Through content analysis of messages posted on Twitter, we categorize the types of content into a matrix — attention, emotion, information, and opinion. We use this matrix to analyze televised political and entertainment programs, finding that different types of messages are salient for different types of programs, and that the frequencies of the types correspond with the program content. Our analyses suggest that Twitter picks up where formal social television systems failed: people are using the tool to selectively seek others who have similar interests and communicate their thoughts synchronous with television viewing.
The evolving role of television
Overview of social television systems
Sharing viewing experiences via social network sites
Data and methods
Discussion and conclusion
Television has traditionally been an important facilitator for social interaction and a popular source of conversation. Scholars have referred to television as a “cultural forum” where people discuss a broad variety of topics (Newcomb, 1994) or an “electronic hearth” that brought people together (Tichi, 1991). Scholars have also found social motivations as distinct factors in uses and gratifications research (Palmgreen and Rayburn, 1979; Rubin, 1983).
Previous research on social television has focused on defining social television in the broad sense — any technology that supports social practices (Harboe, 2009). Thus, earlier research of social television viewing look at two elements — television content as a conversation trigger and television as a medium that physically connects two or more people into a confined space (Morrison and Krugman, 2001).
Social network sites (SNSs), however, allow television viewers to once again enjoy the communal experience of group viewing without being physically together. This study looks at a new practice of television viewing that combines these two elements: viewers are sharing their viewing experiences real–time through computer–mediated communication, which creates a pseudo–communal viewing experience even though they are not collocated.
Examining messages on Twitter, a popular social network site, this study applies a uses and gratifications framework to explore what types of messages people share with others while they are watching television and how those messages correspond to the context of the program they are watching. We propose that this AEIO (Attention, Emotion, Information, Opinion) model is an analytical model for mapping type and flow of content on social media message streams.
Television has had a profound influence on our daily routines since the 1950s (Marling, 1996). In post–World War II America, television was considered a new commodity that joined families together, shaping the middle–class suburbia ideal (Spigel, 1992). Studies conducted in late 1940s and early 1950s showed that Americans believed television would revive domestic life by keeping the family at home and even curing marital problems (Bogart, 1956).
As more people could afford television sets, however, studies began showing how television viewing was moving from a communal to individual experience. According to Nielsen’s 2009 Television Audience Report, only 11 percent of the average U.S. home had three or more television sets in 1975, but 54 percent of homes had three or more TVs in 2009. Likewise, in 1975, 57 percent of homes had one television, but in 2009, the percentage dropped to 18 percent.
Much of the research over the past 20 years on the effects of television viewing has focused on negative effects. The prevalence of television and the impact of its content have raised many alarming messages from the scholarly community. Television has been blamed to foster disengagement from reality and cognitive disorder (Gerbner, et al., 1986; Signorielli, et al., 1995; Newhagen and Reeves, 1992). Academics have also described television viewing as a “fundamentally isolating experience — one that gives the illusion of contact with the world while discouraging actual human contact” (Hoynes, 1994) — as well as an experience that fosters disengagement from reality and threatens formation of self–identity due to the lack of social interaction (Davis, 2000).
Putnam (1995a) pointed out that advanced technology enables individual tastes to be satisfied more fully but “at the cost of the social gratification associated with more primitive forms of entertainment.” He blamed American television for the erosion of social capital and civic engagement (Putnam, 1995b), reasoning that television viewing causes less face–to–face connections with other people. These face–to–face connections are the basis of trust in a community, which in turn builds social capital. Based on analysis of the General Social Survey data from 1974–94, he pointed out that television viewing was strongly and negatively related to social trust, group membership, and voting turnout .
Putnam was certainly not alone in his ideas. Bugeja (2005) expressed concerns about how media and technology created a social gap, eroding our sense of community and creating an “interpersonal divide.” Putnam’s arguments that were later echoed in studies of Internet and mobile phones (Kraut, et al., 1998; Nie, 2001).
Technology, however, is a double–edged sword, and while we do not want to argue that it can create an individual experience, we also want to point out that it can also be a social experience. The Internet supports interactions with a greater number of people who are spatially dispersed (Boase and Wellman, 2004) and social network sites, in particular, enable people to communicate with extended networks (Haythornthwaite, 2005).
Also, the concept of television has been evolved. Television is no longer a rectangular “box” that plays video images transmitted through airborne broadcasting waves: although there is a considerable population that watches TV from the comfort of the couch, an increasing number of people are watching television on their computers through Internet connection or on mobile devices. Now, viewers have the option to watch any time they want, anywhere they want. According to Nielsen’s 2009 Three Screen Report, 131 million Americans watch video on the Internet, a 13 percent increase from 2008, and 13.4 million Americans watch videos on mobile phones, a 4.9 percent increase from the previous year. These new forms of viewing are heavily linked with social media, which extends the viewing experience, enabling a viewer to engage in conversation or other forms of interaction with his or her social networks.
