1. Preface
[1.1] When attempting to explain Tumblr to friends and colleagues, I often tell them to forget about Facebook, Twitter, Snapchat, and Instagram and instead to think of Geocities. The comparison isn't perfect, but it serves to highlight how social connections across the site function less by a social network site logic (boyd and Ellison 2007; Ellison and boyd 2013) and more like the personal websites of the early 2000s (Papacharissi 2002), where content is less sanitized and messier, and where personal connections exert far more influence than raw follower count (Cho 2015). This study is an initial step to understand how community is negotiated on Tumblr, a microblogging site where no formal community architecture exists. I examine behaviors present in two tags from a single fandom in order to articulate behavioral trends within and across these two related tags.
[1.2] The data for this quantitative analysis were collected in 2013 and thus reflect an iteration of the site that no longer exists. Tumblr, like many digital web artifacts, is a continually moving object of study. I have tried throughout to contextualize what Tumblr looked like in 2013 and how a very particular set of tags functioned as a community while being unable to call on familiar community structures found on other social media sites populated by fandom participants, particularly LiveJournal (Busse and Hellekson 2006; Kendall 2007). When trying to speak about what Tumblr is, it is equally important to acknowledge what is isn't—and it isn't an ideal web platform for exerting social control and building community boundaries.
2. Introduction
[2.1] Here I investigate the content circulated and conduct performed on Tumblr within the Yogscast fandom using quantitative content analysis. The Yogscast are a group of online Let's Players who produce YouTube videos in which they play video games, joke with one another, and sometimes engage in long-form storytelling. Some members of the Yogs have been friends for a decade, having met through playing the massively multiplayer online role-playing game World of Warcraft (Blizzard, 2004), while others are YouTube content producers who started out independent and then later joined the group. The Yogscast as a business entity takes a portion of earned revenue from member YouTube channels in exchange for promotional, editing, and legal services. The Yogscast is thus both a friend group and a business arrangement. The Yogscast fandom in turn includes discussion of and fan works about personalities and/or characters based on both content producers who are contractually bound to the Yogscast and various friends of those content producers who appear in their videos.
[2.2] At the time of data collection, the "Yogscast: Lewis and Simon" channel (formerly on YouTube as BlueXephos but now as Yogscast [https://www.youtube.com/yogscast], often referred to by fans as the main channel) had over 4.4 million subscribers and 1.8 billion video views (as of March 2018, over 7.2 million subscribers and over 3.7 billion views).
[2.3] Because the Yogs are themselves social media personalities, the perceived distance between members of the group and their fans (called Yognau(gh)ts) is quite close. Yognau(gh)ts who produce romantic or sexual content do not uniformly agree whether their fandom is real person fiction. Some fans contend that they are writing fictional characters based on the Yogs' on-camera personas, particularly in regard to their Minecraft (2011) Let's Plays, rather than about the YouTubers themselves. Yogs have gone as far as reading fan fiction aloud on podcasts and live streams, sometimes affectionately (as on "YoGPoD 39: Christmas Fun Packed Special 2011") and sometimes dismissively (as on "Christmas Livestreams, 2015"). They have also hired fan artists to produce official promotional images. This proximity between content producers and fans produced some level of anxiety in fans, who then adopted particular tagging practices in order to regulate who gained access to fan works and through what means.
[2.4] In this study, I focus on two specialized Yogscast fandom tags on Tumblr, zoethian and sjips. "Zoethian" is a portmanteau of "Zoeya" and "Rythian." This is a heterosexual ship that is used as a plot device in "The Tekkit Chronicles" series of Minecraft videos. "Sjips" is a portmanteau of "Sjin" and "Sips" (note 1). This is a male homosexual (slash) pairing that is joked about in various series, including multiple Minecraft video series.
[2.5] The establishment of content and conduct norms, as well as fandom participants using these tags as both nouns and verbs, signal that tags take on some of the functions of communities in an online space that lacks formalized community pages. Content analysis is used to code for content and conduct variables and to establish quantitative measures of the norms adopted by fans who engage on Tumblr.
[2.6] Tumblr's architecture lends itself to fandom participation because fan fiction, fan art, and general discussion can all be distributed and retrieved from a single platform. Tagging allows fans to indicate the show, movie, or video game that corresponds to a post's content. Without belonging to a community as strictly defined by Tumblr's architecture, users can nonetheless reach a potential audience that may be searching by tag. Participatory fandom culture has been considered as a possible training ground for developing and practicing the production, distribution, and social skills necessary for successful, meaningful engagement in politics and enacting social change (Jenkins 2006). The viral and memetic properties of digital platforms and applications open up participation, both political and personal, to increasing numbers of individuals and publics (Shifman 2014; Papacharissi 2015).
