individuals. Previous research suggests the following hypotheses for the current research:
H4 – The influence of tweets on one’s entertainment experience will be stronger for tweets from celebrities compared to tweets from unknown people.
Method
ParticipantsFor this research, participants were recruited online, via Facebook. A Facebook post was placed via the social network of the researcher. Participants were also asked to pass on the
14 questionnaire to others, so a snowball effect occurred. There were 191 participants who started the questionnaire, but 8.9% dropped out before finishing. In the end, the final sample consisted of 174 participants. The participants were aged between 18 and 39 years (M = 26.55, SD = 5.26). The majority of the final sample consisted of females (60.3%). Looking at educational level, the biggest part, 73%, of the sample was enrolled in higher education (37.4% University, 35.6% University of Applied Sciences), 26.4% was enrolled in lower education, and 0.6% was not enrolled in any of the given options. All participants were Dutch speaking.
Procedure
To test the hypotheses in this research, an online experiment was conducted. This experiment consisted of a 2x2 factorial design with positive vs. negative tweets and celebrity vs. unknown people as factors. In addition, one control condition without any tweets was used. Participants were randomly assigned to one of the five conditions (celebrity + negative, celebrity + positive, unknown + negative, unknown + positive, control). Anonymity was ensured by providing participants an anonymous link which did not register any personal information. In addition, the option ‘anonymize responses’ was enabled, so no IP addresses were registered. All the questions in the questionnaire were provided in Dutch. To participate, participants had to follow a link that was placed on Facebook. Firstly, a short introduction was presented, and participants had to agree with an informed consent. Secondly, some demographics (age, gender, education) were asked. After that, all participants were shown the same video (part of a TV show) with different tweets aside (depending on condition). Then they were asked some questions about how much they enjoyed the video and if they felt transported into the video. Lastly, some questions were asked about familiarity with the TV show, familiarity with the specific episode, and likeability of the TV show. The participants in the conditions that contained tweets were also asked about the presence of the tweets, valence of the tweets,
15 familiarity with the source of the tweets, and likeability of the celebrity. Finally, participants were thanked for their participation and an e-mail address for further questions and remarks was provided. Participation in the study took approximately five minutes.
Material
For this research, manipulated tweets were used which can be found in appendix 1. Participants in the negative tweet conditions were shown four negative tweets (e.g. ‘What a boring test, no excitement at all. Boring!’, ‘I don’t see any team spirit in this test, please work together!’, How hard can it be to keep the torch burning? Losers!’, and ‘The tests are getting more boring with every episode’). Participants in the positive tweet conditions were shown four positive tweets (e.g. ‘What an excited test, come on!’, ‘What a great team spirit in this test, collaboration is key!’, ‘So nice that Team Red is up front! Heros!’, and ‘The tests are getting more excited every episode!’). The positive and negative tweets were aimed to be as equal as possible in terms of length and content, only differing in valence. The participants in the conditions that contained tweets from celebrities, got tweets from Chantal Janzen (a famous Dutch actress) and Martijn Krabbé (a famous Dutch presenter). The participants in the condition with tweets of non-celebrities, got tweets from non-existing persons. Two tweets in every condition were from a female person and two tweets were from a male person. This is because research has shown that men and women stronger identify with their own sex (Hill & Lynch, 1983), and by providing both sexes in the manipulated tweets, female participants as well as male participants can identify with the source of the tweets.
The video that was shown in this experiment contained a part of an episode of Expeditie Robinson, which is a Dutch reality show about people having to survive on an island and play battles against each other. The video showed the denouement of a battle between two teams and lasted 1 minute and 45 seconds. The video had some interesting parts, like the excitement about who is about to win the battle, but it had also some less interesting
16 parts, like some comments of the competitors. Reality TV and shows with a competition element seem to be very suitable for engaging in Social TV (Buschow et al., 2014). The video of Expeditie Robinson had both of those elements, and was therefore a suitable video for this experiment.