Studies conducted in the past decade that examine the social aspects of television define social television as a technical construct — a machine that facilitates interactivity. The studies are mainly based on experiments conducted in lab settings with prototypes that were created for the purpose of experimentation. Companies and academic institutes created their own interactive television systems; some of which include 2BeOn at the University of Aveiro (Abreu, et al., 2001), Amigo TV at Alcatel (Coppens, et al., 2004), Media Center Buddies at Microsoft Labs (Regan and Todd, 2004), ConnecTV at TNO (Boertjes, et al., 2008), and STV at Motorola (Harboe, et al., 2008). Other studies have looked at the mechanics of shared viewing (Ducheneaut, et al., 2008) in a simulated environment, seeing differences between viewers who were watching in the same location and viewers watching from different locations but were connected with live audio. None of the systems used in these experiments are available to the public.
One of the reasons researchers created these hypothetical settings was due to the fact that interactive features embedded into television sets have not been commercially successful. For a long time, companies tried to tie in interactive features to the medium itself and failed. Time Warner experimented and failed in the 1970s with an interactive cable system called Qube (Wolf and Latane, 1981); another similar attempt in the early 1990s was also unsuccessful. In 1980, Zenith launched Spacephone, a high–end television model that enabled users to talk on the telephone through their television set (Harboe, 2009). Users could make phone calls with their remote control and watch television while talking, but the product was not successful and discontinued after a few years. America On–Line (AOL) launched AOLTV in 2000, which enabled users to surf the Internet, read e–mail, and chat while watching television on the same screen (Harboe, 2009). The service, however fell off the market in 2002.
Although early efforts at creating interactive television sets were made in the United States, commercial interactive television viewing services started gaining momentum at the turn of the century in Asia and Europe along with mainstream adoption of the Internet, blossoming in countries that quickly achieved nationwide broadband coverage. In Europe, many countries launched SMS TV Chat around 2000 (Harboe, 2009). This service allowed people to chat like they would in an online chat room using their cell phones to post messages; the messages would then appear on the television set. The service was hugely popular, although it was flawed in that the messages were displayed in the same interface as the program being broadcast — if the volume of messages went up, it would be difficult to view the actual program.
In the United States, commercial interactive TV services did not catch on until very recently. In 2006, ABC.com was the first network Web site to offer full–length episodes online for free, to be followed quickly by other networks such as CBS, Fox, and joint ventures such as Hulu. These services enabled viewers to comment on programs while they were watching them. Although television discussion forums had existed for a long time on the Internet, these sites were unique in that the discussion could take place on the same page as the video. In 2009, networks have also begun to support live streaming services and synchronous messaging. The U.S. Open, for example hosted live videos of the tournament on its Web site for the first time in 2009. Viewers could watch the video and engage in live chatting. In late 2009, TV.com introduced chatting services so that people watching the same program can engage in synchronous messaging. However, live content that is available at the same time as the original network airing is still extremely limited within the United States.
Although companies have been slow to provide television sets that facilitate conversation, television viewers have found ways to engage in conversation by repurposing social network sites. Social network sites are Web–based services that allow individuals to construct a public or semi–public profile, create a list of others users with whom they share a connection, and view and traverse their list of connections and those made by others within the system (boyd and Ellison, 2008). Services such as MySpace and Facebook allow people to share favorite TV show lists; new applications introduced in the second half of 2009 enable users to watch streaming videos within the sites. There are also 3D virtual worlds such as Second Life and YouTube3D where users can actually create an avatar and “sit with” other users in a rendered environment to view live multimedia as well. These 3D venues, however, are not mainstream.
Among social network sites, Twitter is a microblogging service launched by a San Francisco–based start–up in 2006 that allows registered users to publish short messages. Twitter poses the question: “What’s happening?” and allows users to answer this question, but users do not always post messages that are answers to this question [Before November 2009, Twitter’s question was “What are you doing?”].
We used a uses and gratifications (U&G) perspective as a theoretical framework to guide our research questions. This approach looks at media use as a choice that consumers (users) make to fulfill certain individual needs (see Katz, et al., 1974). Katz, et al. (1973) explain U&G as a cyclical function, where individual needs and socio–psychological factors create different expectations, which leads to different patterns of use and ultimately different gratifications, which then feeds back into individual needs. Much of the television research using U&G, however, has focused the first two functions, looking at how motivations and individual psychological factors lead to different usage (e.g., Rubin, 1983; Rubin, et al., 1985).
While not designed as a television discussion forum service per se, Twitter is an ideal venue to view how people express their uses and gratifications of their television viewing behavior. If we were to examine messages that were posted during the time when the television program is first aired, we could get a glimpse into what immediate reactions of audiences, which may reflect the nature of their use and perhaps even their short–term gratifications. At the same time, Twitter itself is a media, so the gratifications of television may be the motivation that influences the use of Twitter. To our knowledge, there are no studies that examine this chain effect of media use from a U&G perspective.