[2.7] Lothian (2013) separates the subcultural fan from the common use of the term. When I say, "I am a fan of Star Trek; I like watching the show," it means a different level of engagement compared to the subcultural fan, who may say, "I participate in Star Trek fandom, where I read and write fan fiction." I use the term "fan" in this latter participatory sense. However, Tumblr supports both styles of fan practice.
[2.8] Participatory fans on Tumblr adopt specialized fandom tags, which are unlikely to be used by Tumblr members who do not identify with a particular fandom or subsection of fandom. Such tags are not merely composed of the name the fandom (such as a show or movie title), or even one of the primary characters. Rather, specialized fandom tags further delineate the potential audience, trying to draw the attention of a specific subsection of the larger fan base or to conceal content from other fans (and content producers) who do not want to view romantic or sexual content. In using these tags, widely scattered fans take on some characteristics of communities. Norms develop regarding what is posted and how fans interact with each other when a particular tag is deployed. Without an architecturally defined community structure, content and conduct are nonetheless debated and controlled.
3. Tumblr
[3.1] Tumblr, which was founded in 2007, by 2013 hosted over 105 million accounts (About Tumblr [https://www.tumblr.com/about], 2013). Since then, it has grown to over 400 million blogs (About Tumblr, 2018). While sometimes identified as a social network site (Dana 2012; Mlot 2012), Tumblr resists several characteristics that researchers have previously identified as characteristic of such sites (boyd and Ellison 2007; Ellison and boyd 2013). Users can follow one another, but the ties do not need to be mutual, and mutual connections do not allow for increasing levels of privacy or disclosure. Social connections on Tumblr are entirely private and for many years could not be seen by a third party (note 2). Tumblr founder David Karp states that these alterations to the blogging and social network site formula are intentional to circumvent aspects of online interactions that "poison a whole community" (Walker 2012). Gaining followers is not intended to be competitive; those wishing to make disparaging remarks are forced to display the remarks on their own blog.
[3.2] Tumblr's multimedia approach to blogging, as well as the absence of a character limit or restriction to a single media format, affords the platform additional flexibility compared to Twitter or single-input social media sites, which may allow only for sharing images or audio, but not both. Although Tumblr blogs are blogs, they add the viral mechanisms of later social media, particularly the ability to reblog posts. Maintainers of early blogs through the 1990s and early 2000s hosted their blogs on personal websites, primarily dealt with technology topics, and consciously tried to distance themselves from "feminine" online journaling practice, thus setting blogs on a conceptually different trajectory than online journals (Siles 2011, 2012). As the concept of a blog stabilized in the mid-2000s, websites such as WordPress eased technical requirements, such as knowing HTML and purchasing web hosting (Siles 2012). Still, while blogging practice have always included links to other places on the web, the blog posts themselves were largely static, viewable only from the maintainer's blog and thus perpetuating a sense of distance between community members, even if they did interact on- or off-line (Kendall 2007; Siles 2011). Tumblr allows posts to be reblogged—that is, reposted to a new blog that is maintained by a different person. This function contributes to the viral nature of posts, potentially opening up otherwise niche content produced by amateurs to greater audiences (Shifman 2014).
[3.3] Tumblr does not have formalized community, interest or group spaces or accounts like LiveJournal or Facebook. On Facebook, a user can become a fan of an account dedicated to a product or person, formalizing this relationship by means of a link that will appear on the user's personal profile. On LiveJournal, community accounts are organized around interests, products, and people. Joining a community formalizes the relationship between the user account and the community account by means of a link to the community in the user's profile, and vice versa. However, many users treat LiveJournal more as a personal journal rather than a community tool (Kendall 2007). Fandom-specific interview and ethnographic work has found that fans find LiveJournal to be an important community space where they can more freely express themselves (Lothian, Busse, and Reid 2007). Tumblr lacks any of these formalized community structures. Tags are one of just a few ways Tumblr users discover content produced and circulated by other members. This is not entirely unlike Twitter, which uses hashtags to facilitate searching and to form temporary affective publics coalescing around particular tags during the height of their popularity, then dispersing (Papacharissi 2012, 2015; Papacharissi and de Fatima Oliveira 2011). Tags are not medium specific. For example, searching for the tag "Star Trek" will retrieve videos, images, and stories that all have the "Star Trek" tag.
