Home / Articles / Volume 13 (2016) / Stance and Identity in Twitter Hashtags
Document Actions



Hashtags on the social media microblog Twitter were originally intended to indicate categories, topics, or keywords in users’ tweets, but the use of this affordance has shifted to become its own interactional communication. This study examines users with social Twitter accounts to understand how stance-taking – the act of evaluating objects during a dialogic exchange – is enacted tweet-by-tweet, analogous to face-to-face initiation and response. By applying conversation analytic frameworks focused on stance (DuBois, 2007; Myers, 2010), individual tweets are analyzed as dialogic expressions to address the relationship between tweet content and hashtag. The findings indicate that the Twitter users in this study appropriate hashtags to communicate in a particular way: Users often enter into dialogue with themselves by taking a stance through evaluation, positioning, and alignment. This stance-taking over time can contribute to identity construction.


Writing in online spaces should be studied continually – the use of language changes as the nature of interaction changes. Each social media site, for example, is distinctly different in the kinds of expression it affords. Expression is mediated by the interface of each site, which regulates how users can approach, utilize, or apply particular affordances of the site and in what capacity. The interface, therefore, is perhaps the most influential feature of the site for a communicator to consider. Kress and van Leeuwen (2001) argue that the medium always adds to the design of a composition. Thus, when writers choose a medium, they overtly connect the form of their message to its content.

Observing how users interact socially is particularly interesting on a site like Twitter, whose interface has constraints that invite users to manipulate language. Users of Twitter, whether consciously or unconsciously, make language choices as they communicate in 140 characters or less. Moreover, as Honeycutt and Herring (2009) note,

Twitter is a ‘noisy’ environment, due to the large number of tweets and the speed with which they are posted. This, combined with the fact that tweets are posted in the order received by the system, leads to a high degree of disrupted turn adjacency… (p. 3)

One affordance on Twitter is the use of hashtags, a word or a series of words placed after the pound (#) symbol. Twitter.com suggests that hashtags can label keywords in a tweet, for popular topics called “Trending Topics,” or they can be used as categorizing tools. A help page on Twitter.com (2016) suggests that using a hashtag “correctly” means “us[ing] hashtags only on Tweets relevant to the topic.”

This study examines an innovative appropriation of hashtags, whereby users create dialogue within a single tweet through hashtag use, particularly to present a temporary stance. A stance, as defined by DuBois (2007), is “a public act by a social actor, achieved dialogically…through which social actors simultaneously evaluate objects, position subjects (themselves and others), and align with other subjects, with respect to any salient dimension of value in the sociocultural field” (p.163). Although all stances are evaluative, a stance act is more than an opinion; a subject’s stance must be formed in relation to a second subject (social actor) so that both subjects evaluate an object, position themselves by implicating certain sociocultural values, and align according to how their stances converge or diverge.

DuBois’ “stance triangle” characterizes stance as non-linear and as a holistic communication with interrelated elements. The stance triangle provides a nuanced understanding of how evaluation, positioning, and alignment – three characteristics that help to reveal stance – orient individuals to particular objects and other stance takers. DuBois calls the first subject’s stance the “stance lead,” and the second subject’s stance the “stance follow” (p. 161). Because stance is an act of response, DuBois situates the triangle so that it is reliant on two speaking subjects. Thus, a speaker’s stance can only be understood by “referencing the relevant prior stance” (p. 150) in prior discourse. Because very few people make evaluative statements (e.g., “It’s terrible!”) without referencing an object, DuBois’ triangle requires both an object and another speaking subject to form a stance act; a speaker must evaluate an object during dialogue in order to take a stance. This is why a stance lead and a stance follow are required, as in the example, “I love them” and “I love them too.”

Stance-taking in a tweet can function in two ways: It communicates a temporary stance, and it can also work to construct a user’s identity over time. In this study, I focus on analyzing the dialogic interaction that takes place within single tweets, viewing one of the functions of hashtags as communicating, responding to, or clarifying stance in the content of the tweet. The interface of Twitter encourages users to appropriate hashtags in such a way, in that taking a stance within a single tweet removes the need for multiple posts. Users perform dialogic exchanges with themselves to communicate more visibly within the “noisy” environment (cf. Honeycutt & Herring, 2009), whereas using multiple tweets to clarify a stance or expression could lose relevant context across postings.

In this study, I address DuBois’ call to further explore the nuances of stance and Mayes’ (2014) call to study stance in technological environments. In my analysis I explore the following questions: What is the relationship of Twitter users’ hashtags to their tweet content? How do users engage in dialogue with themselves? How do users enact stance with hashtags? How are users constructing identity through hashtags?


This study assumes that written digital communication is interactive, especially on social media platforms like Twitter, which allows for synchronous exchanges through a variety of affordances. Honeycutt and Herring’s (2009) study revealed that some Twitter users exchange collaborative, conversational tweets. Their study demonstrated that the use of the “@” sign to directly address other users led to “a more interactive use of Twitter,” including “expanding the types of content expressed in tweets” (p. 8). Gillen and Merchant’s (2013) analysis of the Twitter interface in 2010 led to a similar conclusion about interaction. They wrote, “To participate in Twitter is to enter into a discursive relationship with others and to expect, to paraphrase [Mikhail] Bakhtin, response, agreement, disagreement and more” (p. 57). Similarly, Pavalanathan and Eisenstein (2015) note that social media sites “capture language use in natural contexts with real social stakes” (p. 188).

Conversational Interaction on Twitter

On Twitter, tweets do not always receive direct replies from other users; however, several affordances of the site allow for conversational exchanges. This includes direct mentions of users (@-exchanges as mentioned by Honeycutt & Herring, 2009), retweets, or hashtags. boyd, Golder, and Lotan (2010) suggest that the act of retweeting allows Twitter users to “loosely inhabit a multiplicity of conversational contexts at once” (p. 10). Even without acknowledgment of another user in a tweet, hashtags can be utilized to encourage interaction. Huang, Thornton, and Efthimiadis (2010) argue that “conversational tagging” relies on the hashtag as “an important piece of the message.” They continue, “The tag can either serve as a label in the traditional sense of a tag, or it can serve as a prompt for user comment” (p. 175). A tweet becomes part of a public feed when a hashtag is included, whether the hashtag is about a topic or functions to potentially “prompt” conversation. Page (2012) suggests that hashtags position users in conversation about a similar topic, even if they are not talking directly with or to each other about that topic. Zappavigna (2015) calls this “searchable talk”: hashtags that “signal the potential presence of other users in the social network” and permit “different dimensions of their discourse to be retrieved and aggregated” (p. 289).