Twitter users have developed methods of directing messages at specific users and developing mechanisms to search for messages of similar topics. We noticed that television programs were often a Trending Topic on Twitter; Trending Topic is a service that Twitter offers, showing current popular topics on Twitter. The fact that the Trending Topic on a specific television program occurred at the time of its network airing gave us anecdotal evidence that people were Tweeting while they were watching the television program, or if they were not watching it, at least communicating about the program while others are watching it. Trying to examine the uses and gratifications of viewers, we formed the following research question:
RQ1: What types of messages are people posting while watching TV?
This research question would give us an idea of how people are expressing themselves or what they are doing in reaction to the content that they are viewing. However, the different typologies alone are not indicative of whether or not the messages are a direct response, reaction, or reflection of the actual program content. That led us to our second research question:
RQ2: Do these messages correspond to the real–time context of the program?
Our third research question examines whether or not people who post message on Twitter are just posting or actually interacting with other people. Desire for social interaction has been identified across literature of numerous different types of media as one of the major motivations to use the media (e.g., Joinson, 2008; Papacharissi and Rubin, 2000; Rubin, 1983). However, in terms of television viewing, are these social motivations fulfilled by watching television or in subsequent discussions?
RQ3: Do people posting on Twitter engage in conversation with other viewers?
We also noticed that many people were using hashtags, links, and other linguistic mechanisms to relay their message and that many were posting their Tweets through mobile devices. This led us to our final research question:
RQ4: What kind of Utility functions (technical features or linguistic features) do people use to share their television–viewing experience?
We analyzed messages on Twitter, which are called Tweets, for two programs that were televised nationwide in the United States — a live political event and an entertainment/show event. The political event was President Barack Obama’s live speech at the White House announcing his acceptance of the Nobel Peace Prize (NPP) on 9 October 2009. The speech was not prescheduled and interrupted major network programming. The second event was an entertainment event: the 7 October airing of an episode of ABC’s So You Think You Can Dance (SYTYCD), a reality/show program similar to the American Idol series where a panel of judges holds dance auditions nationwide and narrow down contestants who have to perform different dance styles with different partners each week to prove their skill and versatility. The episode was a regional audition taking place in Las Vegas.
Although we understand that many people do not watch television programs real–time, we chose to analyze only the messages that were posted during the broadcast time in order to determine if the messages corresponded to the context of the event and to look at messages that were intended to be posted at that specific time. We used Eastern Standard Time (EST) because it was the first airing of the program, in the case of SYTYCD. One hour after the episode airs on EST, it delivers the same content for Central Time Zone viewers. We wouldn’t be able to determine, then, whether or not the tweets were coming from live viewers in Central Time Zone, or viewers in EST who wanted to discuss about the show after they viewed it. The Obama speech was a live broadcast, thus we did not have time zone problems.
These two programs were chosen because of the difference in their content and the fact that the subject matter is stimulating and has potential for discussion. Both programs had very vibrant Tweets while they were being televised. During the Nobel Peace Prize acceptance speech, “President Obama” and “NPP” were Trending Topics on Twitter. SYTYCD was also a Trending Topic at the time the program was televised.
We did not know, however, the algorithm in which Twitter selected messages for its Trending Topics, so we used Twitter’s search API. We used the key words “Obama,” “Nobel,” “NPP,” and “SYTYCD” as key words in Twitter’s search API to filter related messages. Messages posted on Twitter that contained these keywords were documented real–time since the Twitter API has limitations on how far back one could collect data. Screenshots were also taken as a backup measure. We began collecting messages several minutes before the actual event, during the event, and several minutes after the event. All messages that were posted during the program that contained these keywords we designated were collected. We only archived messages that were in English and were completely open to the public, regardless of whether or not the viewer has a Twitter account. We felt comfortable using this data because messages posted on a public Internet forum are deliberately intended for public viewing and when creating a Twitter account, Twitter users acknowledge that their information may be viewed by third parties. We didn’t think a consent form was necessary because we weren’t doing harm to individual users or the social group as a whole (Eysenbach and Till, 2001) and Twitter has very low levels of “perceived privacy” (King, 1996) due to its high number of users.
Our unit of analysis was one Tweet. A Tweet is a message that is 140 characters or less. Tweets may range from one word to incomplete phrases, hyperlinks, emoticons, complete sentences, and combinations of these elements. The messages were then coded for the time it was posted, number of hashtags, and type of message.