[3.4] Previous academic work on Tumblr focused on the platform as one of a number of blogging application options (DeVoe 2009; Welsh and France 2012). While continuing research into Tumblr practice has only just begun to appear in academic discourse, what research has been accomplished recognizes that Tumblr users do share a particular affinity with one another that is unique compared to other social media platforms (Bell 2013; Chang et al. 2014; Hillman, Procyk, and Neustaedter 2014; Thelandersson 2014). While the platform lacks a visible mechanism for articulating the links between users, and there is nothing approaching a community page, which was an architectural feature of LiveJournal (Neill Hoch 2014), Tumblr users do display cohesive behaviors in reblogging practice (Chang et al. 2014) and frequently reference other users with whom they feel they share a relationship (Bell 2013; Tiidenberg 2016). Prior research by boyd (2006; see also Davis 2010) regarding Friendster and MySpace examines alternate means for crafting communities in digital spaces, including curating "top friends." While Tumblr does not have a top friends feature, boyd's work points toward alternative avenues for articulating social and communal ties through social platforms. The present research adds to the existing literature by articulating specific interaction practices occurring within two specialized fandom tags, thus highlighting emerging behaviors and allowing for contrasting the two sets of interconnected fans.
4. Tagging
[4.1] Tags are keywords or phrases that are attached to a specific resource within a system by users of that system rather than by a designated authority (Golder and Huberman 2006; Mathes 2004; Marlow et al. 2006). The resource depends on the system in question. On Tumblr, the resource is the post, while on Flickr the resource is an image. Lin and Chen (2012) separate tagging behavior into three categories. In information organization–oriented tagging, taggers are primarily concerned with personal information retrieval; they tag resources to make it easier for them to personally find a resource again in the future, not for the express purpose of sharing. Social-oriented tagging strives to tag a resource in such a way other system users can discover and further share resources. These two positions "may not necessarily contradict each other. They could logically co-exist" (Lin and Chen 2012, 543). Finally, strategic tagging means that a tagger can work from multiple perspectives and may tag flexibly. Strategic taggers may be "more likely to produce influential tags" (Lin and Chen 2012, 543) and inspire imitation. Tagging imitation occurs when users make use of established, suggested tags. Both Delicious.com, studied by Lin and Chen, and Tumblr provide suggested tags to users on the basis of tag popularity.
[4.2] Tagging systems have been considered as an alternative to taxonomies and controlled vocabularies crafted by information specialists. Controlled vocabularies are hierarchical indexes that require a great deal of up-front work to organize results for easy retrieval later (Macgregor and McCulloch 2006). The vast amount of information already on the internet, as well as the staggering quantity of new content that is continuously generated, means that most internet content cannot be organized into a controlled vocabulary by an authority. Instead, tagging systems develop into folksonomies, where no hierarchical system is planned before tagging begins. Mathes (2004) defines a folksonomy as "an organic system of organization" (3), which is "simply the set of terms that group of users tagged content with[;] they are not a predetermined set of classification terms or labels" (4). Folksonomies "lack hierarchy, parent-child relationships, and sibling relationships between terms" (4), which are standard characteristics of controlled vocabularies.
[4.3] However, quantitative analysis of tag data, primarily from Del.icio.us (now Delicious.com) indicates that tags follow the power law distribution (Golder and Huberman 2006; Guy and Tonkin 2006; Kipp and Campbell 2006). A few tags are used very frequently. Early adopters of the platform establish these heavily used tags, which are then perpetuated by those who join the site later. Many tags are used infrequently. This results in a "long tail" of seldom used tags. Tags in general are therefore much more stable than their anticontrolled vocabulary ethos would suggest (Mathes 2004). Although not predetermined from the onset, tagging systems quickly become dominated by influential users.
[4.4] Marlow et al. (2006) offer a qualitative classification matrix for tagging systems defined by seven dimensions: tagging rights, tagging support, aggregation model, object type, source of material, resource connectivity, and social connectivity. Decisions made by tagging system designers in each of these dimensions have potential implications for tagging behavior. Following this taxonomy, Tumblr is self-tagging, suggests tags to users, uses a bag model (each user tagging a resource can do so independently, allowing for duplicate resources and tags), allows for many different objects (images, video, audio, text), sources material from users, has no resource connectivity outside of the tag, and uses linked but entirely private social connectivity. By this taxonomy, Tumblr is not equivalent to either Flickr or Delicious.com, the two sites commonly researched in regards to tagging behavior. Tumblr offers a different set of constraints that may result in different patterns of tagging behavior.