Measurements
Enjoyment - The measurement of enjoyment was done with the enjoyment scale of Wirth,
Hofer and Schramm (2012). This scale was originally developed to measure hedonic enjoyment of a sad movie, but has been used to measure enjoyment in very different contexts since. The scale consists of four items (e.g. It gave me pleasure to watch the video, watching the video amused me, the video was exciting, and I enjoyed the video). Participants had to answer on a 7-points Likert scale (1=strongly disagree, 7 = strongly agree). To test if the items measure the same concept, a factor analysis was conducted. The factor analysis showed that the scale was unidimensional, with one component with an Eigenvalue higher than 1.00. The dimension, existing of four items, explained 87.13% of the total variance. In addition, to test the reliability of the scale, a reliability analysis was conducted. The reliability analysis showed a Cronbach’s Alpha of .95, which means the scale was reliable. The mean score of the scale was 4.22 (SD = 1.29).
Transportation - Transportation was measured with the engagement scale of Buselle
and Bilandzic (2009), which is often used for measuring narrative engagement. The scale was adjusted for this experiment and consists of four items (e.g. I was mentally involved with the video, I was really pulled into the story, I wanted to learn how the video ended, and I wanted to know how the events would unfold). Two items form the original scale (I was completely immersed in the video and The viewing experience was intense for me) were left out because of an overlap with other items when translated to Dutch. To test if the items measure the same concept, a factor analysis was conducted. The factor analysis showed that the scale was
17 unidimensional, with one component with an Eigenvalue higher than 1.00. The dimension, existing of four items, explained 84.65% of the total variance. In addition, to test the reliability of the scale, a reliability analysis was conducted. The reliability analysis showed a Cronbach’s Alpha of .94, which means the scale was reliable. The mean score of the scale was 3.58 (SD = 1.21).
Familiarity with the show - The familiarity of the show was measured with one item
about how often the participants watch Expeditie Robinson, and was measured on a 5-points scale (1 = never, 5 = always). The mean score of this item was 2.84 (SD = 1.09).
Familiarity episode - If participants had seen the showed episode before was
measured with one item (Have you seen the showed episode before participating in this research?), with three answer possibilities (yes, no, don’t know). Most participants did not see this episode before (64.9%). 25.9% did see it before, and 9.2% did not know if they had seen it before.
Likeability of the show - The likeability of the show was measured with one item
about how much they liked Expeditie Robinson, and was measured on a 5-points Likert scale (1 = not at all, 5 = very much). The mean score of this item was 3.87 (SD = .92).
Presence of the tweets - To check if participants had read the tweets, one question was
asked about whether they had read all or some tweets (Yes all, Yes some, No). To check the manipulation of tweets vs. no tweets (control), a last answer option ‘There were no tweets present’ was added. Most participants read all the tweets (67.1%), 27.9% read some tweets, and 5% did not read any tweets. There were no participants in the conditions that contained tweets, who chose the option ‘There were no tweets present’. All participants in the control condition (without tweets) chose the option ‘There were no tweets present’.
18
Valence of the tweets - To check the manipulation of the valence of the tweets
(negative vs. positive), one item was used to ask the participant’s opinion about the valence of the tweets. This item was measured on a 5-points scale (1 = very negative, 5 = very positive). In the condition with the negative tweets, the mean score was 2.17 (SD = 1.99). In the condition with the positive tweets, the mean score was 4.21 (SD = .58).
Familiarity with the source - To check the manipulation of the source of the tweets
(celebrity vs. non-celebrity), one item was used to ask the participants whether they knew the source of the tweet (Yes, No, Don’t know). In the condition with the celebrity tweets, 75.45% knew the source, 12.75% did not know the source, and 12.80% was not sure if they knew the source. In the condition with the non-celebrity tweets, 87.35% did not know the source, 12.65% was not sure if they knew the source, and no participants indicated that they knew the source.
Likeability of the celebrity - As discussed before, an important determinant of the
Celebrity Endorsement Theory is trustworthiness. Friedman et al. (1978) discovered that likeability tends to be the most important attribute for trust. Therefore, in this research, likeability of the celebrity was also measured. Likeability was measured with one question about how much the participants liked the celebrity, on a 5-points scale (1 = not at all, 5 = very much). The mean score was 3.24 (SD = 0.10) (Chantal Janzen: M = 3.18, SD = 0.09; Martijn Krabbé: M = 3.30, SD = 0.11).