In recent scholarship, studies have examined conversational language construction on Twitter, specifically with regard to the use of hashtags. Some studies that analyzed the affordances of Twitter hashtags are politically focused, examining political preference, political alignment, or political stance (e.g., Golbeck & Hansen, 2014; Mohammad, Zhu, Kiritchenko, & Martin, 2015; Rajadesingan & Liu, 2014; Small, 2011), while some studies analyze hashtag activism (e.g., the corpus of scholarship regarding Arab Spring; Bonilla & Rosa, 2015). Other scholars have looked at language use on social media broadly or specifically in tweets, such as sarcasm and irony (Kunneman et al., 2015), linguistic and discourse acts in tweets (Lockyer, 2014; Wikström, 2014a), and lexical style-shifting (Pavalanathan & Eisenstein, 2015).

Several scholars have focused on individual appropriation of hashtags. For instance, Page (2012) argues that celebrities, corporations, and ‘every day’ Twitter users “self-brand” by using hashtags to promote a representative identity. Wikström (2014b) identifies multiple functions of personal hashtag use, including meta-comments and parenthetical additions. Hashtags that serve as meta-commentary reveal how the tweet should be read (for example, if a user includes “#opinion” in his or her tweet) and can help to clarify self-image. Parenthetical explanations are defined by Wikström as additions “in the sense that the information they provide is likely superfluous to any reader who already knows the context” (p. 138). Pavalanathan and Eisenstein’s (2015) study reveals that Twitter users are aware of audience, as evidenced by hashtag and @-mention uses; their findings suggest that local or personalized hashtags function differently from hashtags that are used broadly as topic tags, in that they are included with the intent to communicate to different audiences.

Zappavigna (2012) has explored how hashtags create relationship bonds between users by creating “ambient affiliation.” She argues:

Perhaps another significant explanatory factor is the human desire for affiliation: we exist within communities of other voices with which we wish to connect. The stances we adopt and observations and evaluations we share all exist relative to the meaning-making of the other members of our social network and to all other potential networks of meaning. In other words, we perform our online identities in order to connect with others. (p. 38)

Zappavigna (2015) suggests that hashtags function as metadata in three ways: experiential (topic related), interpersonal (evaluative), and textual (typographic). She argues that the inclusion of interpersonal hashtags “has little to do with aggregating posts into searchable sets and much more to do with adopting particular attitudinal dispositions involved in enacting different kinds of microblogging identities” (p. 286). The way hashtags function as “interpersonal emphasis” is “difficult to pinpoint” (p. 286). However, the present analysis suggests several ways that Twitter users construct personal, evaluative stances through their hashtags.

Stance and Identity

In keeping with DuBois’ dialogic stance triangle, the examples in the present study – although all single tweets – reveal how Twitter users construct dialogue across tweet content and hashtag to shape stance. In some instances, stance can only be clarified if users are able to enter into dialogue with themselves in single tweets, turning statements into stance acts by utilizing the hashtag as a response to the content of the tweet. The users in this study also engage in intertextuality to index prior discourses. Ochs (1992) notes:

With respect to social history, Bakhtin and Volosinov make the point that utterances may have several ‘voices’ – the speaker’s or writer’s voice, the voice of someone referred to within the utterance, the voice of another for whom the message is conveyed, etc. The voices of speaker/writer and others may be blended in the course of the message and become part of the social meanings indexed within the message. (p. 338)

In some examples in this study, the hashtag appears to encapsulate a variety of ‘voices,’ whether because the user is responding to the ‘voice’ represented in the content of his or her tweet, or to a collection of unnamed ‘voices.’

Stance and identity are related functions, in that taking a stance can contribute to identity construction. Ochs (1992) maintains that social identities (like gender) are enacted over time through language functions like stance-taking. Similarly, Bucholtz (2007) studied the relationship between stance-taking and long-term identity construction, connecting the two with time: “Stance is therefore both a subjective and an intersubjective process, for social identities may be built up through the habitual taking of stances, and interactional dynamics may sediment into social relations” (p. 379). Significant for Bucholtz in identity formation is duration: A stance might have “temporary salience” within an interaction, but repeated stance-taking over time “solidifies into more enduring kinds of identities” (p. 395).

Bucholtz and Hall (2005) understand identity as fluid, temporary, and emergent from discourse, maintaining that it “is shaped from moment to moment in interaction” (p. 591). They include stance as one construction of identity and define identity broadly as “the social positioning of self and other” (p. 586). To explain the construction of identity through interaction, Bucholtz and Hall offer a framework involving five principles: emergence, positionality, indexicality, relationality, and partialness. Their third principle, indexicality, is the principal focus of this analysis:

Identity relations emerge in interaction through several related indexical processes, including (a) overt mention of identity categories and labels; (b) implicatures and presuppositions regarding one’s own or others’ identity position; (c) displayed evaluative and epistemic orientations to ongoing talk, as well as interaction footings and participant roles; and (d) the use of linguistic structures and systems that are ideologically associated with specific personas and groups. (p. 594)

Through these markers, Bucholtz and Hall (2005) conclude that identity can emerge both for speakers and for other participants. Stance can be developed even “in the most fleeting of interactional moves,” and these stances can “build up into larger identity categories” (p. 595).


To explore my research questions, I collected tweets for up to 12 months from 10 users on my personal Twitter feed. These 10 users were chosen because they regularly employed hashtags for a variety of purposes, including appropriating hashtags for personal use, as described above. I examined more than 1,000 individual tweets for evidence of a stance lead and stance follow that included a hashtag; I narrowed my focus to six users who most frequently engaged in this tweet structure. The tweets selected for this study appeared to demonstrate patterns of stance-taking that involve a dialogic exchange between tweet content and hashtag. The users included in this study are not representative examples of Twitter users but are cases that reveal particular processes of stance construction. The purpose of data collection was not to present a systematic method of hashtag stance formation, but rather to demonstrate how users engage in various processes of creating stance through dialogic engagement with the tweet content and hashtag.