To see whether the messages were independent Tweets or interactive discussions, we adapted Henri’s (1992) concept of interactivity. Henri’s model consists of three steps: communication of information, response to this information, and a second response to the first information. She then identified messages as “explicit interaction,” “implicit interaction” and “independent statement.” We found it difficult, however, to assess implicit interactions. An explicit interaction is an obvious two–way communication whereas implicit interaction is one that contains content generated by someone else but does not attribute it. Therefore we decided to categorize messages according to the directionality and target audience. That gave us three types: (1) messages that were sent to an at–large audience; (2) messages that were directed at a specific user; and, (3) messages directed at a specific user that were reciprocated.
In Twitter, users can choose to direct a message at a specific user by adding “@” in front of the recipient’s username (Honeycutt and Herring, 2009). For instance, if Apple wanted to respond to a message posted by Banana, the message would be something like: “@Banana I totally agree with you.” The use of “@” however does not indicate that a discussion took place. For instance, Apple may send Banana a message, but Banana may not respond. We therefore counted how many messages contained “@” but noted whether they were one–directional messages were reciprocated ones.
Re–tweets are messages that duplicate a message another person has written. These messages are identified by “RT @ID of original message.” For instance, “RT @arcticpenguin I love school” indicates that a user with the ID of “arcticpenguin” posted “I love school” and that another person is re–posting this message. Re–tweets were coded as being mutually exclusive from “@” messages because the messages are a repeat of the person identified in the message, not directed at the person identified in the message.
Hashtags are words that are preceded with the “#” sign. Twitter users use hashtags to make searching for relevant topics easier. For instance, using the word “lost” to search for messages about the television program “Lost” would be a bad search mechanism because the search engine would bring up all messages that included the word “lost.” Using “#Lost” instead of “Lost” enables people to create specific threads of conversation.
Messages were also coded for the type of medium used to post — at the end of the Tweet, Twitter provides the name of the platform that was used to post the message. We then categorized these into three types — Web, text message via cell phone, or mobile phone application. Applications that were available on two platforms were noted as such.
Tweet types — Forming an AEIO matrix
Past studies of conversations that groups have while watching television categorized those conversations into five types: content–based, context–based, logistical, non sequitur, and phatic (Ducheneaut, et al., 2008). This typology, however, could not be applied to Twitter messages because the message posters were not collocated and any non sequitur messages (messages that have no relation to the television program whatsoever) would not show up in our keyword search. We wanted to create a typology that would explain all types of messages on social media message streams regardless of the context of the person posting the message.
A large portion of our research involved developing a Matrix that would categorize the types of content to address RQ1 (What types of messages are people posting while watching TV?). We decided to analyze the type of content depending on two criteria: whether the message is subjective or objective, and whether the message is inbound (about oneself/the Tweeter) or outbound (not about oneself — in the case of television, this would be about the television program). This created a 2 x 2 table resulting in four different types of messages: Attention–seeking (an objective message about oneself), Information (an objective message about the program), Emotion (a subjective message about oneself), and Opinion (a subjective message about the program).
Computer–mediated communication notes the importance of social messages (Rice and Love, 1987), which can be defined as a “statement or part of a statement not related to the formal content of subject matter” (Henri, 1992). This was our basis for creating the dichotomy of inbound and outbound.
To reduce ambiguity, we created a strict protocol that included a list of key words or phrases that we would assign to the different categories. Emotional content, for instance, contained verbs such as “love, hate, hope, excited, congratulations … .” Messages that could be seen as “information” but were written entirely in capital letters, with multiple explanation marks, or emoticons were also categorized as emotional content. Attention–seeking messages contained phrases such as “I wonder …” and questions explicitly soliciting response such as “Do you think …,” “Can anyone tell me …?” Information content consisted of dry, objective content about the program (“Obama is delivering a speech.”) but also included posts that had links to articles or blogs. In the case of links to blogs, the blog may be an opinion, but the act of the Tweeter in sharing the information of the site categorized that as “information” in our protocol. A link was not considered as being the primary message unless the entire message only consisted of a link.
Although most Tweets clearly fell into one category, there were a few that dealt with multiple ideas. These were coded for the most salient message. For instance, “Obama is giving a speech. I agree with what he says,” could be Information (Obama is giving a speech) or Opinion (I agree with what he says), but as the former is being presented as contextual (dependent) information for the latter, the main message was coded as Opinion. We did not separate this message and make it two messages because we wanted to see the frequency patterns of types of message, and assigning two types to one message would not create distinct patterns.
A total of 1,307 messages were analyzed for President Obama’s Nobel Prize speech. We had two coders that had a Cohen’s Kappa of .95. Conflicting items were reviewed together to reach an agreement. We started collecting the messages for the Obama speech 11 minutes before it started and ended collection three minutes after it finished. Obama messages were almost equally distributed between inbound and outbound messages, but showed a strong orientation towards subjective messages (61.1 percent) over objective messages (38.9 percent). There were more outbound (61.7 percent) than inbound (38.3 percent) messages. Attention–seeking messages were 15.53 percent, Emotion messages were 22.8 percent, Information messages were 23.3 percent, and Opinion messages were 38.3 percent.