5. "Zoethian" and "sjips" as specialized fandom tags
[5.1] Scholars, particularly Booth, have considered the affordances and constraints of online platforms such as wikis (2009) and MySpace (2008) for fan participation. However, among academics researching fan fiction, LiveJournal has received considerably more attention. While there is still little specific research in regards to fan cultures on Tumblr (Bury et al. 2013), both Tumblr and LiveJournal's primary status as blogging platforms suggest some conceptual overlap.
[5.2] The tension between the private act of journaling and the public sharing of journal entries via a social network architecture makes LiveJournal a site of performance (Kendall 2007; Lindemann 2005). Busse (2006) observes that women sharing fan fiction with one another through their LiveJournal accounts use affectionate and sometimes sexual language with one another, even if they identify as heterosexual in other contexts. Collaborative online discussions with fan fiction writers and readers confirm that some participants consider LiveJournal fandom communities to constitute a "queer female space" where women can be verbally affectionate with one another (Lothian, Busse, and Reid 2007). Fan cultures do develop norms that may differ from the cultural norms enacted by those same individuals outside the context of fandom.
[5.3] The two tags under investigation, "zoethian" and "sjips," are specialized tags that, although not hierarchical in the formal sense, are narrower than a number of other tags in use. Both are portmanteaus that refer to specific romantic ships. Fans may favor one pairing between fictional characters and reject another. Thus, tagging a post "zoethian" or "sjips" indicates that it may be agreeable to a subset of Yogscast fandom, not necessarily all fans. In 2013, Tumblr's tagging system allowed for such distinctions to be made.
[5.4] Previous research on tagging practice does not necessarily attend to the social norms that develop around a specific tag or set of tags. A focus on two specialized tags limits this project's scope and permits an exploration of how tags are used beyond information retrieval or gaining social clout. The use of these tags permits nascent community building and maintenance through social regulation. Previous work by Honeycutt (2005; Yahoo! Groups), boyd (2006; Friendster and MySpace), Danis and Lee (2005; intracompany websites), and Kendall (2002; MUDs), while qualitative in nature, pay special attention to repeated behaviors as they pertain to community maintenance and group identity—aspects I have tried to quantitatively capture here.
6. Conduct control
[6.1] Conveying and reinforcing group behavioral norms in online is normally accomplished through textual rebukes. Smith, McLaughlin, and Osborne (1997) split a corrective episode into four parts: failure event, reproach, account, and evaluation of the account. Their study, and many similar studies looking at corrective episodes online, source data from message boards and Usenet groups. With these platforms, comments can directly follow the post where the failure event occurs. The account, where the at-fault poster describes or justifies why the failure event occurs, may then follow the reproach. Each is presented one after another, or at minimum in the same thread. In 2013, Tumblr did not allow comments, meaning that failures and reproach events may become separated from one another. One option is for reproachers to reblog the failure event, adding their own comments to the reblog, but this requires reproachers to post the offending entry to their own blog. It is just this type of accountability that Tumblr founder Karp intended when he decided that Tumblr would not use comments (Walker 2012).
[6.2] The second option on Tumblr is to post a general reproach that is not directly tied to the failure event via a new post using tags that overlap with those of the offender. Thus, a failure event using the tag "zoethian" may cause a reproach post that is also tagged "zoethian." The reproach might not be formally linked back to the failure event, but all users tracking the "zoethian" tag would see this reproach.
[6.3] Silencing speech, or "enduring noise and flames focused on the target collective identity" (Chua 2009, 235), can be deployed by aggressive, adversarial users to prevent communication among members of the targeted group. When group norms are threatened, elite or established members may also use hazing as a way of establishing boundaries. While such hazing is textual, this does not mean that is less aggressive or damaging than off-line hazing events (Honeycutt 2005). Tumblr's permeable group boundaries mean that users may become victims of silencing speech or face perceived threats to established norms. Although prior research suggests ways conduct has been controlled on different online platforms, none of it has been specific to Tumblr's architecture.
[6.4] Therefore, to extend prior research on tagging and social media within fandoms, as well as research in conduct control in online spaces, the following research questions emerge to assess if users regard Tumblr tags as community groupings or spaces in the absence of architecturally formalized groups:
[6.5] RQ1: How often do users of the "zoethian" and "sjips" tag use their portmanteau outside of the tag field?
RQ2: To what extent do users of the "zoethian" and "sjips" tags practice and reinforce the same content norms?
RQ3: To what extent is conduct control consistent between the "zoethian" and "sjips" tags?