The frameworks of DuBois (2007), Bucholtz and Hall (2005), and Myers (2010) are applied to the data to analyze the various uses of hashtags in the examples provided. While a partial focus of the analysis is Twitter users’ dialogic relationships with themselves, it is important to note that if a Twitter user has followers, every tweet also potentially positions or aligns that user in relation to an audience. When analyzing hashtags in context, one must keep in mind that they could index the ‘voices’ of this audience. Myers’ lexical and grammatical approach to stance-taking on blogs allows him to argue that almost every sentence on blogs and wikis has an evaluative component, and by analyzing these statements, one can better understand how bloggers write to communicate with their audiences. While Myers focuses on epistemic, attitudinal, and stylistic stances, therefore employing a different approach from DuBois, his research contributes to this study’s analytical framework in that it allows for comparison of stance-taking on a microblog with stance acts on a traditional blog. Because DuBois’ framework is intended for spoken conversation and Myers applies his framework to written material, the pairing of the two helps to bridge the spoken-written divide and reveals how some Twitter users employ conversational devices in both tweet content and hashtags.

The analysis below examines how users include hashtags in tweets to engage in dialogue with themselves, invoking the three stance categories found in DuBois’ “stance triangle”: evaluation, positioning, and alignment. Also included is Myer’s epistemic category; Myers writes, “Epistemic markers are relatively infrequent (compared to the other categories of stance), but they are very important, because they affect the way blogs are interpreted as part of news and political discussion” (p. 97). While not applied to interpret political conversation, I include this category of stance act to discuss the potential interference in forming an epistemic stance due to Twitter’s affordances.


Understanding dialogic stance means first understanding both the structure and the relationship between tweet content and hashtag. DuBois (2007) states that a stance lead must be referenced in order for the current stance to be understood, and that “only by referencing the relevant prior stance, locatable anaphorically in the dialogic context, can the meaning of the present agreeing stance be understood” (p. 151). In this section, I show that the Twitter users in this study often compose a stance lead within the content of a tweet and include a hashtag or multiple hashtags to construct the stance follow. However, even when these users do not employ multiple ‘voices’ to enact a stance lead and stance follow, they still engage in dialogue between the content of the tweet and the hashtag. This allows for the formation of an initial stance act and fulfills one subject’s role in DuBois’ stance triangle.

The analysis also addresses stylistic and grammatical devices employed by users to help invoke stance. It is important to understand how stance can be interpreted in online settings as opposed to physical, synchronous interactions. Mayes (2014) argues that the context of online conversational interaction is similar to face-to-face interaction; although many scholars point out that “spoken interaction … unfolds at the moment of speaking,” Mayes maintains that stance and identity in online settings “can be altered incrementally (if not at that moment)” (p. 276). Mayes also suggests that online stance acts are often more lexically and grammatically complicated than the examples found in studies of face-to-face interaction. As Mayes notes, face-to-face or verbal communications can be immediately adjusted, while written asynchronous communications (such as on Twitter) must be cleared up over time or multiple exchanges. Below I show that the Twitter users include writing that employs tone, stylistic and grammatical choices, and conversational devices to make their stances evident without the need for multiple exchanges.

Referential Objects

While DuBois’ stance-taking requires two social actors, sometimes Twitter users do not employ multiple ‘voices’ to fulfill stance lead and stance follow. Instead, they utilize hashtags as if dialoguing with the content of their own tweet; this allows an initial stance act to take place. DuBois notes:

[I]n cases where it may not be obvious that the full stance triangle is in play, it is usually possible to break the triangle down into its component vectors (e.g., an individual stance vector constituting the subject-object evaluative relation), and thereby to achieve an insightful, if partial, analysis of stance. (p. 168)

Even without a stance lead, readers must first understand the “referential object or target toward which the stance is being directed” (DuBois, p. 147) to understand how the subject-object-evaluation relationship takes shape in a single tweet. Once the referential object is identified, stance can be interpreted through various categories of stance acts. Consider the following examples:

  1. @User2: Taye Diggs’ voice in Rent. #Love

  2. @User1: My daughter will be making up school days until July 4. My senior loves the cancellations because no makeup for him. #nomoresnowdays

In example 1, the stance object forms the entirety of the tweet, but it is not until the hashtag that readers can understand the subjective, evaluative stance of “#Love.” The hashtag reveals the stance. In example 2, the reader cannot understand how the user feels about the topic until the hashtag, when she positions herself in relation to the referential object (school weather cancellations) and aligns herself differently from her son. Again in this example, the hashtag reveals the stance of the user. It is important to note that in both of these examples, the users could have composed their tweets to simply state their stances within the 140 character constraints of Twitter’s status box. For example, @User2 could have written “I love Taye Diggs’ voice in Rent” and @User1 could have made a statement about snow days within a sentence. But instead the users employ hashtags dialogically to promote interaction within their own tweets. Due to the disrupted adjacency of tweets and their responses on Twitter, users sometimes create different ‘voices’ within the same tweet or use hashtags to enter into dialogue with the content of the tweet. Using hashtags is one way they create this self-interaction at the moment of writing.

  1. @User3: slacking on keeping up with all the podcasts I listen to, not that anyone was asking, but putting this out there might help me. #pointless

Without the hashtag “#pointless,” the tweet content in example 3 would not reveal the actual stance of the user: What this user attempts does not matter because he believes he will never be able to listen to all of his podcasts. The construction of the tweet content also invokes dialogue itself, and this user positions others by assuming they do not care (“not that anyone was asking”). His hashtag, therefore, also reflects this positioning to align both the pointlessness of mentioning this situation to his audience (since they will not care) and the pointlessness of mentioning the situation to himself (since discussing it will not help him catch up on podcast episodes). Thus in this example, there are two referential stance objects: The entire tweet content that reveals the situation at hand and the ‘voiced’ assumed values of this user’s audience.