Ja#####: So did President Obama basically say he’ll accept the Nobel Peace Prize in advance of what he hopes to accomplish in the coming years? (Attention–seeking)
Opp####R: Obama was the nobel? I’m sorry, but what. The. Fuck. Did I miss something or has everyone absolutely gone fucking crazy? (Emotion)
Ma###er: Obama said Prize is not a recognition of his accomplishements — but an affirmation of American leadership on behalf of global aspirations. (Information)
Sou######ce: Noble peace prize for a President Obama? So soon..too early for his image..for him to be reflected as a Savior..he hasn’t done much (Opinion)
A total of 1,012 messages were collected for So You Think You Can Dance (SYTYCD). Two coders had an inter–coder reliability of above .85 (Cohen’s Kappa) for every variable. Conflicting items were reviewed together to reach an agreement. This program also showed more subjective (60.7 percent) messages over objective (39.3 percent) messages, but slightly more inbound messages (50.9 percent) than outbound messages (49.1 percent). Attention messages (22.1 percent) for SYTYCD were noticeably higher than the percentage of attention messages in Obama messages (16 percent). Emotion messages were 28.7 percent, Information messages were 17.2 percent and Opinion messages were strongest at 31.9 percent. (see Figure 1) These are some examples of the messages in each of the categories:
Sar###gel: Finally sitting down to watch #SYTYCD (Attention)
Jan###ee: sytycd makes me madddd! he wass sooo good :( (Emotion)
Pa###upe: SYTYCD semi finals begin and glee (Information)
The###o: dude one of the dudes on SYTYCD looks like travis wall FREAKY (Opinion)
Figure 1: Comparison of AEIO ratio in Obama and SYTYCD.
To address RQ2 (Do types of messages correspond to real–time context of the program?), we mapped out how many messages of each type were posted in the given timeframe to examine whether the message patterns correlated to the content of the program. The Obama speech, being only seven minutes long, was analyzed by the minute. SYTYCD was analyzed in increments of four minutes.
The graph (see Figure 2) indicates the overall flow of messages; we can easily identify opinion messages as being the most salient. It also shows the traffic of messages at a given time. There was a brief lull just before Obama’s speech started at 11:16 a.m. and peaked at 11:18 after he said he would accept the award as a “as a call to action”. Some messages that were posted at this time were:
By###en: RT @BreakingNews: NOBEL PEACE PRIZE –– President Obama: I will accept this award as a call for action.
Q###n: Obama said, “I do not feel that I deserve to be in the company of so many transformative figures that have been honored by this prize.”
Figure 2: Time–mapping of AEIO for Obama.
The second peak came when he finished his speech at 11:23 a.m., when attention, emotion, and opinion messages shot up again.
For SYTYCD, we saw a spike in information messages before the program began airing at 8 p.m. EST, followed by attention messages where viewers were publishing to the world that they were about to watch the program (see Figure 3):
Ag###ny: SYTYCD, then Glee. See you in two hours, text buddies.
gr#r#p: One minute from #SYTYCD with @megfowler. This is a first, people!
mui##27: its time to watch sytycd!!!!!
Figure 3: Time–mapping of AEIO for SYTYCD.
In the case of SYTYCD, the fluctuation of AEIO messages correlated with commercials. For instance, we saw a surge of emotion, attention, and opinion messages and a drop in information messages at approximately 8:23, 8:33, 8:42, and 8:49 which were commercial breaks.
Viewers also shared live updates (information) about the content of the program. At 8:30 we see a high peak in opinion and emotion, which was when one dancer had to be led off the stage bleeding due to an accident. Opinion and emotion messages also went up when the judges were making decisions about whether or not a contestant would make it to the next level. The peak that we see at 8:48 was right after Ryan, one of the contestants, was voted out of the competition.
Bu###be: So sad that Ryan got cut on SYTYCD tonight. But I understood what Nigel said was true. Still sad though.
Om####33: watching sytycd. billy bell just got a nose bleed. he’s my favorite. and too bad for ryan. :(
E###LP: Sad to see Ryan K. (Evan’s bro) leave on SYTYCD. His talent is immeasurable. They need to have him perform on a results show or the finale.
Mobility was a high component in our television tweets, supporting recent reports by the Pew Internet and American Life Project that Twitter users are a “mobile bunch” — a group that is more likely to be using wireless technologies (Lenhart and Fox, 2009). In Obama, we saw that 29 percent of users used mobile devices to post their message. The top eight mobile applications that were used were: Echofon, UberTwitter, Tweetie, Txt (text messaging), Mobile Web (for smart phones), Twitterrific, and Twitter Berry. Tweetdeck, an application for both desktop and mobile, was categorized as non–mobile devices, so the percentage of mobile users may have been even bigger.
