Introduction

All language use builds on previous instances of what Becker (1995) calls ‘languaging’; therefore all talk is implicitly connected to the voices of past experiences. One way for speakers to negotiate the history of their talk is by using the resource often referred to as reported speech. Reported speech, while not necessarily an exact repetition of a previous speaker’s words, is used by speakers to explicitly assign responsibility for portions of their talk to prior voices. Goffman’s (1981) approach to participation, discussed in more detail below, has laid the groundwork for many studies investigating this tangle of voices in spoken discourse (e.g. Goodwin & Goodwin, 2004; Hill & Zepeda, 1992; Irvine, 1996; Levinson, 1988). Few studies, however, focus on the ways that reported speech and participation function in rapidly changing computer-mediated conversational environments.

In this study, we describe the dialectic relationship between the evolution of the interface of the microblogging platform Twitter and the extant practices and needs of its user base regarding reported speech. In 2009, the designers of Twitter added a new feature to the medium to accommodate the user-created manual practice of sharing prior text from other Twitter users, most often using the notation ‘RT’ to indicate that the material was being ‘retweeted.’ This interface change allowed tweeters to more easily share others’ tweets with their own followers through the simple click of a button, with minimal imposition of the user’s own voice on the prior text. We focus here on this change to the Twitter interface and its impact on the practice of sharing prior texts on this platform, attending to the ways in which users incorporate the new feature into their extant practices and how this allows them to negotiate the interplay of voices and the assignment of responsibility in ways that are particular to Twitter.

We begin with an introduction to Twitter as a platform of communication to situate the analysis in context. Following this introduction, we provide background information on reported speech, a phenomenon that is commonplace on Twitter using the retweet functions. Murthy (2012: 1067) notes that Twitter audiences often attribute retweeted material to the user sharing the prior tweet, despite clear reference to the original tweeter. In the second section, we explore such conflicts of ownership and authorship between the retweeter and original tweeter using the Goffmanian participation framework, updating it for a new medium with new rules and constraints. The case of retweeting illustrates the dialectic between the practices of users and designer-implemented interface modification. To illustrate, in the third section, we describe user practices as they have been affected by the change in modal affordances, resulting in two forms of retweeting that each perform different interactional work in terms of displaying alignment through participation structures and attribution. We find that this situation has resulted in the coexistence of two different forms of retweeting, nuanced in their participatory affordances and complementary in their usage, and we discuss the relationship of these retweet practices to participatory concepts such as principalship and authorship. Finally, we illustrate how the co-existence of these two forms of retweeting has resulted in a complementary set of emergent practices. This study thus contributes to a growing body of research on participation in social media and the use of reported speech and prior texts in interpersonal online interaction.

Twitter as a Platform of Communication

Twitter, founded in 2006, represents a mode of social media known as micro-blogging. Micro-blogging, as defined by Murthy (2012: 1061), is:

an internet-based service in which (1) users have a public profile in which they broadcast short public messages or updates whether they are directed to specific user(s) or not, (2) messages become publicly aggregated together across users, and (3) users can decide whose messages they wish to receive but not necessarily who can receive their messages; this is in distinction to most social networks where following each other is bi-directional (i.e., mutual).

Twitter users are often referred to as ‘tweeters,’ based on the name of the short messages that they produce, ‘tweets.’ An individual user’s tweets appear in a collected feed on his or her ‘homepage,’ which other users can visit to see all recent tweets produced by the user. When one user decides that she wants to see messages from another tweeter regularly, she can ‘follow’ that user’s account, resulting in this second user’s tweets being displayed in the first user’s ‘timeline,’ which presents tweets from all the tweeters that she has followed in reverse chronological order. As noted by Murthy, this act of following does not require reciprocity; User A can follow User B without any need for User B to follow User A in return. This can result in a uni-directional spread of information, with User A receiving the messages that User B posts without User B ever seeing anything posted by User A.

‘Retweeting,’ an iconic characteristic of Twitter, is the practice of reproducing another user’s tweet on one’s own homepage and redistributing that tweet to one’s own followers. Retweeting originated as a user-generated practice to share informative, humorous, or creative tweets produced by other tweeters; this practice was indicated by the retweeter by manipulating the format of the basic tweet by typing the acronym ‘RT,’ usually followed by the original tweeter’s username (indicated with @username) and then the content of the tweet being shared. Retweeting is a way of showing engagement on Twitter, and discovering what kind of material gets retweeted has been one way of increasing engagement on the platform. Bongwon Suh and colleagues (2010) analyzed 74 million tweets to discover which features of tweets contribute to ‘retweetability.’ Their results indicated that tweets that included URLs and hashtags were highly retweeted, indicating that content is being shared in some relevant way; furthermore, the age of the account and the number of follower and followees increased retweetability, indicating that the network of the original author is influential in the level of engagement.

When something is retweeted, what is the role of the author of the tweet, and what is the role of the retweeter? Who gets the attribution for the tweet? While this RT notation was understood by users to indicate that content from another user was being shared, the retweeting user could modify the content of the original tweet or add more content of his or her own, often resulting in some confusion as to which parts of the tweet had been produced by the original tweeter and which had been produced by the retweeter. This confusion of authorship and attribution was addressed in boyd, Golder, and Lotan’s (2010) study of retweeting, and it also caught the attention of the designers of Twitter. In 2009, Twitter changed its interface to incorporate a tool for retweeting more simply and without alteration of the content of the original tweet. This change gave users a new option for sharing material produced by others. However, while the new retweet function facilitated rebroadcasting of tweeted material, it did not allow retweeters to engage with the content of the original tweet, as they could with the older RT form, by adding their own content or modifying the original tweet. As a result, two popular forms of retweeting currently co-exist: the new retweet function built into the Twitter interface and the older, user-generated retweet convention. We argue that users deploy these two forms to indicate different relationships to retweeted text.