  1. @User2: Cady wont do Russian twists and crazy Ivans with me. “Anything with the words twists and crazy is not my kind of workout.” #dying #hilarious

In example 4 there are two prior utterances, that of @User2 and that of Cady. Cady’s stance is indicated by the use of quotation marks. Myers refers to this quotation structure as “reported speech” which can be used to “dramatize a situation from which the reader can infer the writer’s view” (p. 108). What is learned in the tweet content, however, is only Cady’s point of view, that she refuses to do seemingly odd workout exercises, the object of Cady’s stance.

Not until @User2 explains that this is funny can the reader understand her stance and how she positions and aligns herself. Identifying the referential object in the tweet content that the hashtag responds to is necessary in order to understand the stance she takes. @User2 situates herself differently from Cady’s stance about the workout, and her hashtags support their differing alignment (although in an amused way). Interestingly, the hashtag “#dying” also functions to reveal this user’s stance but only in relation to the additional hashtag “#hilarious.” If the user included “#dying” as the only hashtag, it could be assumed she was in physical pain from her workout and employing a figure of speech (dying to signify an exaggerated claim). When placed into conversation with “#hilarious,” the tweet content helps to explain that this user is metaphorically dying from laughter, giving additional context to her stance.


The clearest forms of stance revealed in hashtags include assessments, which are commonly found at the end of a tweet, where hashtags are often employed. Although all stances are evaluative, including a hashtag as an evaluation “implicates those dimensions of sociocultural value which are referenced by the evaluative act” (DuBois, p. 141). DuBois notes that this kind of evaluative stance taking is the most recognized facet of the stance triangle; he defines it specifically as “the process whereby a stancetaker orients to an object of stance and characterizes it as having some specific quality or value” (p. 143). While Myers’ approach conceptualizes evaluation in terms of several sub-categories (aesthetic, judgmental, affective), his analysis encourages blog readers to look beyond explicitly written words to understand how specific lexical and grammatical choices lead to both evaluative stances and identity construction.

  1. 5) @User3: had a fantastic conversation with @keelyjoy today about goals, the future and passions. #revitalizing

In example 5, the hashtag responds to the prior utterance: “#revitalizing” refers to the “fantastic conversation with @keelyjoy” that included a discussion of “goals, the future and passions.” Followers of this user, however, may understand his stance act in a more nuanced way. For example, a user might know of his career as an illustrator and that he loves spending time with his wife, two topics that he tweets about often and could easily be part of a conversation about “goals, the future and passions.” Thus, his hashtag might also function to index these objects that have been voiced in prior tweets.

Bucholtz and Hall (2005) make an important distinction between interactional identity construction and long-term associations:

On the one hand, the interactional positions that social actors briefly occupy and then abandon as they respond to the contingencies of unfolding discourse may accumulate ideological associations with both large-scale and local categories of identity. On the other, these ideological associations, once forged, may shape who does what and how in interaction, though never in a deterministic fashion. (p. 591)

When users respond to their own ‘voices,’ they create intertextuality among tweets that can help other users interpret stance. It is possible that some stance acts might reference the user’s previous tweets, but because the homepage of Twitter displays a users’ tweet within a feed of other users, those previous stances might not be readily viewable. Thus, the stance found in example 5 can be interpreted quickly as evaluating the entire content of the tweet, rather than offering a complex understanding of the specific referential objects (“goals, the future, and passions”) that are “#revitalizing” for the user.

  1. @User3: a beautiful woman laying beside him & his dog laying on top of him, all while a beard warms his face, is the blanket a man can have #manup

In example 6, the same Twitter user describes his situation (a beautiful lady friend, a loving dog, a well-grown beard), likening it to a blanket, something that is comfortable and cozy. His hashtag “#manup” responds to this idea by revealing a gendered ideology. Placing the hashtag in dialogue with the rest of the tweet, it appears that this user believes that one can have these things only if one takes on the responsibilities of a man. Presumably, recognizing the stance objects as “a beautiful woman,” a “dog,” and “a beard” as indicators of a manly lifestyle lead the reader to understand how the user positions himself within his understanding of gender and relationship norms.


When users position themselves or others, readers can also discover aspects of users’ identities. DuBois explains that positioning is “the act of situating a social actor with respect to responsibility for stance and for invoking sociocultural value” (p. 143). The tweets below are examples of how users position themselves in relation to specific objects in the moment, and their hashtags act in dialogue with the content of the tweet to reveal assumed social values.

  1. @User3: talking on the phone inside a quiet coffee shop #petpeeve COUGH @keelyjoy COUGH

In example 7, @User3 mentions a situation many people would find impolite (people talking on the phone in quiet venues) but uses the hashtag “#petpeeve” to assign this particular annoyance to himself personally. While his hashtag stance makes clear that this event bothers him, the additional comment to a fellow Twitter user @keelyjoy – and personal acquaintance, as understood from tweet example 5 – after the hashtag makes the stance less severe as he calls his friend out on her actions by placing her name between fake coughs. In the content of his tweet, he calls upon readers who understand what this situation might be like; with the hashtag he positions himself with those who are annoyed by certain social behaviors in coffee shops.

  1. @User2: Eyes are freakishly blue today. #mutant

  2. @User2: I feel prettiest in a cut off with messy hair and cheap sunglasses. #judgeme

Examples 8 and 9 both invoke values of beauty, and both tweets involve claims by the user of not meeting particular notions of societal beauty. In example 8, @User2 calls herself a “#mutant” because her eyes appear “freakishly” blue. The alignment of the tweet and hashtag includes two words to suggest she looks abnormal or unusual, and neither word suggests the appearance is appealing. This user employs two ‘voices’ to create a stance lead with the content of the tweet and a stance follow with the hashtag. With the hashtag “#mutant,” the user positions the appearance of her eyes as abnormal by further evaluating them in response.

In example 9, the same user appears to be self-conscious about when she feels most beautiful, challenging her audience to judge her (and because she calls them out on it, she assumes some will). The tweet content and hashtag work to form a stance: The user does not care if you judge her for having “messy hair” or wearing “cheap sunglasses” because she is comfortable with those aspects of herself. In her hashtag, she invites someone to take her up on the challenge, extending the dialogicality of her stance act to other social actors. As evidenced by this stance, the user reveals particular sociocultural expectations of her audience about beauty and fashion standards.