SYTYCD Tweeters showed a higher mobile rate of 39 percent. Again, Tweetdeck was categorized as non–mobile devices, so the percentage of mobile users may be even bigger. Some explanatiosn for the higher mobile use for SYTYCD could be that the average age of viewers is lower or that users tweet from their couch. Another explanation could be that given that the Obama speech was delivered on a weekday during the day, many people were watching from their office and thus being at their desk warranted less use of mobile technology.
Television viewers were not very interactive: interactive tweets made up of less than four percent. The mapping of hashtags, links, and re–tweets according to time showed similar frequency patterns to the patterns of the message types (see Figure 4 and Figure 5). The use of hashtags suggests that people want to share their message with a group of people who have the same interest.
Figure 4: Time–mapping of links, hashtags, and re–tweets for Obama.
Figure 5: Time–mapping of links, hashtags, and re–tweets for SYTYCD.
Social media is recreating a pseudo “group viewing” experience of television. Tweets indicate that people are using Twitter to express themselves. Although television viewers aren’t communicating directly with each other while they are viewing, the use of hashtags and re–tweets suggests that although users aren’t directly interacting with specific individuals, they want to be part of a larger group.
This study was mainly exploratory in trying to look at what type of messages people post on Twitter when they are watching television. Using a uses and gratifications framework, we conceptualized Tweets as both a means of gratification (for television) as well as a type of use (for Twitter). Based on two axes of subjectivity/ objectivity and inbound (about oneself)/ outbound (about the TV program), we were able to create a matrix that categorized messages into four types: Attention, Emotion, Information and Opinion. We found that the frequencies of the message types strongly correlated with the content of the program. This does not indicate a causal relationship because there are so many other variables that could affect the type of messages people posted, but has strong implication that people post in reaction to what they see. Of particular note, we saw very little time lapse between a situation that occurred on television and a message responding to that situation, suggesting that some people Tweet without giving much thought to what they are Tweeting about. People were also Tweeting more during commercial breaks. One explanation could be that the storyline tends to get more dramatic just before the commercial break, which causes people to engage in livelier discussion. Another explanation could be that viewers are bored and Tweet to pass time. Future studies should look more closely at motivations of Tweeting.
The frequencies of the messages may also reflect to some extent individual engagement in program content. For instance, after Obama concluded his Nobel Peace Prize acceptance speech, emotional and attention messages decreased while opinion messages steeply increased. This could imply that after a political event, people engage in more opinionated discussion. In the case of SYTYCD, messages sharply decreased after the program.
We found that utility was very important in social media messages. People were using various linguistic and technical tools to convey their message. For instance, hashtags, links, re–tweets, and “@” messages all served as tools that people used to interact with each other. Mobile phones also served as an important utility; at least 30 percent of Twitter uses were tweeting from a mobile device.
Since we only looked at messages on Twitter in the context of television, we cannot generalize that the AEIO Matrix would apply to other social media content although we have no reason to believe it would not. Also, in the case of both Obama and SYTYCD, the behaviors that we observed were those of people who have average or above average interest in those topics. We do not claim that the characteristics of these Tweeters represent the general television viewer audience; they could be extreme fans with entirely different patterns from general viewers.
We cannot claim that the messages we sampled were those of people who were actually watching the program; however, the contents of the tweets were fairly specific and given that the data was collected real–time, it was unlikely that comments that was directly related to specific content on TV were being made by those who were not watching. Still, this limitation should be taken into consideration in designing future studies.
One of the biggest limitations of this research is that manual coding is extremely dependent on labor and time resources. Although we think this analysis is more valid than linguistic word counts, which doesn’t detect sarcasm or other types of speech that would be undetectable by a computer, future studies should examine if this typology could be coded into a computer algorithm, which would be useful in tracking frequencies of discussions on certain topics. Limitations of this study also include the fact that we only examined two types of programs: futures studies should see if any frequency patterns can be found in other genres such as news, sports, or television dramas, which most likely will have different patterns in terms of how audiences react, reflect, or respond.
It would also be interesting to see whether or not the type of messages people post on Twitter could be used to evaluate television programs, or predict the program’s ratings. From the diversity and frankness of the messages that we saw, television producers could at least use Twitter as means of quick feedback. Although how representative those messages would be of the general viewer audience is questionable, one could certainly gauge an idea of what enthusiasts of the program are feeling. By looking at the quantity of posts at a given time (which is very easy to measure) and then analyzing the qualitative content of the posts that happen during a surge of tweets, one could start to see patterns of what viewers are interested in.
About the authors
D. Yvette Wohn is a Ph.D. student at Michigan State University in Media and Information Studies. She is interested in new media effects, with a specific interest in prosocial outcomes of social networks. Other projects include socio–economic behavior in virtual worlds and media continuance.
E–mail: yvettewohn [at] gmail [dot] com
Eun–Kyung Na is a Ph.D. candidate at Keio University and associate at Samsung Electronics’ Media Solution Center. She is interested in social media content and business implications.