7. Methods
[7.1] The entire population of posts tagged either "zoethian" (n = 283) or "sjips" (n = 64) between February 18, 2013, and March 4, 2013, were collected for content analysis. The unit of observation was the Tumblr post. Two weeks of posts were collected to allow sufficient time for all four YouTube channels to release multiple videos. During data acquisition, Yogscast Sjin and Yogscast Sips released new videos almost every day; Yogscast Rythian released videos several times a week; and Yogscast ZoeyProasheck released one or two videos a week. During the observed period, all four Yogscast members appeared in at least one YouTube video with their partner and at least one video by themselves. Tumblr posting volume appears to increase as videos are released on YouTube, although the current study is not intended to investigate this trend.
[7.2] Because the two tags are related through the larger Yogscast fandom, a small number of posts (n = 3) contained both tags, in which case they were counted twice in the data set, once for each tag. One post was eliminated during data collection because its tags were changed by the original poster within 48 hours of initial posting, thus removing the post from further tag searches.
[7.3] After receipt of institutional review board approval for nonhuman subject research, I collected posts manually within 24 hours of initial posting. Notes attached to posts were counted at the time of coding, approximately one month after initial collection. Eight posts were deleted between initial collection and final coding. In these cases, an earlier note count, collected between forty-eight and seventy-two hours of initial posting, was used. These posts could not be assessed for the reproach in notes and account variables because they were no longer assessable, and therefore they were excluded from analysis.
[7.4] Two coders assessed posts with a 10 percent (n = 35) overlap to calculate intercoder reliability. Posts were proportionally assigned to each coder at approximately a 4:1 (zoethian:sjips) ratio to ensure both coders would be coding posts from both sets. Coders had prior knowledge of how to use the Tumblr platform. Coders also had prior familiarity with Yogscast YouTube videos. Intercoder reliability for content and conduct variables was computed using Perreault and Leigh's (1989) reliability index (table 1). Intercoder agreement of 0.90 or above is considered sufficient agreement for coders to continue coding independently without additional overlap (Perreault and Leigh 1989).
| Variable | Perreault and Leigh's (1989) Reliability Index |
|---|---|
| Notes | 1 |
| Content variables | |
| Tag as adjective | 1 |
| Tag as noun | 1 |
| Post type | 0.95 |
| Visual sexual content | 1 |
| Verbal sexual content | 0.9 |
| Conduct variables | |
| Reproach in post | 0.96 |
| Reproach in notes | 1 |
| Account | 1 |
| Apology rituals | |
| Ritual 1 | 0.93 |
| Ritual 2 | 0.96 |
| Ritual 3 | 0.96 |
| Ritual 4 | 0.93 |
| Ritual 5 | 1 |
[7.5] Notes are the combined total of "likes" and "reblogs" by other Tumblr members. ("Replies" are now also included in "notes," but this was not functional at the time of data collection.) This is a publically visible number that appears alongside each post. Notes indicate some level of endorsement by other Tumblr users. Reblogging distributes content to additional users who may not be looking at a specific tag, and likes are publically visible when viewing all notes associated with a post, but they do not redistribute the post.
[7.6] Content variables were defined to code for specific content found within each post. References to the tag as an entity location (tag as noun) or descriptor (tag as adjective) were coded as present or absent. Posters sometimes address the tag as if it were a community, or at least as if ties exist between Tumblr members who use a specific tag with frequency. For tag as noun or tag as adjective to be present, a reference to "zoethian," "sjips," or the phrase "the tag" had to be present in the body of the post in addition to "zoethian" or "sjips" in the tag field. Posts were tagged as tag as noun if the phrase "the tag" or the words "zoethian" or "sjips" were used as a noun to refer to a place object, such as in the phrase, "I wrote zoethian." If the words "zoethian" or "sjips" were used as an adjective, such as the phrase, "Hello sjips people!," the post was coded as tag as adjective. By using the portmanteau outside of the tag field, the words "zoethian" and "sjips" are fulfilling a function other than personal information retrieval or social-oriented tagging. Using the portmanteau in the body of the post suggests that the tag is used as a proxy for a formalized group space that does not exist on Tumblr. Tumblr members may be "zoethian people" even though there is no formalized "zoethian" community space.
[7.7] All posts were coded as belonging to one of three post types: fan fiction, fan art, or general discussion posts, with the latter used for all posts not fitting into fan fiction or fan art. General discussion posts include talking about Yogscast videos (including reactions), posting screen captures from videos, and reposting Twitter conversations between Yogscast members.
[7.8] Sexual explicitness was coded as either visual (most often in the case of fan art) or verbal (most often in the case of fan fiction). Visual explicitness is present when "character's actions [or depictions] imply a sense of likely sexual intimacy" or behavior (Downs and Smith 2010, 276). "Sex talk," or verbal explicitness, includes dialogue or text between characters or posters that suggest sexual issues or sexual behavior at any point in time (Kunkel, Cope, and Biely 1999).