Reported Speech and Sharing of Prior Text

While early work on reported speech and prior text largely focuses on the role of repetition and attribution, later studies, including Tannen’s (1986, 2007) research on constructed dialogue and Johnstone et al.’s (1994) research on intertextual repetition, have focused on the ways that so-called reported speech is altered when imported into new interactional environments in spoken language. Tannen (2007: 112) argues that:

when speech uttered in one context is repeated in another, it is fundamentally changed even if 'reported' accurately. In many, perhaps most, cases, however, material represented as dialogue was never spoken by anyone else in a form resembling that constructed, if at all. Rather, casting ideas as dialogue rather than statements is a discourse strategy for framing information in a way that communicates effectively and creates involvement.

For Tannen, analyses of attribution of talk to a previous speaker should center less on any possible repetition and more on the new construction, including new frames for interpretation of the information that is being presented in the form of dialogue. Gordon (2006: 546) continues this theme, noting that “prior text is shaped and reshaped to create different alignments and identities in different interactions and for different audiences”.

While we continue the use of the term ‘reported speech’ in this article (due in part to the possibility for near-exact repetition in the sharing of previously produced talk on Twitter), we note the objections of Tannen and others to the idea of reported speech as an exact repetition of earlier talk, and we bear in mind the role of creativity and new construction in the sharing of prior text on Twitter. Regardless of its faithfulness to the original form, reported speech is viewed here not just as a way of reproducing content from another person’s talk, but also as a rhetorical device for the layering of multiple voices in one’s own discourse. The features of Twitter allow for many options for reproducing the talk produced by others, including fully faithful, largely modified, and wholly created representations.

Participation frameworks in reported speech

Much of the interactional work done by participants in dialogue that takes place through the use of reported speech centers around the concept of participation. In this study, we draw on the framework for participation proposed by Goffman (1981), in which he elaborated participant roles beyond those of physically uttering the words (speakers) and of receiving the acoustic signals of this speech (hearers). For Goffman, participation status is “the relation of any one such member to [an] utterance.” Of primary of interest in this work are his distinctions for participants in production roles: animator, author, and principal:

animator: the participant who physically produces the utterance, “the sounding box”

author: the participant who composes the utterance, “the agent who scripts the lines”

principal: the participant responsible for the utterance, “the party to whose position the words attest” (Goffman, 1981: 226)

By separating out these roles, Goffman enables a more useful analysis of the relationship of participants involved in producing the talk than a simple conception of the “speaker.”

In addition to these traditional roles, Draucker (2012) notes that the features of Twitter have given rise to a divergence in the role of transmitting talk, which, for Goffman, could be neatly packaged into the animator role. Draucker suggests that when considering Twitter interaction, it is useful to distinguish between the roles of animator, defined as “the person who is physically producing the text,” and broadcaster, “the followable entity that makes the talk available to recipients.” She notes that all tweets necessarily have both a broadcaster and an animator, but that the two roles need not be filled by the same participant. This distinction can often be seen clearly in the talk of organizational Twitter accounts, in which the organization itself is set up as the broadcasting party that makes the talk available to the account’s followers (e.g., the Pittsburgh Steelers American football team for the @steelers account), while some (often unknown) individual or group performs the actual physical animation of the text for the tweets (most likely through typing, but also potentially through voice-to-text software). Draucker’s definition of the broadcaster role is intentionally specific to the features of Twitter, but it has applications in other mediums such as television or print, particularly those that allow for the animator to be obscured from the audience in some way. We adopt this distinction in transmission of talk in this article and add the broadcaster to Goffman’s traditional production framework, as it is particularly relevant to a discussion of reported speech and the sharing of prior text.

In addition to recognizing the possibility for different participants to take up different roles in their production of talk, it is important to take into account the possibility for multiple participants to perform overlapping roles. Bakhtin argues that all talk is multi-voiced in some way, drawing on the usage of words and phrases that have gained meaning from their prior uses, but he also makes a special designation for “double-voiced discourse,” or the use of “another’s speech in another’s language” (1981: 324). Bakhtin argues that this double-voicing imbues words with three types of meaning: the direct meaning of the quoted speaker, the meaning created through the current speaker’s use of reported speech, and the dialectic meaning created in the interplay between the two. Similarly, Tannen’s (2007) argument regarding constructed dialogue asserts that the attribution of talk to previous speakers is a rhetorical device, used to enact the present speaker’s stance towards the talk he or she is supposedly reproducing.

The meaning of reported speech, then, is created not just in the content of the words, but in the ways in which speakers layer the multiple voices to which those words can be attributed. Besnier (1992, adapted from Pascal, 1977: 8) provides the following example to illustrate the manipulation of voicing and attribution when reporting the talk of others:

  1. She stopped and asked, “Is that the car I saw here yesterday?”
  2. She stopped and asked if that was the car she had seen there the day before.

In example (a), the speaker is attributing the speech back to the woman represented by the pronoun ‘she,’ indicating that the speaker is reporting the speech directly as it was uttered by the woman. Example (a) employs multiple voices, but the direct quotation style is meant to indicate no overlap in production roles – the speaker clearly takes on the role of animator, but assigns the author and principal roles to the woman he or she is quoting. In example (b), however, the speaker employs indirect quotation, allowing for more overlap in the production roles. While the speaker is still the sole animator, the authorship of the sentential complement of if is unclear – the hearer does not know to what extent the woman or the current speaker has authored or is the principal of the final part of the utterance.

Additionally, it is important to differentiate the content of a speaker’s utterance from metalinguistic ideas such as patterns of preferred usage, responsibility, and authorship. Norms for reporting speech in a community or medium reflect local theories of information, authorship, and meaning (Besnier, 1992). Our analysis investigates the common practices for Twitter users in sharing the prior tweets of other users, asking how the available features of the platform influence the forms of sharing and reproducing talk on Twitter.

Participation and reported speech on the Internet

Goodwin and Goodwin (2004) note that technology is changing the face of participation, opening up possibilities for wider participation and communication; consequently, in attempts to apply traditional concepts of participation frameworks to online contexts, Goffman’s roles no longer neatly apply. In Jirotka et al.’s (1991) study of an online strategy game, the authors demonstrated that while users design their talk for certain participants, the text-based nature of conversation confused some speakers about whether they were viewing a conversation or a series of individual announcements. In his work on blogging, Myers (2010) notes that the many roles held by each individual participant often make it difficult to use Goffman’s categories to separate the intended participants (and participant roles) from the unintentional roles that users take on.