In the following examples, users clearly reveal a stance lead in the content of the tweet and then echo the evaluation in the hashtag as the stance follow. DuBois characterizes alignment as “the act of calibrating the relationship between two stances” (p. 144). The hashtags in the examples below clarify the stance lead by offering an additional nuance of meaning.

  1. @User2: Working main bar with Lisa. Setting up to be a good morning! #baristawin

In example 10, @User2 supports the stance lead with the hashtag. She predicts a “good” morning with her fellow barista Lisa, and her hashtag aligns with her optimism to articulate specifically how “good” of a morning it will be – a “win[ning]” scenario. This user also assigns herself a barista identity and speaks with authority on what makes for a pleasant work environment for baristas (in this case, it would be Lisa). This is important to note because the user specifically includes “barista” in her hashtag to align the “win” with the object of her stance (“main bar”) in the stance lead; her stance follow reveals a “win” not because she is spending the morning with Lisa, but because she is working as a barista with Lisa that morning.

  1. @User1: Hey, self-acclaimed genius smarty-pants, it’s “affected” not “effected”! #yourenotmensa

Example 11 sounds like a stereotypical moment when a school bully teases the “smarty-pants” and a sidekick friend chimes in with a supporting retort (“#yourenotmensa”). @User1 employs two ‘voices’ to form a stance lead and stance follow, responding to the “self-acclaimed genius smarty-pants” and the grammar error in the reported speech. This user’s hashtag, aligning with the tweet content, enters into dialogue with the stance lead to echo the taunting of “smarty-pants,” the shared subject of both stances.


Focusing on single facets of the stance triangle allows for a nuanced understanding of how those facets function within a single tweet, but DuBois argues that evaluation, positioning, and alignment are always present when stance is taken. His stance triangle accounts for “causal and inferential linkage” of the three functions (p. 164), which creates a “unified stance act” (p. 145). In the examples below, I analyze how Twitter users in this study dialogue with themselves in tweets in order to evaluate referential objects, position themselves according to certain sociocultural values or expectations, and align the hashtag with the content of the tweet.

  1. @User2: Just wanna give a big “fuck you,” to all the homework that I have to get done. Graduation in 34 days can’t come soon enough. #OverIt

In example 12, @User2 makes very clear she is displeased with her homework load. The use of “Just wanna” indicates she is aware that neglecting homework is actually not an option so close to graduation. Her hashtag, therefore, functions to express her frustration as she evaluates her mood and motivation, aligns with the stance lead, and positions her as apathetic about her current homework condition.

  1. @User4: Went to the pool and realized I’m at least seven shades pastier than everyone else. If I was a paint swatch, I’d be “Heavy Cream.” #sowhite

@User4’s use of the word “pastier” works to position him as unhealthy, since ‘pasty’ is a physical feature one does not aspire to. The user’s hashtag evaluates his whiteness with an intensifier (“so”), which positions him as an extreme outsider. But the hashtag also aligns with the content of the tweet, which states his paint swatch would be a milky shade of white. “Heavy” does the same work as “so” in his hashtag: It intensifies the literal whiteness.

  1. @User2: I couldn’t look like any more of a dirty hippie right now. #longhairdontcare

The hashtag in example 14 further evaluates the referential object (this user’s appearance) from the stance lead: The user does not care that she looks like a “dirty hippie,” and her phrasing implies it is a state of appearance that one should care about. The user also positions herself in a specific social discourse with the linguistic structure ‘___ hair, don’t care,’ a current meme often invoked by users on social media when displaying or discussing their hair. This phrasing of the hashtag aligns with the image of a “dirty hippie” in the stance lead and reveals the user’s stance through an evaluation.

In example 14, @User2 once again writes about her physical features. The stances in examples 8, 9, and 14 evaluate correlated referential subjects, suggesting that this user believes she does not meet conventional beauty standards. Patterns of stances such as these for these Twitter users contribute to what Bucholtz (2007) refers to as sedimentation of identity.

Epistemic Stance-Taking

What might seem like firm statements of fact in the following tweets are actually stance acts. In these tweets, users provide evaluations of certainty with the inclusion of hashtags. Myers (2010) analyzes epistemic stance through the incorporation of facts on blogs, noting that the authors he examined actually used few facts, and when they were uncertain, they were careful to say so. He also discusses grammatical decisions bloggers make to communicate stance, but on Twitter these must sometimes be condensed to function within the character constraints. Thus the qualified verb-plus-clause structure, ‘I think Audis are the sleekest looking cars,’ might be changed to a simple assertion like ‘Audis are sleek’ in order to fit the comment into 140 characters. So while Myers explains that the use of modal verbs and hedging (could, might, sort of, maybe, etc.) reflect a writer’s certainty or uncertainty, Twitter users might construct epistemic stances that communicate certainty because they do not have the required characters to hedge toward their stance.

Epistemic stance-taking is a different facet of stance than evaluation because it positions these users according to specific sociocultural values that do not appear temporary; instead, these users seem to present their preferences like general truths, which contributes to the sedimentation of their identity.

  1. @User6: My mother feels it necessary to invite every single human being that doesn’t have family to our Easter. That’s why I love her #goodmom

In example 15, @User6 uses the hashtag to declare his mother is a “good” one, and the rest of the tweet works to position him (and his mother) and aligns with his declaration. The parallel alignment of stance lead and stance follow position this user as someone who appreciates and values his mother.

  1. @User5: I want to audition for #thevoice. Not because I can sing but because I want to be in the sexy presence of @UsherRaymondIV #truth

The first hashtag in example 16 functions as what Zappavigna (2015) calls experiential hashtags, which are organizational and about a topic or practice. The inclusion of “#thevoice” inserts this tweet into a public feed about the musical television show to potentially generate dialogue with other users. The second hashtag is a bit more difficult to analyze, as it is not clear if the “#truth” the user mentions is that @UsherRaymondIV (a former judge on The Voice) is “sexy” or if it is true she only wants to appear on the show because she wants to be in the presence of @UsherRaymondIV. Either way, it is ‘truth’ that positions the user in relation to the stance object (@UsherRaymondIV’s physical appearance), revealing features she finds attractive in the opposite sex.