[7.9] For conduct variables, posts were assessed for reproach in post, reproach in notes, and accounts. The method for assessing these three variables differs somewhat from methods previously used when coding Usenet and message board content, although the definitions of the categories themselves are similar to those of Smith, McLaughlin, and Osborne (1997). A reproach is any statement that draws attention to an infraction of behavioral norms. This can range from simply pointing out that an infraction has occurred to cursing and threats. Because Tumblr does not work on a thread-reply structure like a message board or Usenet group, reproaches can be difficult to tie back to a single specific behavioral infraction. A reproach in post occurs when the body text of a post draws attention to an infraction, even if a specific failure event is not identified.
[7.10] A reproach in notes is present when one or more of the notes attached to a post was a reblog with the addition of a reproach. When reblogging a post, rebloggers have the option to add additional text that appears below the original post, potentially to add a reproach. These reproaches will generally not appear when searching for a tag because reblogging on Tumblr by default removes all the original tags from the reblog. Such tag stripping means that when a post is reblogged thousands of times, it does not appear thousands of times when someone searches for a particular tag. In the case of conduct control, it means that the reproach might be more concealed, although still completely public, than it would be on a Usenet group or message board. If a reproach in notes occurred, coders also coded the presence or absence of an account. Accounts are excuses or justifications for the failure event (Schlenker and Darby 1981; Smith, McLaughlin, and Osborne 1997). An account had to follow a reproach in notes in order to be correctly identified as a response to particular reproach. More general admissions of guilt were captured using the apology rituals variables.
[7.11] Apologies were considered as a means of regulating conduct. An apology indicates that the poster acknowledges that an infraction has occurred, whether or not others have identified it as a failure event. Apologies could consist of any combination of five rituals (Goffman 1971; Schlenker and Darby 1981), as follows. Ritual 1 is a "statement or apologetic intent" (Schlenker and Darby 1981, 272) such as "I'm sorry." Ritual 2 is an acknowledgment that the poster feels bad about the infraction: "I feel awful about this." Ritual 3 offers restitution: "I'll make it up to you next time I post." In ritual 4, the poster disparages him- or herself for misbehavior, including self-deprecation: "I know my drawing is awful." Ritual 5 involves directly asking for forgiveness: "Please forgive me." Apologies did not need to follow a reproach in post or reproach in notes. Any time one of these five rituals was present, an apology occurred, even if the offended party or recipient of the apology was unclear or imaginary.
8. Results
[8.1] For notes, "Zoethian" posts average 21.53 notes per post and "sjips" posts average 63.78 notes per post. Posts using the "sjips" tag receive significantly more notes than those using the "zoethian" tag (t = 4.393, p < .001). Notes are used in the analysis below as a proxy for popularity and endorsement.
[8.2] The first research question asks how frequently "zoethian" and "sjips" users use their specialized portmanteau outside the tag field. There is no statistically significant difference (Kunkel, Cope, and Biely 1999) between the two tags when considering either tag as noun or tag as adjective. Users of the tags "zoethian" and "sjips" use these specialized tags for reasons other than information retrieval or socially oriented reasons (Lin and Chen 2012). "Zoethian" and "sjips" users use their respective specialized fan tag equally as a noun (χ2 = 0.04) or verb (χ2 = 1.06). They may be using it to signal to other Tumblr users that they are part of a nascent community.
[8.3] The second research question asks if users of the "zoethian" and "sjips" tags enforce the same content norms. There were statistically significant differences between both visual and verbal sexual explicitness, and for the distribution of fan art and general discussion posts. Table 2 shows the percentage of posts in each tag that use the specialized fan tag as a noun or adjective and both verbal and visual sexual explicitness variables. Users of the "sjips" tag post significantly more sexually explicit content, both visually (χ2 = 31.40, p < .001) and verbally (χ2 = 30.32, p < .001). Not only are posters using the "sjips" tag posting more sexual content than users of the "zoethian" tag but they are endorsing that content with notes.
| Tag | Tag as Noun | Tag as Adjective | Visual Sex | Verbal Sex |
|---|---|---|---|---|
| ***p < .001. | ||||
| Zoethian (het pairing) | 7.07% | 12.37% | 1.14% | 4.24% |
| Sjips (slash pairing) | 7.81% | 17.19% | 17.18% | 25.00% |
| Pearson's chi-square | 0.04 | 1.06 | 31.40*** | 30.32*** |
[8.4] Table 3 summarizes the distribution of posts by post type. Users of the "zoethian" tag tend to engage in more general discussion (65.02 percent) than do users of the "sjips" tag (31.25 percent) (χ2 = 24.57, p < .001). Users of the "sjips" tag post more fan art (43.75 percent) than do users of the "zoethian" (17.67 percent) tag (χ2 = 20.38, p < .001).