To this end, it is important to keep in mind the modal affordances of the medium of study. Originally introduced by Gibson (1977), the term ‘affordance’ is used to refer to the potential of an object to allow a user to perform an action. The term has been taken up in studies of computer-mediated communication and human-computer interaction; it is defined here as “options for interaction that the environment provides for the participants residing in it, and especially those options that are acted upon by the individual” (Örnberg Berglund, 2009: 187). Awareness of the modal affordances of the medium allows the analyst to understand the process by which norms manifest and become standard practice (Cherny, 1995). These standard practices arise out of a soup of possibilities afforded by the medium, and their acceptance ultimately depends on context and discourse dynamics (Androutsopoulos, 2011). For instance, Herring (2010) argues that even though interactions on the Internet could be technically classified as ‘typing’, they are conversational in nature and can be classified like ‘talk’ – as such, we use the term ‘talk’ to refer to the conversations and interactions happening in an online space, even though the talk is (usually) produced by a keyboard instead of the vocal cords. Constraints on the actions of users are another aspect of modal affordances that must be considered, for example, Twitter’s limit of 140 characters per tweet. This character limit imposed by the medium is one motivation for users to shorten word forms, elide words and characters, and modify retweeted material to fit under this limit. However, as described by Hargittai and Walejko (2008), in the practice of creating and remixing content to post online, there is still a ‘participation divide’ based on a number of socioeconomic factors such a race, gender, and education that leads some groups to post and share content more than others.

With this in mind, we turn to previous research on Twitter to more fully elaborate the modal affordances of this platform. One cue that has been appropriated on Twitter (and in other Web 2.0 contexts) is the process of addressing or tagging a user by using the @ symbol. This is an example of a process that was generated by Twitter users and was later built into the Twitter interface by the developers so that typing the sequence ‘@username’ created a clickable link to the addressed user’s profile as well as notifying the user that they had been mentioned. In a massively multi-participant public environment like Twitter, a high degree of addressivity is required if a user is attempting to conduct a conversation (Honeycutt & Herring, 2009). Addressivity using @ was not always intended to be conversational; for example, another significant use of @ was to attribute talk to other authors, specifically in retweets. Another method of tagging tweets is using the hashtag symbol #, which tags tweets with keywords that are searchable and followable. In an analysis of tweets following the 2008 United States Presidential election, Michele Zappavigna (2011) describes hashtags as a way to create community on Twitter by making talk searchable online. Using hashtags, users can find and interact with other users who are discussing the same topics; this can lead not just to conversations with other users and an expansion of a user’s network of connections around shared topic interest, but also discovery of material for their own Twitter timeline through retweeting.

boyd et al. (2010) described early practices of retweeting, showing how attribution may be obscured by users attempting to conform to the character limit while also adding comments and retaining the content of the original tweet. They found several strategies that users employed to formulate retweets, the most frequent of which was restricting attribution to one user rather than a series of users. This particular practice was the main cause of obfuscation of authorship — sometimes, when presented with a string of usernames in a multiply-retweeted message, new retweets would reduce the string and inadvertently omit the original tweet’s author. An example from boyd et al. (2010: 8), below, shows how the original tweet’s author, zephoria, has been deleted in a subsequent retweet as part of a string of authors.

  1. @jtoddb: RT @mStonerblog RT @zephoria: new blog post “Is Facebook for old people?” based on interviews w/ teens in Atlanta http://bit.ly/v0aPS.
  2. @sparepixel: RT @mStonerblog: new blog post “Is Facebook for old people?” based on interviews w/ teens in Atlanta http://bit.ly/v0aPS. Always insightful!

Murthy (2012: 1066) also discusses the use of retweeting for the purpose of introducing new voices into an interaction on Twitter, suggesting that retweets not only enable users to bring the words of others into a new context much in the same way as reported speech or constructed dialogue, but create the impression of a conversational exchange. With changes to the Twitter interface, automatic retweets preserve the identity of the original author which can lead to greater exposure; therefore, exposure-minded users of Twitter attempt to craft tweets that may lead to many retweets and increase their own popularity. In a study of the likelihood of tweets being retweeted, Wang, Chen, and Kan (2012) found that short, humorous tweets expressing the author’s personal opinion are likely to be retweeted, while personal updates and complex tweets are less likely to be retweeted. Xu and Yang (2012) investigated features of tweets that promote retweeting, and found that the primary factor influencing retweet numbers on tweets is a social connection between the two users – that is, if User A has retweeted User B previously.

It is also interesting to note that retweets in boyd et al.’s study were often seen as a tacit signal of agreement or approval of the content. This follows on Bakhtin’s argument that repetition of the words of others will have one of two different basic meanings – to parody the source or to honor or agree with the source (Morson & Emerson, 1990). The primary way for users to disagree with the content of an original tweet was to add a comment to the retweet expressing the contrary stance, a modal affordance not available to users of the retweet form later added by Twitter designers.

While the change in the Twitter interface may not have changed users’ motivations for retweeting, it has altered not only the potential form of the retweets, but also the ways in which multiple voices can be represented in a single tweet. In this article, we look at how the interface changes have impacted practices of retweeting, specifically with regard to attribution of roles in the production of talk. We argue that the change has given rise to two co-existing forms of retweeting which allow for differing participation structures and different ways of framing the information for interpretation by a tweeter’s audience.

Defining the Two Classes of Retweets

The retweet examples used in this study come from the Pittsburgh Corpus, tweets from the summer of 2012 that were collected by searching and archiving the keywords Pittsburgh, Pitt, and Pgh. Using this sample, we analyze the discursive strategy of retweeting in mundane talk on Twitter. We deploy a qualitative approach due to the in-depth user knowledge required for understanding modal affordances as they impact usage. We focus on a descriptive analysis of the practices currently in place for the Twitter userbase and use example tweets (also referred to as “tweet utterances” by Murthy 2012: 1067) to illustrate the practices gleaned from our corpus. Using this method of analysis, we define the two classes of retweets with associated examples drawn from the corpus and discuss how they relate to reported speech, participation, and attribution.