19
Results
Randomization check
To test whether participant’s gender was equally distributed over the five conditions, a Chi-square test was conducted. This analysis showed that gender was equally distributed among the conditions, Chi-square(4) = .16, p = .997.
To test whether age was equally distributed over the five conditions, an ANOVA was conducted with experimental condition as independent variable and age as dependent variable. From the ANOVA, it could be concluded that age does not differ across conditions, F(4,169) = 1.04, p = .391.
To test whether education level was equally distributed over the five conditions, a Chi-square test was conducted. This analysis showed that there were no differences between the conditions concerning educational level, Chi-square(12) = 12.54, p = .403.
To test whether the participants read the same amount of tweets in every condition (that contained tweets), a Chi-square test was conducted. This analysis showed that there were no differences between the conditions concerning the attention to tweets, Chi-square (6) = 9.76, p = .135. It thus seemed that the randomization across conditions was successful.
Various other control variables were measured in the questionnaire. To make sure that those variables were equally distributed among the conditions, a randomization check was needed. For testing the equal distribution of Familiarity with the show, an ANOVA was conducted, with experimental condition as independent variable and familiarity as dependent variable. The ANOVA showed an equal distribution, F(4,169) = .24, p = .918. For Familiarity with the episode, a Chi-square test was conducted. The analysis showed that there was no difference between the conditions, Chi-square(8) = 1.48, p = .993. For Likeability of the show,
20 an ANOVA was conducted with experimental condition as independent variable and likeability as dependent variable. The ANOVA showed an equal distribution, F(4,152) = 1.28, p = .280 across conditions. For Likeability of the celebrity, another ANOVA was conducted with experimental condition as independent variable and likeability as dependent variable. Also for this variable, the ANOVA revealed an equal distribution across the two conditions, F(1,49) = .06, p = .806.
Manipulation check
Presence of the tweets - In the conditions that contained tweets, 5% of the participants did not
read any tweets. This data can be excluded from further analysis, and therefore six cases have been indicated as missing data (N = 168).
Valence of the tweets - To test whether the valance of the tweets was manipulated
correctly, an independent-samples t-test was conducted, with experimental condition (positive tweets / negative tweets) as independent variable and perceived valence as dependent variable. Results showed that the perceived valance of the tweets in the positive condition (M = 4.21, SD = .58) significantly differed from the perceived valance of the tweets in the negative condition (M = 2.15, SD = 2.39), t(137) = -7.10, p < .001, CI = [-2.63, -1.49], Cohen’s d = .12. Therefore, it could be concluded that the manipulation of the valance of the tweets was successful.
Familiarity with the source - To test whether the source of the tweets was
manipulated correctly, a Chi-square test was conducted. The analysis showed that there was a significant difference between the conditions with tweets from celebrities and the conditions with tweets from unknown people, Chi-square(2) = 93.14, p < .001. Significantly more people knew the source in the celebrity-condition. Therefore, it could be concluded that the manipulation of the source of the tweets was successful.
21 Test of Hypotheses
The first three hypotheses were all about the direct effect of the valance of tweets on entertainment experience. To test the hypotheses 1) Positively framed tweets will increase one’s entertainment experience of a TV show; 2) Negatively framed tweets will decrease one’s entertainment experience of a TV show; and 3) The effect of negatively framed tweets on one's entertainment experience will be stronger compared to the effect of positively framed tweets, a one-way ANOVA test was conducted with valance of tweets as independent variable and enjoyment and transportation as dependent variables (N = 168).
As seen before - at the randomization checks - the control variables were all equally distributed among the different conditions. In addition, a bivariate correlation test was conducted to make sure the control variables were not correlated with the dependent variables (enjoyment and transportation). Results show that there was no significant correlation (all p > .08), and therefore, the control variables were not included in further analyses.