Stance becomes problematic to interpret when Twitter users invoke a sarcastic, ironic, or contradictory tone or reference personal social habits or expectations. When users dialogue with themselves, clarifiers typically found in verbal conversation are sometimes lacking. When users do not effectively communicate their tone, or their hashtag does not align with the content of their tweet, their stance can become unclear.

  1. @User2: Straight munchin’ celery right now. #normal [also includes picture of user and acquaintance with celery stalks held to mouth]

  2. @User4: I just had one of those evolution smoothies from Starbucks. It tasted like celery and genitalia. #healthy

In example 17, @User2’s tone is unclear. Her use of the hashtag “#normal” on the surface suggests what she is doing is an accepted practice, yet if it were an acceptable activity she probably would not feel obligated to explain. If eating celery is an abnormal activity, the sociocultural value she is calling upon is unclear: People do not eat celery or ‘munch’ on plain celery? Or is it is normal for her to do this abnormal thing, and therefore she is a bit strange?

Example 18 is perhaps the most complicated, sophisticated tweet and hashtag pairing of the included examples. The Twitter user makes it clear his smoothie was not the best tasting drink by comparing the taste to “genitalia.” (He also states it tastes like “celery” which perhaps sheds light on example 17.) The included hashtag, however, offers a positive connotation of the drink. Kunneman et al. (2015) suggest hyperbolic sarcastic intensifiers such as ‘fascinating’ reveal an evaluation. The authors argue, “[T]he inclusion of a sarcastic hashtag reduces the use of linguistic markers that otherwise would be needed to mark sarcasm” (p. 508). Without a hashtag, they note, sarcastic tweets are sometimes difficult to distinguish.

The confusion in example 18 comes from the disalignment of the tweet content with the hashtag. When placed in dialogue with each other, this user could be critiquing the marketing of the product (“evolution smoothies” are “healthy” but disgusting); could be critiquing the vulgar taste of all healthy foods by using the stance object as representation; could be evaluating his own reaction to healthy foods; or could simply be remarking that the drink tasted unsatisfying yet is, indeed, healthy. Most likely the user is employing sarcasm through the disalignment of tweet content and hashtag; he still positions himself and evaluates the stance object, which reveals an ironic evaluation.

Stylistic and Grammatical Choices

Stylistic and grammatical writing choices help to reveal stance by explicitly positioning subjects in relation to referential objects. Myers (2010) considers stylistic stance markers such as adverbs and metacomments as representations of tone in blogs. Character constraints on Twitter can affect the notions of style that Myers examines – manipulation of grammar, word choice, and abbreviations become important writerly choices when expressing tone and stance on Twitter, similar to the careful choosing and execution of hashtags.

  1. @User1: Can’t.Sleep.Enough #vitamindme

  2. @User6: Athens sucks this time of year #out #of #towners

Examples 19 and 20 demonstrate how tone can be revealed in the tweet content or in the hashtag through punctuation or spacing. @User1 uses periods between her words for emphasis, just as one might speak slowly, pausing between words, when upset, surprised, or frustrated. @User6 makes a similar stylistic choice, separating his words with the hashtag (pound) symbol. @User1 positions herself clearly in the stance content – she is tired and wants more sleep – and the hashtag “calibrates” (DuBois, p. 144) this stance to align by suggesting vitamins to compensate for a lack of sleep. @User6 reveals his stance in the tweet, but he places the stance object in the hashtags where he assigns blame to out-of-town visitors and self-identifies (positions himself) as a local member of the town (“Athens”). In both instances, the use of separation between words emphasizes the users’ frustration and helps construct their stances.

  1. @User1: My kids getting best and second best scores in their classes with their essays>>>>>winning the lottery! #englishgeek

Example 21 is notable because of the repeated use of the ‘greater than’ sign, a typographic style choice sometimes found in digital writing. The greater than sign in mathematical usage communicates that whatever is to its left is greater than or has more value than what is to its right. In digital communications, it can also function as a kind of arrow, pointing to a result. In either use, it is a symbol that expresses evaluation and positions the user when reduplicated; the use of multiple signs in a row (reduplication) functions as an intensifier.

In this example, the user explains that her children excelling in school either means more to her than a windfall of money or is equivalent to winning a windfall of money. Her hashtag aligns with the tweet content by further self-positioning in her stance follow. It is most likely that this user intends a positive connotation to the identity of “englishgeek,” so it positions her as someone who cares very much about good writing, as evidenced by her evaluation of her children’s high scoring essays. This identity aligns with the repeated ‘greater than’ sign found in the stance lead.

  1. @User1: 2) The past: It wasn’t the great. Music and hair were bad and drugs didn’t help you. At all. Let.It.Go. #tantrumtuesday

In example 22, @User1 reveals two stances: Frustration at an unnamed audience and awareness that she is probably overreacting. Her tweet functions as a plea to move on from the past and position her as frustrated or angry, implied by the brief utterances, negative words, and the typographic separation of “Let.It.Go.” Her hashtag, however, works to slightly disalign with the tweet content. Choosing the word “tantrum” rather than another word like ‘rant’ or ‘complain’ reflects self-awareness that she is letting her temper control her reaction. Because of space constraints, she cannot explain this latter notion to her audience in the tweet content, so she communicates it through the use of carefully chosen words.

Conversational Devices

Conversational cues are also found in tweeting, lending the writing a conversational tone. Myers (2010) understands imagined speech sounds as conversational devices (p. 112). Similarly, Wikström (2014b) argues that hashtags like ‘#sigh’ or ‘#laughs’ represent intentional “face-to-face paralinguistic cues” (p. 140). Wikström notes that these verbal cues often occur organically in physical interaction, but to type them requires intentional communication of the emotion. Imitations of speech sounds make tweets interactional, and they can also function as stance markers. This becomes especially clear when the conversational device is employed as the stance follow.