At the time of the data collection, two primary options were available to users in the interface and commonly used in practice for retweeting. In defining these two retweet forms, we borrow from boyd et al.’s (2010) classification of ‘preserver’ and ‘adapter’ types of retweeters. As boyd et al. found, “in deciding how to adjust a tweet for retweeting, a significant divide arises between respondents who seek to preserve as much text of a tweet as possible and those who are willing to adapt retweets by removing various parts of the tweet that were, in their opinion, nonessential” (2010: 5). The new interface-enabled form of retweeting requires the retweeter to preserve the original tweet in full form, including the username and profile picture of the original tweeter, and thus we use the term Preserving Retweet (RT) to describe this tool. Alternatively, the user-generated form of retweeting allows retweeters to adapt or alter the tweets to fit their needs, including possibly deleting any reference to the original source of the tweet. We therefore use the term Adapting Retweet (RT) to refer to this style of retweeting; while altering the content of the original tweet is not required, it is always an option for the retweeter. Additionally, the original tweet is removed further from its original context, as the audience of the retweet will see the profile picture of the retweeter and that tweeter’s username at the top of the tweet, as opposed to the identifiers of the original tweeter. It may be helpful to think of the Adapting RT as a ‘manual’ retweet, indicating that the user must manually reproduce content, and the Preserving RT as an ‘automatic’ retweet in which the content is automatically ported by the Twitter interface.

The Preserving Retweet

The newer style of retweeting, the Preserving Retweet, was introduced into the Twitter interface in 2009 as a response to the widespread user practice of redistributing the tweets of others. By including this function in the Twitter toolbox, the developers attempted to improve upon the user-generated retweet function, specifically aiming to deal with confusion and inaccuracy in attribution and allowing users to retweet the entire content of an original tweet without having to modify the tweet (plus the RT notation and the username of the original tweeter) to fit into the 140-character limit (Stone, 2009). However, as was anticipated by former Twitter developer Ev Williams in his 2009 blog post discussing the function (Williams, 2009), differences in the appearance and function of the retweets led users to adopt the new Twitter-introduced style for some purposes and maintain the original retweet style for others.

In Preserving Retweets, the original tweet cannot be modified – it is simply reproduced in its original form, with no additional content added. When User A retweets User B with the Preserving RT style, the retweet shows up in user A’s timeline as if it has been cut out of user B’s timeline and pasted into user A’s. This is indicated by the inclusion of user B’s name and avatar along with the text, as well as the information that it has been retweeted by user A at the bottom of the tweet. Figure 1,1 below, shows a Preserving Retweet. The original author, LuigiLemieux, has both his username and avatar appear in the newsfeed, and the retweeter, PensNation_Ken, has his name appear underneath the retweet with a double-arrow icon.

Figure 1 

Preserving Retweet: The user PensNation_Ken retweets a tweet originally created by the user LuigiLemieux, “A Pittsburgh Penguin walks into a bar … No timetable for his return.”

This contrasts with the Adapting RT in that it reproduces both the avatar and username of the original tweeter. This facilitates attribution to the original author more easily than the Adapting RT, which allows for erasure of attribution, as seen in tweets observed by boyd et al.

The Adapting Retweet

The Adapting RT is similar to the most common user-generated form of retweeting studied by boyd et al. (2010). In Adapting RTs, user A can indicate that they are reproducing text from user B by marking the quoted text off with the capital letters “RT” followed by the username of the original tweeter. The text from the original tweet can be modified (although it does not have to be), and user A can add their own text, which is often placed before the RT and retweeted material. The most common syntax of Adapting Retweets is: {comment text} RT @{author} {retweeted text}. As can be seen in Figure 2, in this method of retweeting the avatar and username of the user performing the retweet (here, user Sta7ic) appear in the timeline, just as they would for regular tweets produced by this user. The retweeted material appears after the acronym RT, and the identification of the original author with “@ThePittNews,” and Sta7ic’s added comment (“Yeah, no. @PosvarHall is one of the WORST i’ve seen”) appears at the beginning of the tweet.

Figure 2 

Adapting Retweet: Sta7ic retweets material from ThePittNews (“Posvar a better-planned building than the Cathedral [link]”) with a preposed comment indicating the user’s stance towards the retweeted content (“Yeah, no. @PosvarHall is one of the WORST I’ve seen”).

Sta7ic is highlighted as the participant who is sharing this tweet, as we see his username and profile picture prominently featured, while the author of the retweeted material, ThePittNews, is less readily apparent on first glance at the tweet.

While other forms of sharing previous tweets also continue to exist, such as the use of quotation marks around the retweeted text or the retweeted text followed by the notation “(via @username),” these forms are used only occasionally while the Preserving and Adapting RT styles are by far the most prevalent.

Analysis: Relationship of RTs to Principalship and Authorship

Retweeting, as a form of sharing prior talk, raises interesting questions of principalship, authorship, and the mixing of voices in a stretch of talk. In this section, we show how Preserving and Adapting Retweets build different participation frameworks on the production side of talk. For tweets that do not include reported speech, a single Twitter user can be assigned all four roles of animator, broadcaster, principal, and author. However, beyond this simple production framework in which one participant assumes multiple roles, there are many instances in which multiple participants can be assigned to a single role, or where the connection between participants and the roles that they play is unclear, notably in many retweets. In this section, we discuss examples of tweets demonstrating variations of this ambiguity of participant roles.

Figure 3 illustrates the participant roles in a straightforward way with a basic tweet (by ‘basic,’ we mean a tweet that includes only text, not any retweets or @ mentions of other users). The user Ord410 inhabits the roles of animator, broadcaster, author, and principal (as shown in the accompanying participant roles table). The first person deictic pronoun ‘I’ points to Ord410 as its referent and therefore indicates his or her principalship. As for the authorship and animatorship of this tweet, while it is possible that this tweet was not composed or physically written by the user Ord410, the content of the tweet does not give us any reason to believe that this is the case. Finally, since this tweet was visible in Ord410’s timeline, Ord410 is the broadcaster of this talk.

Figure 3 

A basic tweet by the user Ord410 representing a single participant acting as author, principal, animator, and broadcaster of the message. The table indicates the participant roles held by Ord410.