The results of the one-way ANOVA test showed that participants differed across conditions in terms of how much they enjoyed the video, F(2, 165) = 46.15, p < .001, η² = .036. This means that 3,6% of the explained variance in enjoyment can be explained by the valance of the tweets. Furthermore, the results also showed that participants differed across conditions in terms of how much they felt transported into the video, F(2, 165) = 56.92, p < .001, η² = .047. This means that 4,7% of the explained variance in transportation can be explained by the valance of the tweets. Results of a Tukey’s HSD post-hoc test (homogeneity of variances assumed) showed that there was no significant difference in enjoyment between participants in the positive condition (M = 4.88, SD = 1.09) and participants in the control condition (M = 4.68, SD = .81), p = .621. Furthermore, no significant difference in transportation between participants in the positive condition (M = 4.25, SD = 1.07) and
22 participants in the control condition (M = 4.06, SD = 1.18) was found, p = .604. Therefore, H1 can be rejected.
Moreover, results of the post-hoc test show that participants in the negative condition (M = 3.25, SD = 1.07) enjoyed the video significantly less than participants in the control condition (M = 4.68, SD = .81), p < .001. Meaning, when negative tweets were read during a TV show, enjoyment decreased with 1.43 scale points compared to not reading any tweets. Furthermore, participants in the negative condition (M = 2.62, SD = .57) also felt more transported into the video than participants in the control condition (M = 4.06, SD = 1.18), p < .001. Meaning, when negative tweets are read during a TV show, transportation decreases with 1.44 scale points compared to not reading any tweets. Therefore, H2 can be accepted.
To test the third hypothesis that posed that the effect of negatively framed tweets on one's entertainment experience will be stronger compared to the effect of positively framed tweets, we had to look at the difference in effect of the negative tweets on entertainment experience and the effect of positive tweets on entertainment experience. As discussed above, there was no significant effect for positive tweets on either enjoyment or transportation. Whereas the effect of negative tweets on enjoyment and transportation, were both significant (enjoyment: Mdifference = 1.43, p < .001; transportation: Mdifference = 1.44, p < .001). Therefore, H3 is supported.
The fourth hypothesis stated that the source of the tweet can moderate the effect of the valance of the tweets on the entertainment experience, and that the influence of tweets on one’s entertainment experience will be stronger for tweets from celebrities compared to tweets from unknown people. To test this hypothesis, firstly a two-way ANOVA with source of tweet and valance of tweet as independent variables and enjoyment as dependent variable was conducted (N = 134). Results showed that there was a significant direct effect of valance
23 of tweets on enjoyment, F(1, 132) = 80.17, p < .001, η² = .036. No significant direct effect was found for the source of tweets on enjoyment, F(1, 132) = 2.08, p = .175.
However, a small significant interaction effect between the source of the tweet and the valance of the tweet on enjoyment emerged, F(1,132) = 6.41, p = .012, η² = .030. This means that 3% of the explained variance in enjoyment can be explained by the valance of the tweets and the source of the tweets. Respondents who read the positive tweets enjoyed the video more when the tweets were from celebrities (Mpos_cel = 5.24, SDpos_cel = 1.16; Mpos_unk = 4.52, SDpos_unk = .89), and respondents who read the negative tweets enjoyed the video less when the tweets were from celebrities (Mneg_cel = 3.13, SDneg_cel = .96; Mneg_unk = 3.34, SDneg_unk = 1.16) (see Figure 1). Therefore, the fourth hypothesis can be accepted.
Additionally, Some t-test were conducted to see which means actually differ significantly. Results showed that for the positive condition, participants with tweets from celebrities (M = 5.24, SD = 1.16), enjoyed the video significantly more than participants with
1 1,5 2 2,5 3 3,5 4 4,5 5 5,5 Negative Positive
Figure 1. Interaction effect of valence tweets and celebrity endorsement on enjoyment
24 tweets from unknown people (M = 4.52, SD = .89), t(66) = 2.83, p = .006, CI = [.21, 1.21], Cohen’s d = .70. However, for the negative condition, no significant difference was found between tweets from celebrities (M = 3.13, SD = .95), and tweets from unknown people (M = 3.35, SD = 1.16), t(64) = -.81, p = .424, CI = [-.74, .32], Cohen’s d = .04. This was interesting, because it seemed that negative tweets were influential no matter who the source was, but for positive tweets it seemed to matter whether the tweets were posted by a celebrity or an unknown person.