E–mail: naccy80 [at] gmail [dot] com
Thanks to First Monday’s anonymous reviewers and session participants of the International Communication Association’s 2010 conference in Singapore for constructive feedback.
1. Putnam, 1995b, p. 678.
J. Abreu, P. Almeida, and V. Branco, 2001. “2BeOn: Interactive television supporting interpersonal communication,” In: J.A. Jorge, N.M. Correia, H. Jones, and M.B. Kamegai (editors). Multimedia 2001: Proceedings of the Eurographics Workshop in Manchester, United Kingdom, September 8–9, 2001, Vienna: Springer, pp. 199–215.
J. Boase and B. Wellman, 2006. “Personal relationships: On and off the Internet,” In: A. Vangelisti and D. Perlman (editors). The Cambridge handbook of personal relationships. New York: Cambridge University Press, pp. 709–723.
E. Boertjes, J. Klok, and S. Schultz, 2008. “ConnecTV: Results of the field trial,” Proceedings of EuroITV 2008. Tampere, Finland: Tampere University of Technology, at https://soc.kuleuven.be/com/mediac/socialitv2/papers/ConnecTV_Results_of_the_Field_Trial.pdf, accessed 14 February 2011.
L. Bogart, 1956. The age of television: A study of viewing habits and the impact of television on American life. New York: F. Ungar.
d. boyd and N. Ellison, 2008. “Social network sites: Definition, history, and scholarship,” Journal of Computer–Mediated Communication, volume 13, number 1, pp. 210–230.
M. Bugeja, 2005. Interpersonal divide: The search for community in a technological age. New York: Oxford University Press.
T. Coppens, L. Trappeniers, and M. Godon, 2004. “AmigoTV: Towards a social TV experience,” Proceedings of EuroITV 2004 (Brighton, U.K.), at http://www.csd.abdn.ac.uk/~jmasthof/EuroITV04/S04.pdf, accessed 14 February 2011.
J. Davis, 2000. Identity and social change. New Brunswick, N.J.: Transaction.
N. Ducheneaut, R. Moore, L. Oehlberg, J. Thornton, and E. Nickell, 2008. “Social TV: Designing for distributed, sociable television viewing,” International Journal of Human–Computer Interaction, volume 24, number 2, pp. 136–154.http://dx.doi.org/10.1080/10447310701821426
G. Eysenbach and J. Till, 2001. “Ethical issues in qualitative research on Internet communities,” British Medical Journal, volume 323, number 7321, pp. 1,103–1,105, and at http://www.bmj.com/content/323/7321/1103.full, accessed 14 February 2011.
G. Gerbner, L. Gross, M. Morgan, and N. Signorielli, 1986. “Living with television: The dynamics of the cultivation process,” In: J. Bryant and D. Zillman (editors). Perspectives on media effects. Hilldale, N.J.: L. Erlbaum Associates pp. 17–40.
G. Harboe, 2009. “In search of social television,” In: P. Cesar, D. Geerts, and K. Chorianopoulos (editors). Social interactive television: Immersive shared experiences and perspectives. Hershey, Pa.: Information Science Reference, pp. 1–13.
G. Harboe, N. Massey, C. Metcalf, D. Wheatley, and G. Romano, 2008. “The uses of social television,” Computers in Entertainment, volume 6, number 1, pp. 1–15.http://dx.doi.org/10.1145/1350843.1350851
C. Haythornthwaite, 2005. “Social networks and Internet connectivity effects,” Information, Communication & Society, volume 8, number 2, pp. 125–147.http://dx.doi.org/10.1080/13691180500146185
F. Henri, 1992. “Computer conferencing and content analysis,” In: A. Kaye (editor). Collaborative learning through computer conferencing: The Najaden papers. New York: Springer, pp. 117–136.
C. Honeycutt and S. Herring, 2009. “Beyond microblogging: Conversation and collaboration via Twitter,” HICSS ’09: Proceedings of the 42nd Hawaii International Conference on System Sciences, pp. 1–10.
W. Hoynes, 1994. Public television for sale: Media, the market, and the public sphere. Boulder, Colo.: Westview Press.
A. Joinson, 2008. “Looking at, looking up, or keeping up with people? Motives and use of Facebook,” CHI ’08: Proceedings of the 26th Annual SIGCHI Conference on Human Factors in Computing Systems, pp. 1,027–1,036.
E. Katz, J. Blumler, and M. Gurevitch, 1974. “Utilization of mass communication by the individual,” In: J. Blumler and E. Katz (editors). The uses of mass communication: Current perspectives on gratifications research. Beverly Hills, Calif.: Sage, pp. 19–32.