In Figure 4, below, we see a basic tweet in which the principal role is filled by a participant other than the animator/broadcaster, and the author role is shared and to some degree unclear. The user josephperson is the animator and broadcaster of the tweet, having physically produced the tweet utterance and shared it in his timeline. The actual content of the tweet, however, is attributed to another person, Ron Rivera, who is the principal of the message. (Note that the original utterance by Ron Rivera is not in the form of a tweet, but rather was made in another medium, hence this is not a retweet.) Finally, the authorship of the exact wording of the tweet is unclear; because the tweet makes use of an indirect quotation form, it is not clear whether Ron Rivera’s words are being repeated as he originally uttered them, or if josephperson has changed the words to some degree to make them fit into the character limit or to portray the message more clearly. Much like Tannen’s (2007: 110) argument for constructed dialogue, it is possible that much of this quote is constructed from memory and only loosely matches the original form of Ron Rivera’s actual talk. While it remains clear then that the talk is meant to be attributed to Rivera, we cannot know the extent to which the words that appear in the tweet have been composed by Rivera himself or by josephperson. This ambiguity is noted in the accompanying chart in Figure 4, indicating that both Rivera and josephperson may have had some role in the authorship of this tweet.

Figure 4 

At tweet representing a Twitter user (josephperson) acting as an animator and broadcaster, but attributing the talk to another participant (Ron Rivera) as principal. Due to the use of indirect quotation, authorship of the tweet is unclear. A gloss of this tweet reads: josephperson says that Ron Rivera says that starters will only play a minimum Thursday against Pittsburgh, if at all.

In Figure 5 below, the content of the tweet in Figure 4 has been retweeted by the user jjones9 in an Adapting Retweet. Ron Rivera is still the principal, josephperson is still the animator, having physically typed out the text, and the author role is still unclear with both Rivera and josephperson as possible contributors, but there is a new broadcaster for this retweet. jjones9 takes up the broadcaster role by retweeting the content of another person’s tweet to his followers. (jjones9 is the animator, broadcaster, author, and principal for the added RT notation; however, as this notation is simply an indicator of sharing a prior tweet and no new substantive content is added, we will not discuss the roles for such notation further.)

Figure 5 

An Adapting Retweet of the original tweet in Figure 4. Here jjones9 is acting as broadcaster, sharing the talk with his followers, but not acting as a principal, author, or animator of the talk. A gloss of this tweet reads: jjones9 reports that josephperson said that Ron Rivera says that starters will only play a minimum Thursday against Pittsburgh, if at all.

The Adapting Retweet, as stated above, can also be used by retweeters to add their own comment to the retweeted material. Figures 6 and 7 provide an example of this, where we can see that the retweeter has added additional content to the original tweet he or she is sharing. Figure 6 shows a basic tweet from user PittGHOTCH, in which PittGHOTCH acts as the animator, broadcaster, author, and principal of the tweet, while Figure 7 illustrates an Adapting RT of the tweet in 6 with a preposed comment depicting the stance towards the retweeted material of user OaklandZoo (named for the student support section at the University of Pittsburgh men’s basketball games).

Figure 6 

A basic tweet from PittGHOTCH reporting some news from the University of Pittsburgh.

Figure 7 

An Adapting RT with a comment. The tweet in 6 is retweeted by @OaklandZoo, with an additional comment (“Wow”) added before the retweeted text as a response. The two tables illustrate participant roles for each of the two parts of the tweet.

In the Adapting RT in Figure 7, there are two sets of participant roles in play. For the preposed comment “Wow,” OaklandZoo is the principal and broadcaster, with an unidentified member of the student group acting as author and animator. However, for the retweeted text, the original user PittGHOTCH retains all of his participant roles except broadcaster, which has been taken over by OaklandZoo in this retweet. Figures 5 and 7 illustrate the dual utility of the Adapting RT – the retweeter can simply act as a broadcaster as seen in Figure 5, or retweeters can add content indicating their own stance towards the retweeted comment for which they take on additional production roles, along with being the broadcaster of the entire content of the tweet, as exemplified in Figure 7. In Figure 5, the retweeter’s stance can only be assumed by the audience, as no indication of stance is given. Comparatively, in Figure 7, OaklandZoo makes his or her amazement clear through the addition of the stance-taking comment “Wow.” These examples illustrate the flexibility of the Adapting RT; this flexibility is an important motivation for the continued use of this retweet form.

Figure 8 illustrates a tweet by josephperson, while Figure 9 illustrates a Preserving RT by peggyskid. In Figure 8, josephperson is the animator, broadcaster, author, and principal of his original tweet, which is broadcast on his homepage. In the Preserving RT in Figure 9, peggyskid takes over as broadcaster, having rebroadcast the tweet to his or her own followers. While peggyskid assumes the broadcaster role, josephperson remains the animator, author, and principal, as the tweet is still clearly attributed to him and no reanimation of the text is required for the retweet. The shift in broadcasters is the only participant role shift that is available to users of Preserving RTs, as the affordances of this type of retweet do not allow for the addition or modification of content or for erasure of attribution to the original tweeter.

Figure 8 

A regular tweet from josephperson reporting news about the Pittsburgh Steelers football team.

Figure 9 

Josephperson’s tweet from 8 is retweeted by peggyskid, as indicated by the retweet symbol and text “by peggyskid” below the content of the tweet.

The primary differences between the Preserving and Adapting RT forms might be seen in the ways in which they assert the heteroglossia of the retweeted material. The Preserving RT retains clear ties to the participant(s) in the animator, author, and principal roles for the original tweet content, both visually and in the text, by posting the original tweeter’s avatar picture and username with the retweet and giving no opportunity for erasure or change that could lead to confusion in attribution or new authorship of the content. The essential component of the Preserving RT, then, is that it only allows a retweeter to take over the broadcaster role, eliminating the possibility of confusion over other roles, but also not giving the retweeter any chance to express their stance towards the material they are sharing. The Adapting RT, in contrast, gives retweeters more control over the shared material, allowing them to add or alter the content and form of the retweet to fit their purposes and to indicate their feelings about the shared text, as well as to highlight their own participation through the inclusion of their own avatar in the tweet (instead of the avatar of the original tweeter). As the following section will show, however, it is in this freedom that the participant roles begin to intertwine, leading to potential for attribution to be obscured or erased in the process of using the Adapting RT form.