| Tag | Fan Fiction | Fan Art | General Discussion |
|---|---|---|---|
| ***p < .001. | |||
| Zoethian (het pairing) | 17.31% | 17.67% | 65.02% |
| Sjips (slash pairing) | 25.00% | 43.75% | 31.25% |
| Pearson's chi-square | 2.03 | 20.38*** | 24.57*** |
[8.5] Table 4 summarizes reproaches and accounts as a percentage of all posts in each tag. There is a statistically significant difference between groups for the variable reproach in notes (χ2 = 5.17, p < .05). There is no statistical difference between groups when considering reproach in post (χ2 = 1.38) and account (χ2 = 0.63).
| Tag | Reproach in Post | Reproach in Notes | Account |
|---|---|---|---|
| *p < .05. | |||
| Zoethian (het pairing) | 2.12% | 8.30% | 1.08% |
| Sjips (slash pairing) | 0 | 0 | 0 |
| Pearson's chi-square | 1.38 | 5.17* | 0.63 |
[8.6] Although the "zoethian" tag had significantly more reproaches, Tumblr members using the "sjips" tag exhibited more apologetic behaviors. Those using the "sjips" tag used more self-deprecating remarks in their posts, engaging in apology ritual 4 (χ2 = 9.49, p < .01). Mostly these self-deprecating remarks stated that the posters considered themselves to be poor writers or artists when posting fan fiction or fan art. Table 5 summarizes apology rituals by tag.
| Tag | Ritual 1 | Ritual 2 | Ritual 3 | Ritual 4 | Ritual 5 |
|---|---|---|---|---|---|
| **p < .01. | |||||
| Zoethian (het pairing) | 7.77% | 1.06% | 3.53% | 13.78% | 1.06% |
| Sjips (slash pairing) | 3.13% | 3.13% | 3.12% | 29.69% | 0 |
| Pearson's chi-square | 1.75 | 1.57 | 0.03 | 9.49** | 0.68 |
9. Discussion
[9.1] In this study of two tags used by subcultural Yogscast fans on Tumblr, differences in content and conduct between the two tags may be based on norms developed by the users of each tag in the absence of Tumblr's architectural limitations or opportunities, or the influence of moderators. Although this research is exploratory, there is evidence that online communities will emerge from available architecture, even if the explicit intent of the given architectural element may not have been community building. This is consistent with more ethnographically informed findings from boyd (2006) and Davis (2010) concerning user profiles on MySpace. Users of the "zoethian" and "sjips" tags use specialized fandom tags for reasons other than personal organization or sharing resources with others (Lin and Chen 2012). When the specialized tag is moved from the tagging field and into the body of the post, either as a noun (Zoethian, 7.07 percent of posts; Sjips, 7.81 percent of posts) or an adjective (Zoethian, 21.37 percent of posts; Sjips, 17.19 percent of posts), it serves a function unrelated to organization and retrieval. What may have begun as an organizational term, allowing fans of a specific pairing to distribute and locate posts that might interest them, becomes a way to suggest a particular community affiliation.
[9.2] The content and conduct difference between the two groups is striking. Tumblr members using the "sjips" tag post more sexually explicit content to greater endorsement, and their posts receive more notes. Posts that use the "sjips" tag receive fewer reproaches, but the posters are more inclined to apologize. Because the use of apologies differs across cultures (Maddux et al. 2011), it is reasonable to infer that different communities may develop different strategies for using apologetic remarks, such as preemptively addressing failure events (Williams, Morgan, and Cameron 2011). Because reproaches appeared to be infrequent in these tags, apologies may substitute as a mechanism for behavioral control. Instead of waiting for the (nonexistent) moderator to determine if a post is not allowed, tag users may use apologies to prevent conflicts from arising. Without the threaded structure of a message board, where reproaches and accounts are clearly linked to the failure event, castigating others may be more difficult on Tumblr. Any castigation is posted to the reproacher's blog and only appears as one of many notes attached to the original post. While the original poster may be aware of committing an infraction, the reproach may not be as visible to others compared to message board architecture. Unprompted apologetic rituals may be an adaptation of conduct control unique to Tumblr and its architecture.