E. Katz, H. Haas, and M. Gurevitch, 1973. “On the use of the mass media for important things,” American Sociological Review, volume 38, number 2, pp. 164–181.http://dx.doi.org/10.2307/2094393
S. King, 1996. “Researching Internet communities: Proposed ethical guidelines for the reporting of the results,” Information Society, volume 12, number 2, pp. 119–127.http://dx.doi.org/10.1080/713856145
R. Kraut, M. Patterson, V. Lundmark, S. Kiesler, T. Mukopadhyay, and W. Scherlis, 1998. “Internet paradox: A social technology that reduces social involvement and psychological well–being?” American Psychologist, volume 53, number 9, pp. 1,017–1,031.
A. Lenhart and S. Fox, 2009. “Twitter and status updating,” Pew Internet & American Life Project (12 February), at http://www.pewinternet.org/Reports/2009/Twitter-and-status-updating/Part-1.aspx, accessed 2 January 2011.
K. Marling, 1996. As seen on TV: The visual culture of everyday life in the 1950s. Cambridge, Mass.: Harvard University Press.
M. Morrison and D. Krugman 2001. “A look at mass and computer mediated technologies: Understanding the roles of television and computers in the home,” Journal of Broadcasting & Electronic Media, volume 45, number 1, pp. 135–161.http://dx.doi.org/10.1207/s15506878jobem4501_9
H. Newcomb, 1994. Television: The critical view. Fifth edition. New York: Oxford University Press.
J. Newhagen and B. Reeves, 1992. “The evening’s bad news: Effects of compelling negative television news images on memory,” Journal of Communication, volume 42, number 2, pp. 25–41.http://dx.doi.org/10.1111/j.1460-2466.1992.tb00776.x
N. Nie, 2001, “Sociability, interpersonal relations, and the Internet: Reconciling conflicting findings,” American Behavioral Scientist, volume 45, number 3, pp. 420–435.http://dx.doi.org/10.1177/00027640121957277
P. Palmgreen and J. Rayburn, 1979. “Uses and gratifications and exposure to public television: A discrepancy approach,” Communication Research, volume 6, number 2, pp. 155–179.http://dx.doi.org/10.1177/009365027900600203
Z. Papacharissi and A. Rubin, 2000. “Predictors of Internet use,” Journal of Broadcasting & Electronic Media, volume 44, number 2, pp. 175–196.http://dx.doi.org/10.1207/s15506878jobem4402_2
R. Putnam, 1995a. “Bowling alone: America’s declining social capital,” Journal of Democracy, volume 6, number 1, pp. 65–78.http://dx.doi.org/10.1353/jod.1995.0002
R. Putnam, 1995b. “Tuning in, tuning out: The strange disappearance of social capital in America,” PS: Political Science and Politics, volume 28, number 4, pp. 664–683.
T. Regan and I. Todd, 2004. “Media center buddies: Instant messaging around a media center,” NordiCHI ’04: Proceedings of the Third Nordic Conference on Human–computer Interaction, pp. 141–144.
R. Rice and G. Love, 1987. “Electronic emotion: Socioemotional content in a computer-mediated network,” Communication Research, volume 14, number 1, pp. 85–108.http://dx.doi.org/10.1177/009365087014001005
A. Rubin, 1983. “Television uses and gratifications: The interactions of viewing patterns and motivations,” Journal of Broadcasting & Electronic Media, volume 27, number 1, pp. 37–51.http://dx.doi.org/10.1080/08838158309386471
A. Rubin, E. Perse, and R. Powell, 1985. “Loneliness, parasocial interaction, and local television news viewing,” Human Communication Research, volume 12, number 2, pp. 155–180.http://dx.doi.org/10.1111/j.1468-2958.1985.tb00071.x
N. Signorielli, G. Gerbner, and M. Morgan, 1995. “Violence on television: The Cultural Indicators Project,” Journal of Broadcasting & Electronic Media, volume 39, number 2, pp. 278–283.http://dx.doi.org/10.1080/08838159509364304
L. Spigel, 1992. Make room for TV: television and the family ideal in postwar America. Chicago: University of Chicago Press.
C. Tichi, 1991. Electronic hearth: Creating an American television culture. New York: Oxford University Press.
S. Wolf and B. Latane, 1981. “If laboratory research doesn’t square with you, then Qube it: The potential of interactive TV for social psychological research,” Personality and Social Psychology Bulletin, volume 7, number 2, pp. 344–352.http://dx.doi.org/10.1177/014616728172026
Received 20 January 2011; revised 30 January 2011; accepted 2 February 2011.
“Tweeting about TV” by D. Yvette Wohn and Eun–Kyung Na is licensed under a Creative Commons Attribution–NonCommercial–ShareAlike 3.0 Unported License.
Tweeting about TV: Sharing television viewing experiences via social media message streams
by D. Yvette Wohn and Eun–Kyung Na.
First Monday, Volume 16, Number 3 - 7 March 2011
A Great Cities Initiative of the University of Illinois at Chicago University Library.
© First Monday, 1995-2017. ISSN 1396-0466.