Retweet forms and the obfuscation of attribution

Potential confusion of authorship and attribution can arise through the use of the Adapting RT form. An example is Figure 8, an Adapting RT of the tweet in 9, specifically the Preserving RT done by peggyskid. The user Meemah2AOBL has added the tweet-final comment “AWESOME!” which was presumably one motivation for using the Adapting RT form instead of the Preserving RT form. The most common Adapting RT syntax is {comment} RT {@username}{original content}, with possible recursion of ‘RT {@username}{content}’ for a chain of retweets. However, Meemah2AOBL deviates from this syntax, and as a consequence, the authorship of “AWESOME!” may be incorrectly assigned by a viewer who has not seen the original tweet by josephperson or the Preserving RT by peggyskid.

Following the most common retweet syntax, the comment “AWESOME!” could appear to belong to the original tweet by josephperson. This practice of adding tweet-final comments is one of the primary sources of confusion in retweeting according to boyd et al. (2010). In the conventional Adapting RT practice, comments added before the RT notation provide a clear barrier between the added comment and the original retweeted material. Furthermore, this barrier indicates to the reader the point at which the retweeter shifts participant roles from author to broadcaster. In Figure 10, the postposed comment eliminates this barrier, creating syntactic opacity, which creates ambiguity about the source of the “AWESOME!” comment and its associated stance.

Figure 10 

Adapting RT that embeds the Preserving RT found in Figure 9. josephperson’s original tweet is included in this Adapting RT along with peggyskid’s RT, but it also includes a postposed comment “AWESOME!” made either by Meemah2AOBL or another participant that Meemah2AOBL is copying text from. Many of the participant roles are unclear for a viewer of this tweet in isolation, indicated by “??s”.

While there is confusion over the attribution of the postposed comment in Figure 10, one thing that the Adapting RT allows users to accomplish is the indication of a chain of sharing prior tweets. In Figure 10, Meemah2AOBL was able to indicate where he or she found the retweeted material, showing others that it was because of peggyskid’s broadcast that Meemah2AOBL came across the tweet from josephperson. The possibility to include a chain of attribution is one of the facets that distinguishes an Adapting RT from a Preserving RT. While users do not always include the chain of attribution in an Adapting RT, it is possible to do so. The chain of attribution is obscured in a Preserving RT, which shows only the original author and not any subsequent retweeters.

The tweets in Figure 9 demonstrate this problem of attributing intermediary sources through Preserving RTs. Figure 11 is an original tweet by pittprofdude, while Figure 12 shows two Preserving RTs by parnopaeus and ftd318. These Preserving RTs were produced for the purposes of illustrating this phenomenon, and the retweet from ftd318 was produced from parnopaeus’s retweet. Unlike Figure 10, which shows Meemah2AOBL’s intermediary source, in Figure 12 the intermediary sources of ftd318’s Preserving RT cannot be traced. Due to the modal affordances of the Preserving RT, there is no possibility for ftd318 to give credit to parnopaeus’s role as a broadcaster, which led to ftd318 discovering pittprofdude’s tweet. The affordances of the Preserving RT, then, while quite clearly retaining the original author of the tweet, erase any chain of attribution in broadcasting. For many tweeters this lack of attribution to intermediary sources may not be seen as a problem — the original tweeter is given credit for the material they produced and it might not be relevant to others how the retweeter came across this tweet. However, in a culture highlighting ‘micro-celebrity’ on Twitter (see Marwick and boyd, 2011, and Page, 2012, for further discussion), in which users gain popularity based on numbers of followers, retweets, interactions, and overall exposure, it can be important to some users that they gain attribution for their role in the production of the retweeted material, even if that role is only one of broadcasting. For some users, then, the constraints of the Preserving RT that keep them from attributing the broadcasting chain (or being attributed themselves) might be enough to prevent them from using this retweet form.

Figure 11 

A tweet by pittprofdude including a link.

Figure 12 

Two Preserving RTs of the tweet in Figure 11, as retweeted by users parnopaeus and ftd318. From the information contained in Preserving RTs, any potential intermediary sources are not attributed.

While the functional affordances of Adapting RTs allow for the intermediary attribution that Preserving RTs do not, it is still the user’s decision whether or not to include any sources at all. One of the consequences of using an Adapting RT is that users can erase the author by deleting indicators of attribution to the original author. According to boyd et al. (2010), this is often done by retweeters to reduce the character count of their message so that it fits in the 140 character limit imposed by Twitter. While the “RT” notation is meant to signify to readers that the initial stretch of talk is authored by the original tweeter, the “MT” is commonly used to make it clear to the audience that they have, in fact, modified the original content in some way. Figure 13 shows an example of this phenomenon in which the user TheMadHessian has used the acronym MT (for ‘modified tweet’) to indicate modification of the original replicated content. Many users now use this MT notation to mark a distinction between Adapting-style RTs that are faithful to the original form and those that the retweeting user have altered in some way. Here, TheMadHessian has deleted the original author’s username in the retweeted material. Presumably, the author’s username was erased so that TheMadHessian could include the comment “BAZINGA!” at the beginning of the tweet and stay within the character limit. This motivation likely also explains the modification of the retweeted text, although it is unclear to what extent TheMadHessian has taken over an authorship role for this portion of the tweet. Due to the deletion of the username, it is difficult to trace the original tweet and therefore difficult to discern the ways that the text has been modified. Although it is clear that something has been modified in the tweet, thanks the TheMadHessian’s use of MT, because the original tweet is not easily traceable for those reading the tweet, there is no immediately clear answer as to which parts of the modified tweet are authored by TheMadHessian and which are authored by the unknown and unattributed author.

Further compounding the confusion of participant roles in Figure 13 are the quotation marks included in the modified tweet. As typically used in written forms of reported speech, quotation marks indicate direct representations of actual talk by a referenced speaker; similarly, on Twitter, quotation marks are a resource that can be used in modified tweets to indicate which portions of the original tweet are unaltered. The quotation marks around the text “a high school offense to a pro-style/college offense” create ambiguity resulting in two possible readings: either the quoted text is meant to represent an actual spoken utterance from Tino Sunseri, or the quoted text is unaltered content retweeted from the unnamed author’s original tweet. The modified retweet in Figure 13 seems to suggest the first interpretation, in which the quoted material should be attributed to Tino Sunseri.