For transportation, another two-way ANOVA with source of tweet and valance of tweet as independent variables and transportation as dependent variable was conducted (N = 134). Results showed that there is a significant direct effect of valance of tweets on transportation, F(1, 132) = 118.49, p < .001, η² = .047. No significant direct effect was found for source of tweets on transportation, F(1, 132) = .193, p = .661. Moreover, no interaction effect emerged, F(1, 132) = .63, p = .429. Therefore, H4 can only be partly accepted. Only the influence of tweets on enjoyment was found to be significantly stronger for tweets from celebrities compared to tweets from unknown people.
Discussion
The aim of the current study was to investigate the possible effects of tweets (positive vs. negative) on the entertainment experience of a TV show, and the possible moderating effect of the source of the tweet (celebrity vs. unknown person). Results show that negative tweets can have a small effect on both enjoyment and transportation. Respondents who were exposed to negative comments reported less enjoyment and felt less transported compared to respondent who did not see any tweets. Thus, the current study shows that negative tweets lead to a more negative entertainment experience. This is in line with the Social Conformity
25 Theory from Asch (1951), which states that individuals form their opinion based on other’s opinion. Furthermore, this also supports previous research on the effects of evaluations of others on entertainment experience (e.g. Waddell & Sundar, 2017; Möller et al., 2018; Möller & Kühne, in press). This means when engaging in Social TV, the content of the TV show is not the only predictor of one’s entertainment experience, also the valance of the tweets can influence how much viewers enjoy the show.
In contrast to the significant result for negative tweets, no significant effect was found for positive tweets on the entertainment experience. This is interesting, because it is not in line with previous findings (e.g. Möller et al., 2018; Möller & Kühne, in press), which all found an effect for both positive and negative comments. However, the findings are in line with the study of Waddell & Sundar (2017), they also only found an effect for negative comments. This could be explained by another finding from the study of Möller et al. (2018), in which a negativity bias occurred among the participants. It appeared that participants paid more attention to negative social information compared to positive social information. Therefore, it might be in the current research, that less attention was paid to positive comments, thus the significant effect did not occur. This is also in line with the theory that states that people pay more attention to negative information because of human instincts to detect danger (Baumeister, 2011).
Möller et al. (2018) also found that the effect of social information was stronger for negative social information. This is supported by the current research, from which the results also show that the effect of negatively framed tweets on one’s entertainment experience was stronger compared to the effect of positively framed tweets.
Whereas Möller et al. (2018) were the first to found the negativity bias effect in the context of online videos and enjoyment, this research is the first to find the negativity bias effect in the context of Social TV and entertainment experience. The main difference between
26 those two is that with online videos, the comments are always present, and in Social TV, you proactively search for the comments yourself when going to Twitter.
The findings of the current study also show that the source of the tweet (celebrity vs. unknown) has a small moderating effect on the relationship between the valance of the tweets and enjoyment. It seemed that negative tweets were influential no matter who the source was, but for positive tweets it seems to matter whether the tweets came from celebrities or unknown people. There was only a significant difference in enjoyment when the positive tweets came from a celebrity.
The findings of the positive tweets are in line with the Source Credibility Model which states that the effectiveness of a communication message depends on the level of trustworthiness and expertise the endorser is perceived to have (Hovland & Weiss, 1951). Furthermore, according to the Celebrity Endorsement Theory, celebrities seem in general more trustworthy and expert than unknown people (e.g. Amos et al., 2008; Ohanian, 1990), and therefore the statements of a celebrity could influence one’s beliefs, attitudes, behavior or opinion more strongly. However, the finding for negative tweets is not supported by this theory. It seems that the negativity bias was so strong that the source of the tweet did not matter.
In contrast to the significant results for enjoyment, no significant difference was found for the effect of celebrity endorsement on transportation. An explanation for this could be that transportation is a more elaborated process than enjoyment, and therefore not reached or influenced as easily. Van Laer et al. (2014) suggest that narrative transportation can be reached through for example identifiable characters, attention and imaginable plot. So, it could be true that factors like celebrity endorsement do not affect transportation, but can still influence enjoyment. Furthermore, the episode that was shown in this research, only lasted
27 1:45 minutes. Assuming that transportation is a more elaborated process, it could be that the video was simply too short for participants to feel transported into the story line.