[9.3] This study raises questions about the relationships between content producers and fans, as boundaries between the two groups are rapidly changing and being renegotiated. While research has looked at the legal gray areas that arise when distributing fan fiction and fan art (Chander and Sunder 2007; Jenkins 2006; Tushnet 2014), additional questions arise when producers and fans overlap on the same online space. All four Yogscast members studied maintained Tumblr, Reddit, and other social media accounts to interact with fans. One reproach post in the sample addressed the issue of the perceived difference between how Tumblr users treat fan art and fan fiction. Fan art is perceived as receiving more notes. Rythian reblogged this post and added additional text, supporting the idea that writing is also an art and takes a great deal of time and skill. This post garnered 156 notes—well above the average 21.52 notes per post for the "zoethian" tag. Zoey also used the "zoethian" tag once during the observed period. Her post received 204 notes—again, well above the 21.52 average for the "zoethian" tag.
[9.4] Marwick and boyd (2011) describe the context collapse that occurs when perceived online audiences come into contact. What one user might deem appropriate material for Facebook may not be appropriate content for Twitter. Likewise, for Yogscast fandom, Tumblr content is not considered appropriate for Reddit, where semiofficial interaction between Yogs and Yognau(gh)ts occurs. The fandom adage "don't like, don't read" is extended to the subjects of the fan works themselves. Although Zoey and Rythian are supportive of continued fan works with themselves/their characters as the focus, other Yogs have expressed discomfort with romantic or sexual fan-produced content. The imagined audience (Marwick and boyd 2011) for Yogscast fandom had to include the Yogs themselves, given the limitations of Tumblr's architecture (no privacy settings), as well as technological affordances built into the platform, especially reblogging, meaning that Yogs might see fan content about themselves on their Tumblr dashboard, even if the original poster had meticulously tagged it.
[9.5] This closeness between content producers and fans renegotiates behavioral norms and privacy. While the relationship between media entities and fans has been in a state of flux for some time (Busse and Hellekson 2012; Chin 2014; Jenkins 2006; Jenkins, Ford, and Green 2013), the Yogscast's direct use of fan artists, as well as their reading of fan fiction, signals a complicated relationship between content producers and fans that may account for observed patterns of unprompted apologies in the "sjips" tag.
[9.6] Fandom scholarship has advocated for protecting fans' right to privacy even though their work may technically be publically accessible (Busse and Hellekson 2012; Nielsen 2016; Reid 2016). Fan works occupy a semiprivate location online, where consumption by other fans but not the general public is assumed (Busse and Hellekson 2012). Elm (2009) views privacy as a spectrum; simply because a piece of text is publically accessible online does not mean that it was intended for a broad audience. For this reason, I have not directly quoted any texts or used any user names, save for those of the Yogs themselves. Fan scholars advance the idea that researchers should ask permission of fans when directly or indirectly quoting their work (Busse and Hellekson 2012; Fathallah 2016), but this may be untenable when handling large data sets. Indeed, as this study demonstrates, it becomes near impossible to contact the writer of every contribution, thus raising additional questions regarding the ethics of quantitative fandom studies.
10. Limitations and conclusion
[10.1] This exploratory project comes with a number of limitations. There may be a different perceived level of involvement by the content producers between tags, which may account for some of the difference observed in content and conduct. Either interviewing or surveying Tumblr users about what considerations they take into account when determining if they will or will not post a specific fan fiction or fan art work, and how they choose to tag that post, would allow for a richer examination of the norms developing on Tumblr. Online surveillance should not be considered unidirectional, and social media users are always participating in their own surveillance and the watching of others (Albrechtslund 2008). This can be addressed by querying fans on Tumblr if they think about whether a content producer may be reading or looking at their fan work when they decide to post.
[10.2] This study expands the existing research on tagging, online platforms, and fandom. Tags, while often considered quantitatively, are generally considered for their social potential only insofar as distributing resources or exerting influence over what tags are used frequently. This research considers the possibility that those who use a particular tag may come to constitute a community that expresses particular norms when it comes to content and conduct. Further, few studies consider the architectural possibilities and limitations of Tumblr. Through an exploration of the practices of these two specialized fandom tags, this research addresses how the Tumblr platform is actually used by members, including potentially different purposes, depending on their needs. Finally, as a content analysis, this research provides some strategies for considering fandom communities in a quantitative manner by coding for specific content, such as sexual explicitness, as well as how fans interact with one another, such as apologizing to avoid a negative response from others. It adapts the existing social science method of content analysis for an online space where content is diverse and continually updated. As norms develop on any online platform, it is important not only to investigate the norms themselves but also to understand the reasons why such norms develop and then stabilize.