Figure 13 

An Adapting-style RT with the notation ‘MT’ indicating that the original text has been modified. The user TheMadHessian inserted the preposed comment “BAZINGA!” to take a stance towards another user’s tweet containing the quote from Tino Sunseri, however attribution to the original creator of the tweet containing the Tino Sunseri quote has been erased. The chart explains participant roles only for the text which appears outside of quotation marks.

The modified retweet by Fiegs2point0 in Figure 14, below, however, shows an instance of quotation marks being used to indicate the parts of the tweet that belong to the original author. Fiegs2point0 uses the following syntax to distinguish the parts of the tweet that he has authored and the parts attributed to bcuban: MT: “@{author}{retweeted [altered] text}” {postposed comment}. The quotation marks set off the retweeted text from Fiegs2point0’s postposed comments, demarcating the boundaries of authorship for the text “Welcome back James McDonald #pirates” as shared text from bcuban, while Fiegs2point0 has added the hashtag “#scallywag” and the username “@jfamacdonald” (presumably for the purpose of letting MacDonald know that bcuban has spelled his name wrong). Fiegs2point0 has also employed square brackets within MacDonald’s last name in the retweeted text to indicate his correction to bcuban’s original tweet (and thus the “modified” part of the MT), indicating that Fiegs2point0 is the animator and author of this added letter, and the principal of the correction. The use of quotation marks here is in contrast with Figure 13, in which the quotation marks used in conjunction with the MT and the deleted author add ambiguity regarding authorship and principalship of the retweet material. In both Figure 13 and Figure 14, we can see that this heteroglossic style of reproduction and alteration serves to both add voices to the talk and, at times, confuse the assignment of these voices.

Figure 14 

User bcuban’s original tweet, with user Fiegs2point0’s Adapting RT with modifications and additional comment indicating that bcuban had misspelled James MacDonald’s name. Fiegs2point0 calls bcuban a “scallywag” for this error using a hashtag, and then tags the Twitter account of James MacDonald (@jfamacdonald). Brackets are used to indicate an addition of a letter to correct a spelling error in bcuban’s tweet. Fiegs2point0 is the animator only for the material that appears in brackets to correct bcuban’s spelling. Fiegs2point0 occupies all participant roles for the postposed comment “#scallywag @jfamacdonald”.

While it is clear from the examples above that there are many retweeting practices in use that allow for ambiguity in representation of authorship and principalship, the core practices are familiar enough for most Twitter users that these practices can be exploited for creative and humorous ends. While the “MT” notation allows users to overtly indicate to others that they are altering the retweeted text, the examples in Figure 15 demonstrate creative uses of retweet conventions by taking advantage of the norms of interaction through a flouting of the rules. Here, we have included two humorous tweets in the form of Adapting Retweets from robdelaney demonstrating this clever appropriation of stance-taking norms.

Figure 15 

Two tweets fully authored by comedian robdelaney, created in the form of an Adapting RT, as an attempt to humorously (and falsely) attribute talk to President Barack Obama. The participant roles of author, principal, and animator are suggested to belong to BarackObama through the syntax of the retweet.

Here, robdelaney uses the Adapting RT form to construct tweets which he falsely attributes to user BarackObama as a tool for humor. This option would not be available for the Preserving RT form since the tweets ascribed to BarackObama do not actually exist, and therefore cannot be retweeted through Twitter’s built-in retweet tool. The Adapting RT format can be used to fictitiously share the text, however, and robdelaney adds a preposed comment to each of these tweets to validate his use of the Adapting RT form, as well as to distance himself from the material in the retweet. robdelaney uses the Adapting RT form to ascribe authorship and principalship of the constructed text to BarackObama. The humor in these tweets is in the fact that the text was not actually produced by BarackObama, and therefore the authorship is retained by robdelaney and the principalship is ascribed to robdelaney’s voiced version of BarackObama. By strategically taking advantage of recognized patterns of use, this method of imposing multiple voices onto one tweet exemplifies the heteroglossic nature of the Adapting RT.

Emergence and standardization

An underlying theme of the co-existence of Adapting and Preserving RTs is the tension between emergent practices and the standardization of those practices. In describing emergent grammar, Hopper (1987: 142) defines emergence as “a continual movement towards structure, a postponement or ‘deferral’ of structure, a view of structure as always provisional, always negotiable, and in fact as epiphenomenal, that is, at least as much an effect as a cause”. Hopper’s definition primarily deals with structure; however, such structures emerge through use and practice, and it is this aspect of emergence that we apply to the retweeting phenomenon.

As seen by the move by some users towards incorporation of the “MT” notation to indicate modified content, the Adapting RT shows clear evidence of being an emergent practice, continually evolving as it spread throughout the population of Twitter users. Because of the nature of emergence, it is continually negotiated in its rules for use. When Twitter codified the retweet in its interface, creating the Preserving RT, suddenly there was a standard, institutionalized structure to retweeting that had not existed previously. There became structure in retweeting, including constraints imposed by the interface. which did not precisely match up to the continually evolving norms in Adapting RT practice.

It is evident from Twitter’s early release notes on the Preserving RT that they did not anticipate this function to replace the Adapting RT. Ev Williams writes in his blog post about the release of the Retweet Button:

Keep in mind, there’s nothing stopping you from simply quoting another tweet if that’s what you want to do. Also, old-school retweets are still allowed, as well. We had to prioritize some use cases over others in this release. But just as Twitter didn’t have this functionality at all before, people can still work around and do whatever they want. This just gives another option. (Williams, 2009)

This signals that the developers understood retweeting as an emergent practice, and while they were standardizing one particular incarnation of retweeting and giving officially-sanctioned structure to it, they were not taking away the ability of the userbase to continue their emergent practice. This showed that although Preserving RTs existed as a single snapshot of retweeting practice, the very nature of retweeting as an emergent practice suggested that it would continue to evolve outside of the structure being imposed on it.