Altogether, the current research contributes to the already existing literature about the influence of evaluations of others on the entertainment experience. However, compared to previous research about online (YouTube) videos, this study investigated the influences of Social TV. Results show that negatively framed tweets can decrease one’s entertainment experience of a TV show. Furthermore, a new insight to the field of Social TV is the effect of celebrity endorsement: the influence of tweets on enjoyment of a TV show is found to be stronger for tweets from celebrities compared to tweets from unknown people, but only for positive tweets.
Hence, this study showed that tweets play an important role in the entertainment experience of a TV show in two ways: Negatively framed tweets can decrease one’s entertainment experience no matter who the source is, and positively framed tweets can increase one’s enjoyment of a TV show only when the source is a celebrity. Next to filling in this gap in the literature, the findings are also of value for broadcasters and television producers of Reality TV, because they now know that positive celebrity endorsement is an effective way to make your audience enjoy the show more, and in the end, probably watch it more often.
Limitations and further research
The current study has some limitations that need to be mentioned and potentially addressed in future studies. First of all, for this study an experiment was conducted and respondents were exposed to tweets alongside a video. In other words, respondents were forced to look at the tweets. It could be possible that the respondents of this research actually never engage in Social TV, and therefore the effects that are found in this experiment, would not happen in the
28 real world. Or the other way around, it could also be that for those who engage in Social TV, the effects are stronger because they seek out this kind of additional information. Further research should consider to study the effects of Social TV in the ‘real world’ and only include people in their sample that engage in Social TV, which would enlarge the ecological validity of the results.
Secondly, as previous research has shown, different genres of TV shows can evoke different forms of Social TV (Buschow et al., 2014). For this research only one kind of TV show is used, which was a form of reality TV. It could be possible that the effect of the source of tweets (celebrity endorsement) on entertainment experience will be different for different TV genres. Future research should look at the possible effect of celebrity endorsement for different TV genres like animation, talk shows or talent shows.
In addition, for this research, not much personal information was collected from the respondents. Only some demographics like gender, age and educational level were asked. It could be possible that some personality traits could influence the effect of tweets on entertainment experience. For example, Tsao et al. (2015) studied the influence of hotel reviews on booking intentions, and they also took into account the individual’s level of conformity. Results of his research show that people with a high level of conformity showed higher booking intentions when exposed to positive reviews. For people with low conformity, the effects were less strong. For further research, it is thus worth looking into personality traits, like conformity level, as moderators for the relationship between valance of tweets and entertainment experience.
Furthermore, this research was the first one to look into the moderating role of celebrity endorsement on the effects of tweets on entertainment experiences. A small effect was found for the role of celebrity endorsement on the effect of tweets on enjoyment, but more research is needed to give this results more meaning. Because this was one of the first
29 studies to investigate the moderating role of celebrity endorsement in this context, a first orientating research has been conducted. Therefore, besides likeability – which did not influence the results - this research did not take into account any other possible control variables of celebrity endorsement. Research shows that variables like source credibility and attractiveness of the celerity are determining factors for the effect of celebrity endorsement (Amos et al., 2008). Future research should take into account variables like source credibility and source attractiveness to further develop the research on celebrity endorsement and Social TV.
Lastly, the sample of this study was quite small and not equally divided (gender, educational level, age), and consisted mostly of higher educated females. Therefore, this sample was not representative for the whole population. Further research in this field should use bigger and more representative samples to make more valid statements.
References
Amos, C., Holmes, G., & Strutton, D. (2008). Exploring the relationship between celebrity endorser effects and advertising effectiveness: A quantitative synthesis of effect size. International journal of advertising, 27(2), 209-234.
Asch, M. J. (1951). Nondirective teaching in psychology: An experimental study. Psychological Monographs: General and Applied, 65(4).
Atkin, C., & Block, M. (1983). Effectiveness of celebrity endorsers. Journal of advertising research.
Austin, E. W., Vord, R. V. D., Pinkleton, B. E., & Epstein, E. (2008). Celebrity endorsements and their potential to motivate young voters. Mass communication and society, 11(4), 420-436.
30 Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001). Bad is stronger than good. Review of General Psychology, 5(4), 323–370. doi:10.1037//1089-2680.5.4.323
Berry, M., & Schleser, M. (Eds.). (2014). Mobile media making in an age of smartphones. Springer.