  1. @User1: Screaming kids sounding like tortured cats….. #shhhhhh

  2. @User5: Today is one of those days I should not have logged onto Facebook…. two more engagements and a pregnancy #gahhh

  3. @User5: Made boyfriends mom’s meatloaf for dinner #yummmmmm

  4. @User4: Why do we read restaurant reviews on an app named after “a short sharp cry, esp. of pain or alarm?” #YELP

These examples include hashtags that mimic noises. In two examples, the hashtags align with the tweets to reiterate the emotions of the users: In example 23, the hashtag reflects the user’s desire for quiet, and in example 24, the user is distressed by her Facebook log-in experience. In example 25, the hashtag works to reveal the user’s stance through evaluation. Without “#yummmmmm,” it would be unclear how the user felt about her boyfriend’s mother’s recipe. In example 26, @User4 asks a rhetorical question and uses his hashtag to demonstrate what he means. Because the user has placed it in all capital letters, the reader can imagine he is actually yelping the name of the restaurant review website (Yelp.com). The stance object (the contrast of the app’s purpose and definition) and the hashtag align to illustrate this user’s point.

Temporary and Sedimented Identity

Both stance and identity are ephemeral and fluid. Moreover, identity construction is difficult on Twitter because the interface positions any given tweet within a newsfeed of others’ tweets. Thus, a user’s tweets may not create a clear identity narrative for his or her followers. However, if a follower becomes familiar with the stances a user takes – for example, if the user regularly tweets evaluations about a repeated topic, as in the examples below – an identity may begin to sediment, in the words of Bucholtz (2007). The key to this identity creation is the ability to interact with a user’s tweets over a long period of time. It is still necessary to interpret each stance as responsive to a user’s current context rather than reflecting a predetermined identity.

  1. @User5: Now even Facebook ads are taunting me about weddings and engagement rings #notfair

  2. @User5: Had another dream about engagements last night :( #itslikeasickness #nobueno

  3. @User6: Don’t know what I would do without the saint that is my momma #mommasboy4lyfe

Examples 27 and 28, along with example 24, reveal that @User5 is anxious to be engaged and married. Example 29 and example 15 work in the same way – this user loves and respects his mother, and he uses tweets and hashtags to communicate that stance to his audience when his evaluation warrants the adopting of a “#mommasboy” positioning. It is likely that these users have aspects of their identities that have sedimented in their physical lives that are similar to the stances they take in these tweets; for example, loving one’s mother is not a temporary stance or identity. But because their followers might not know these personal details, these users must position themselves to form stances that communicate who they are online – identities that Bucholtz and Hall (2005) define as “interactionally emergent rather than assigned in an a priori fashion” (p. 587).

The stances of @User5 and @User6 above are expressed toward specific referential objects on a tweet-by-tweet basis, rather than stemming from fixed identities that are applied to every tweet. However, recurrent stances about the same topics can begin to accrue for these users, sedimenting elements of their Twitter identities.


This study contributes to current scholarship that analyzes various appropriations of hashtag use (Page, 2012; Pavalanathan & Eisenstein, 2015; Wikström, 2014b; Zappavigna, 2015) by offering a specific process for how some users construct the kinds of evaluative, personal hashtags and identities observed in previous studies. As shown in the examples above, the Twitter users in this study appropriate hashtags to communicate stance dialogically in multiple ways:

  1. The hashtag in tweets can function to evaluate an object, position, or align with content in the tweet.

  2. The tweet content can function as the stance lead; the hashtag can function as the stance follow.

  3. With stance-taking over time, users’ Twitter identities can sediment.

This newly observed form of constructing stance within a single tweet is significant for conversation analysis scholarship: The users demonstrate that they can fulfill all facets of DuBois’ stance triangle by evaluating, positioning, and aligning with objects without the explicit presence of another social actor. As Hutchby and Wooffitt (2008) describe, one of the aims of conversation analysis is to “discover how participants understand and respond to one another in their turns at talk, with a central focus on how sequences of actions are generated” (p. 12). Studies like the present analysis productively challenge the definitions that ground conversation analysis – expanding concepts such as turn-by-turn interactions, what it means to be a participant in a conversation, and how sequences of interaction that involve digital modalities function. Zappavigna (2015) notes that “further work is needed that considers the particular linguistic structure of the conversation-like ‘interchange’ afforded by social media, since it is likely that it differs fundamentally to forms of turn-taking or ‘exchange structure’ seen in offline conversation (Berry 1981; Martin 1992)” (p. 275). The complex interactions found on social media appear to challenge conventional understandings of discourse and are worthy of continued observation and analysis.

If tweets are truly to be treated dialogically, future sociolinguistics scholars might consider how identity sediments by analyzing the entirety of a user’s feed. Bakhtin (1986) argues that utterances cannot be interpreted alone: “There can be no such thing as an isolated utterance. It always presupposes utterances that precede and follow it. No one utterance can be either the first or the last. Each is only a link in the chain, and none can be studied outside this chain” (p. 136). While a user’s tweets appear as single, momentary bursts on their followers’ newsfeeds, all users have profiles where their tweets are collected. Future analysis could examine the construction of identity over the course of a profile’s timeline, how a user’s stances operate as links in the identity chain, and how a user’s timeline might suggest a different identity from stances taken in individual tweets.

The present study is limited by the small sample size, which could have hindered discovery of more patterns in how users create stance across tweet content and hashtag. Future research should examine a larger corpus of tweets to investigate if this phenomenon generalizes across Twitter. Future sociolinguistics research about Twitter could also focus on the inclusion of additional modes in tweets to reveal how they contribute to stance; as Twitter users employ photos, videos, and emoji in their tweets, their interactions and accompanying language choices become more complicated.


I am grateful to three anonymous Language@Internet reviewers for their constructive, encouraging feedback and am grateful to Susan Herring for her detailed attention to this manuscript. I am also indebted to Pat Mayes, whose suggestions and support guided me through early drafts of the manuscript.