As expected, Adapting RTs continued to thrive, and as such, Twitter responded again with a new addition to their retweeting interface on April 6, 2015. Users can now choose the conventional Preserving RT form, but they are now given the option to “Add a comment” before retweeting. If the user does not add a comment, the Preserving RT is added to their timeline as it has been since the addition in 2009. However, when the user chooses to add a comment, it will now show up in their timeline as if they tweeted an original tweet, with the retweeted text linked below the added comment. This allows the user to make use of the full 140 character limit for their own comment without cutting out any of the retweeted material. The following example in Figure 16 illustrates such a retweet from fieldhousemedia.

Figure 16 

fieldhouse media’s retweet with comment of CoachTomHerman’s original tweet. Retweet created using Twitter’s recently “Add a comment” feature on Preserving RTs.

Here, @fieldhousemedia has retweeted original content from CoachTomHerman. We can see the retweeted content below the added comment of “Coach Herman is killing it on Twitter. #SMsports from fieldhousemedia. The layout of this tweet clearly identifies the added comment as coming from the fieldhousemedia account, while the retweeted text is set off from this text and associated with the original tweeter, as well as being a hyperlinked field that takes the reader to the original tweet. This new option allows users to take advantage of one of the key features of the Preserving RT – that is, preserving the full format of the original tweet – while also getting the opportunity to add their own stance to the tweet.

The addition of this feature will undoubtedly have many implications for participation structures and the continually evolving practices of sharing talking on Twitter. As Twitter users begin to make regular use of this feature, it will add interesting wrinkles to the ways that users can impose their own voice in reproduced talk. This new feature will help to clear up authorship of retweeted material, while still allowing users to author their own talk; however, users will undoubtedly continue to evolve with creative practices that allow them to blur the lines indicating where their voices end and the voice of the original tweeter begins. This feature also helps to demonstrate the tension between the interface and user practice. While many Twitter users interface with the site through mobile applications, this feature is currently only available on the web interface at Twitter.com. As the feature becomes more and more popular, it will likely be incorporated into mobile apps, but in the meantime, access to this feature of sharing talk may impact both the linguistic practices and the access points through which users choose to interact with the social media site. This new retweet form will now have the chance to do what Adapting RTs have long been doing in the Twitter universe – act as an emergent practice with fuzzy rules and ever-changing norms, co-existing with more standardized and familiar versions in the Preserving and Adapting RTs described in this paper.

Conclusion: The Relationship between Modal Affordances and User Practices

In this article, we demonstrated that the two forms of retweeting entail distinct affordances which can be both beneficial and problematic in emergent user practices. While the Preserving Retweet form removes much of the confusion about authorship and principalship, the Adapting Retweet is still deployed by Twitter users for evolving heteroglossic, stance-taking purposes.

The Adapting RT began as an emergent practice that was constantly evolving as the practice spread through the user base. Retweeting became so widely used that the developers of Twitter decided to standardize it into what became the Preserving RT. The top-down implementation of the Preserving RT was intended to help users negotiate the character limit and allow them to retweet more material more easily; however, as the Preserving RT did not exactly replicate the affordances of the extant Adapting RT form, the use of the Adapting RT persisted. The co-existence of these two forms of retweeting, the user-created and the standardized versions, resulted in a new complementary set of emergent practices incorporating both forms.

The introduction of the Preserving Retweet added a multimodal dimension of attribution by including the original tweeter’s avatar in addition to a faithful representation of the retweeted text. The Preserving Retweet uncomplicates authorship and attribution; however, not only does it constrain the retweeter’s ability to comment on the text, it also has the drawback of obscuring possible broadcast chains (as seen in Figures 11 and 12 above). These broadcast chains are important for the micro-celebrity aspect of Twitter, allowing users to not only claim content they have produced but also content they have discovered and shared with others.

The Adapting Retweet allows for the addition of comments, modification of tweet content, and highlight of the retweeter through visual cues. However, as boyd et al. (2010) noted, these additions create the potential for complication of authorship and attribution, and Squires (2011: 5) notes the potential for the erasure of features from a stretch of talk to “privilege the voice of animator and alter the voice of author.” The increase in affordances available to Twitter users since boyd et al.’s study has ultimately not resolved the ambiguity of authorship and attribution; as seen in Figures 13 and 14 above, even the use of MT to indicate that material has been modified does not clearly resolve the issue of who authored the content, as the original author can be erased entirely. The changing dynamics of retweeting have instead created a multitude of opportunities for both exact reproductions of the original text (Preserving RTs) as well as potential derivations through which the retweeter can impose their own voice on the shared material (Adapting RTs and MTs).

We suggest that the continued use of Adapting RTs by Twitter users (despite the constraints on their use which may lead to potential confusion) is part of Twitter’s micro-celebrity culture. The emergent practice of the Adapting RT can be attributed to users’ desire to add their own voice to the text and display their stance towards the retweeted text. In this, Adapting RTs reflect the characteristics of spoken interactions, wherein speakers cannot avoid imposing their own stances on repeated dialogue. Even in exact quotation, something seemingly inconsequential, like intonation, can reveal the speaker’s attitudes towards the talk (Besnier, 1992; Tannen, 2007) and give them an authorship role without adding lexical content. As Johnstone et al. (1994) and Tannen (2007) have shown, repetition can have many functions resulting in subtle changes in meaning and participation, as demonstrated here with the two types of retweets presented. Although the sharing of prior texts necessarily implies some stance towards those texts, direct stancetaking acts are not permitted by the affordances of the Preserving RT. As Bakhtin suggests, the sharing of texts can either parody or honor the prior text, but with the Preserving RT it is up to readers of the retweet to decode the stance being taken by the retweeter. The use of the Adapting RT permits retweeters to remove the question of their stance towards the retweeted material, which is a valuable commodity on a platform where a user’s expression is already constrained by character limits. This heteroglossia can result in potential confusion of authorship and attribution, but this is price users pay for the ability to add their own voice to shared talk. The persistence of the Adapting RT demonstrates that this price is a reasonable one.

Competing Interests

The authors declare that they have no competing interests.