Blumler, J. G., & Katz, E. (1974). The uses of mass communications: Current perspectives on gratifications research. Beverly Hills, CA: Sage.
Buschow, C., Schneider, B., & Ueberheide, S. (2014). Tweeting television: Exploring communication activities on Twitter while watching TV. European Journal of
Communication Research, 39(2), 129-149.
Busselle, R., & Bilandzic, H. (2009). Measuring narrative engagement. Media Psychology, 12(4), 321-347.
Cameron, J., & Geidner, N. (2014). Something old, something new, something borrowed from something blue: Experiments on dual viewing TV and Twitter. Journal of Broadcasting & Electronic Media, 58(3), 400-419.
Chevalier, J. A., & Mayzlin, D. (2006). The effect of word of mouth on sales: Online book reviews. Journal of marketing research, 43(3), 345-354.
Daugherty, T., & Hoffman, E. (2014). eWOM and the importance of capturing consumer attention within social media. Journal of Marketing Communications, 20(1–2), 82–102. doi:10.1080/13527266.2013.797764
Davis, C. J., Bowers, J. S., & Memon, A. (2011). Social influence in televised election debates: A potential distortion of democracy. PloS one, 6(3).
Deller, R. (2011). Twittering on: Audience research and participation using Twitter. Participations, 8(1), 216-45.
31 Dijksterhuis, A., & Aarts, H. (2003). On wildebeests and humans: The preferential detection of negative stimuli. Psychological Science, 14(1), 14–18. doi:10.1111/1467-9280.t01-1-01412
Fein, S., Goethals, G. R., & Kugler, M. B. (2007). Social influence on political judgments: The case of presidential debates. Political Psychology, 28(2), 165-192.
Friedman, H. H., Santeramo, M. J., & Traina, A. (1978). Correlates of trustworthiness for celebrities. Journal of the Academy of Marketing Science, 6(4), 291-299.
Garthwaite, C. and Moore, T.J., 2013. Can celebrity endorsements affect political outcomes? Evidence from the 2008 US Democratic presidential primary. The journal of Law, Economics & Organization, 29 (2), 355–384.
Giglietto, F., & Selva, D. (2014). Second screen and participation: A content analysis on a full season dataset of tweets. Journal of Communication, 64, 260–277.
Harrington, S., Highfield, T., & Bruns, A. (2013). More than a backchannel: Twitter and television. Participations: Journal of Audience & Reception Studies, 10, 405–409.
Hill, J. P., & Lynch, M. E. (1983). The intensification of gender-related role expectations during early adolescence. In Girls at puberty (pp. 201-228). Springer, Boston, MA.
Hovland, C. I., & Weiss, W. (1951). The influence of source credibility on communication effectiveness. Public opinion quarterly, 15(4), 635-650.
Ji, Q., & Raney, A. A. (2015). Morally judging entertainment: A case study of live tweeting during Downton Abbey. Media Psychology, 18(2), 221-242.
Katz, E., & Lazarsfeld, P. F. (1955). Personal influence: The part played by people in the flow of mass communications. Glencoe, IL: Free Press.
32 Laer, van, T., De Ruyter, K., Visconti, L. M., & Wetzels, M. (2013). The extended
transportation-imagery model: A meta-analysis of the antecedents and consequences of consumers' narrative transportation. Journal of Consumer research, 40(5), 797-817.
Larson, R. J., Woloshin, S., Schwartz, L. M., & Welch, H. G. (2005). Celebrity endorsements of cancer screening. Journal of the National Cancer Institute, 97(9), 693-695.
Larsson, A. O., & Moe, H. (2011). Studying political microblogging: Twitter users in the 2010 Swedish election campaign. New Media & Society, 14, 729–747.
Möller, A. M., & Kühne, R. (in press). The Effects of User Comments on Hedonic and Eudaimonic Entertainment Experiences When Watching Online Videos. Communications.
Möller, A. M., Kühne, R., Baumgartner, S. E., & Peter, J. (2018). Exploring user responses to entertainment and political videos: An automated content analysis of YouTube. Social
Science Computer Review, 0894439318779336.