Twitter Data Sources

User 1. [User1]. Tweets [Twitter page]. Retrieved September 2012-April 2013 from http://www.Twitter.com/User1

User 2. [User2]. Tweets [Twitter page]. Retrieved March-April 2013 from http://www.Twitter.com/User2

User3. [User3]. Tweets [Twitter page]. Retrieved March-April 2013 from http://www.Twitter.com/User3

User4. [User4]. Tweets [Twitter page]. Retrieved April 2012-April 2013 from http://www.Twitter.com/User4

User5. [User5]. Tweets [Twitter page]. Retrieved January-April 2013 from www.Twitter.com/User5

User6. [User6]. Tweets [Twitter page] Retrieved March-April 2013 from www.Twitter.com/User6


Bakhtin, M. (1986). From notes made in 1970-71. In C. Emerson & M. Holquist (Eds.), Speech genres & other late essays (pp. 132-158). Austin: University of Texas Press.

Bonilla, Y., & Rosa, J. (2015). #Ferguson: Digital protest, hashtag ethnography, and the racial politics of social media in the United States. American Ethnologist, 42(1), 4-17. doi:10.1111/amet.12112

boyd, d., Golder, S., & Lotan, G. (2010). Tweet, tweet, retweet: Conversational aspects of retweeting on Twitter. In Proceedings of the 43rd Hawaii International Conference on System Sciences (HICSS-43). Kauai, HI: IEEE Press.

Bucholtz, M. (2007). Shop talk: Branding, consumption, and gender in American middle-class youth interaction. In B. S. McElhinny (Ed.), Words, worlds, and material girls: Language, gender, globalization (pp. 371-402). Berlin: Mouton de Gruyter.

Bucholtz, M., & Hall, K. (2005). Identity and interaction: A sociocultural linguistic approach. Discourse Studies, 7(4-5), 585-614. doi: 10.1177/1461445605054407

DuBois, J. W. (2007). The stance triangle. In R. Englebretson (Ed.), Stancetaking in discourse: Subjectivity, evaluation, interaction (pp. 139-182). Amsterdam/Philadelphia: John Benjamins.

Gillen, J., & Merchant, G. (2013). Contact calls: Twitter as a dialogic social and linguistic practice. Language Sciences, 35, 47-58. doi:10.1016/j.langsci.2012.04.015

Golbeck, J., & Hansen, D. (2014). A method for computing political preference among Twitter followers. Social Networks, 36, 177-184. doi:10.1016/j.socnet.2013.07.004

Honeycutt, C., & Herring, S. (2009). Beyond microblogging: Conversation and collaboration via Twitter. In Proceedings of the 42nd Hawaii International Conference on System Sciences (HICSS-42). Los Alamitos, CA: IEEE Press.

Huang, J., Thornton K., & Efthimiadis, E. (2010.) Conversational tagging in Twitter. In Proceedings of the 21st ACM conference on Hypertext and hypermedia, 173-178. doi:10.1145/1810617.1810647

Hutchby, I., & Wooffitt R. (2008). What is conversation analysis? In Conversation analysis (pp. 11-40). Malden, MA: Polity Press.

Kress, G., & van Leeuwen, T. (2001). Multimodal discourse: The modes and media of contemporary communication. New York: Oxford University Press.

Kunneman, F., Liebrecht, C., van Mulken, M., & van den Bosch, A. (2015). Signaling sarcasm: From hyperbole to hashtag. Information Processing & Management, 51(4), 500-509. doi:10.1016/j.ipm.2014.07.006

Lockyer, D. (2014). The emotive meanings and functions of English ‘diminutive’ interjections in Twitter posts. Journal of Theoretical Linguistics, 11(2), 68-89.

Mayes, P. (2014). Interactive advertising: Displays of identity and stance on YouTube. In H. Lim, & F. Sudweeks (Eds.), Innovative methods and technologies for electronic discourse analysis (pp. 260-284). Hershey, PA: Information Science Reference.

Mohammad, S. M., Zhu, X., Kiritchenko, S., & Martin, J. (2015). Sentiment, emotion, purpose, and style in electoral tweets. Information Processing & Management, 51(4), 480-499. doi:10.1016/j.ipm.2014.09.003

Myers, G. (2010). Opinions: Where do I stand? Discourse of blogs and wikis (pp. 95-113). New York: Continuum.

Ochs, E. (1992). Indexing gender. In A. Duranti & C. Goodwin (Eds.), Rethinking context: Language as an interactive phenomenon (pp. 335-358). New York: Cambridge University Press.

Page, R. (2012). The linguistics of self-branding and micro-celebrity in Twitter: The role of hashtags. Discourse & Communication, 6(2), 181-201.

Pavalanathan, U., & Eisenstein, J. (2015). Audience-modulated variation in online social media. American Speech, 90(2), 187-213. doi: 10.1215/00031283-3130324

Rajadesingan, A., & Liu, H. (2014). Identifying users with opposing opinions in Twitter debates. Lecture Notes in Computer Science, 8393, 1-8. Retrieved June 10, 2015 from http://arxiv.org/abs/1402.7143

Small, T. A. (2011). What the hashtag? A content analysis of Canadian politics on Twitter. Information, Communication & Society, 14(6), 872-895. doi: 10.1080/1369118X.2011.554572

Twitter Help Center. (2016). Using hashtags on Twitter. Retrieved January 8, 2016 from https://support.twitter.com/articles/49309#

Wikström, P. (2014a). & she was like "O_O": Animation of reported speech on Twitter. Nordic Journal of English Studies, 13(3), 83-111.

Wikström, P. (2014b). #srynotfunny: Communicative functions of hashtags on Twitter. SKY Journal of Linguistics, 27, 127-152.

Zappavigna, M. (2012). Discourse of Twitter and social media. London: Continuum.

Zappavigna, M. (2015). Searchable talk: The linguistic functions of hashtags. Social Semiotics, 25(3), 274-291. doi: 10.1080/10350330.2014.996948

Biographical Note

Ash Evans [evans39@uwm.edu] is a doctoral candidate at the University of Wisconsin-Milwaukee. Her research interests include online and digital writing, classical and new media rhetorics, and composition pedagogy.


Any party may pass on this Work by electronic means and make it available for download under the terms and conditions of the Digital Peer Publishing License. The text of the license may be accessed and retrieved at http://www.dipp.nrw.de/lizenzen/dppl/dppl/DPPL_v2_en_06-2004.html.