Nabi, R. L., & Hendriks, A. (2003). The persuasive effect of host and audience reaction shots in television talk shows. Journal of Communication, 53(3), 527-543.
Nielsen. (2012). Nielsen’s state of the media: Advertising & audiences (part 2: by demographic). Retrieved from http://www.nielsen.com/us/en/newswire/2012/report-u-s-media-trends-bydemographic.html.
Ohanian, R. (1990). Construction and validation of a scale to measure celebrity endorsers' perceived expertise, trustworthiness, and attractiveness. Journal of advertising, 19(3), 39-52.
Oliver, M. B., & Raney, A. A. (2011). Entertainment as pleasurable and meaningful: Identifying hedonic and eudaimonic motivations for entertainment consumption. Journal of Communication, 61(5), 984-1004.
Petty, R. E., Cacioppo, J. T., & Schumann, D. (1983). Central and peripheral routes to
advertising effectiveness: The moderating role of involvement. Journal of consumer research, 10(2), 135-146.
33 Ramanathan, S., & McGill, A. L. (2007). Consuming with others: Social influences on
moment-to-moment and retrospective evaluations of an experience. Journal of Consumer Research, 34(4), 506-524.
Raney, A. A. (2003). Disposition-based theories of enjoyment. In J. Bryant, D. R. Roskos-Ewoldsen & J. Cantor (Eds.), Communication and emotion: Essays in honor of Dolf Zillmann (pp. 61–84). Mahwah, NJ: Erlbaum.
Rodkin, P. C. (2004). Peer ecologies of aggression and bullying. In D. L. Espelage & S. M. Swearer (Eds.), Bullying in American schools: A social-ecological perspective on prevention and intervention. Mahwah, NJ: Lawrence Erlbaum.
Russell, M. (2012). “Celebrities Get Twitter Buzzing for Papa John’s and Coke,”. Advertising Age(13).
Shedlosky-Shoemaker, R., Costabile, K. A., DeLuca, H. K., & Arkin, R. M. (2011). The social experience of entertainment media. Journal of Media Psychology.
Smith, A., & Boyles, J. L. (2012). The rise of the ‘‘connected viewer.’’ (Pew Research Center’s Internet and American Life Project). Retrieved from
http://www.pewinternet.org/_/media//Files/Reports/2012/PIP_Connected_Viewers.pdf.
Tan, E. S. H. (2008). Entertainment is emotion: The functional architecture of the entertainment experience. Media psychology, 11(1), 28-51.
Trussler, M., & Soroka, S. (2014). Consumer demand for cynical and negative news Ffames. The International Journal of Press/Politics, 19(3), 360–379. doi:10.1177/1940161214524832
Tsao, W. C., Hsieh, M. T., Shih, L. W., & Lin, T. M. (2015). Compliance with eWOM: The influence of hotel reviews on booking intention from the perspective of consumer conformity. International Journal of Hospitality Management, 46, 99-111.
34 Vorderer, P., Klimmt, C., & Ritterfeld, U. (2004). Enjoyment: At the heart of media
entertainment. Communication theory, 14(4), 388-408.
Waddell, T. F., & Sundar, S. S. (2017). # thisshowsucks! The overpowering influence of negative social media comments on television viewers. Journal of Broadcasting & Electronic Media, 61(2), 393-409.
Weaver, J. B., Huck, I., & Brosius, H. B. (2009). Biasing public opinion: Computerized continuous response measurement displays impact viewers’ perceptions of media messages. Computers in Human Behavior, 25, 50–55.
Whiting, A., & Williams, D. (2013). Why people use social media: a uses and gratifications approach. Qualitative Market Research: An International Journal, 16(4), 362-369.
Wirth, W., Hofer, M., & Schramm, H. (2012). Beyond pleasure: Exploring the eudaimonic entertainment experience. Human Communication Research, 38(4), 406-428.
Wohn, D. Y., & Na, E. K. (2011). Tweeting about TV: Sharing television viewing experiences via social media message streams. First Monday, 16(3).
Attachment
Attachment 1: manipulated tweets35 Condition 2 - Positive_celebrity
36 Condition 4 – Positive_unknown
