“How deep do I have to cut?“: Non-suicidal self-injury and imagined communities of practice on Tumblr

Abstract

This paper concerns itself with the study of non-suicidal self-injury (NSSI) content on Tumblr. Adding to existing valuable studies on social media and NSSI, we contribute an anthropological understanding of the communal formations and relationships between people who use social media to express their thoughts and feelings about NSSI. Using online ethnography as a method, we approach our data from the perspective of someone who is new to NSSI Tumblr and is learning how to engage with other people who self-injure sharing content on the site. We argue that people who share and interact with NSSI content on Tumblr form part of imagined communities of practice, through which they create shared meaning and interpretations of their experiences. Our results suggest that Tumbleloggers who engage with NSSI content draw on a variety of communal practices and norms to create a non-judgmental space away from societal stigma. As such, future research can benefit from an approach centered around communal practices to understand how people who self-injure connect and communicate on social media.

1. Introduction

Previously understudied, research on non-suicidal self-injury (NSSI) has seen an uptick since the early 2000s (Klonsky et al., 2014; Seko et al., 2015). Though there now are some ethnographic (Csordas and Jenkins, 2018; Lavis and Winter, 2020; Winter and Lavis, 2020) and auto-ethnographic studies on NSSI (Stirling, 2020; Stirling and Chandler, 2021), anthropological research in this area is still rare. If it does find mention, it is often subsumed under broader work on body modification (Sargent, 2003). However, “body ‘modification’ is not ‘self-injury’ as long as it is socially normative” (Sargent, 2003, p. 26) (cf. Bareiss, 2021b; Lewis and Mehrabkhani, 2016), and self-injurious behavior is highly stigmatized (Seko et al., 2015; Staniland et al., 2021; Steggals et al., 2020). NSSI is defined as the act of intentionally injuring one’s own body tissue through cutting, scratching, burning, or bruising, for non-suicidal purposes (International Society for the Study of Self-Injury, 2021). While “enacted without intent to die” (Staniland et al., 2021, p. 312), NSSI is also often an indicator of “increase[d] risk for both suicidal desire and ability” (Hamza and Willoughby, 2016, p. 414). It is colloquially known as self-harm, a term commonly used among people on social media who engage in NSSI. While we did consider using ‘self-harm’ as a term to reflect the language of people with lived experience, we decided to speak of NSSI. ‘Self-harm’ as a colloquial term encompasses other behavior that does not result in immediate physical injuries, such as disordered eating or substance abuse (Shanahan et al., 2019), which we do not include in the scope of this paper. Additionally, we endeavored to avoid value-laden language which could increase the risk of a potentially vulnerable group being subjected to misinterpretation or bias.

The internet has afforded those who self-injure, especially young people, a relatively safe space to talk about the practice anonymously and without judgment (Alvarez, 2021; Hilton, 2017; Seko et al., 2015; Seko and Lewis, 2018). Ongoing research on the relationship between NSSI and social media has provided great insights on the development of niche spaces in which people come together to share their experiences and struggles. Researchers such as Lavis and Winter (2020), Moreno et al. (2016), Seko and Lewis (2018), Seko et al. (2015), and Shanahan et al. (2019) have contributed greatly to this field of study. However, there is a research gap in knowing what exactly (young) people post in online NSSI spaces and discussions (Lavis and Winter, 2020; Shanahan et al., 2019), and how NSSI is communicated or functions as a means to communicate (Alvarez, 2021; Bareiss, 2021b). We also do not know much about potential communal aspects of NSSI spaces on social media or what such communities may mean to people who self-injure (Alvarez, 2021). Anthropology with its “grasp of the context specific nature of social processes” (Lambert and McKevitt, 2002, p. 211) contributes to the study of these questions.

In this paper, we assess NSSI content on Tumblr from an anthropological perspective and argue that people sharing and engaging with it form part of an imagined community of practice. Three research questions formed the pillars of this central argument:

1. Which practices do Tumblr users (Tumbleloggers) who self-injure engage in that indicate a sense of social interaction and community formation?

2. Which practices and norms do Tumbleloggers who self-injure establish to create shared meaning and mutual understanding in their imagined community of practice?

3. Which other communities of practice (CoP) do Tumbleloggers who self-injure interact with and/or distance themselves from?

These questions allow us to explore four areas of NSSI content on Tumblr: (a) community formation, (b) social practices, (c) language use, and (d) intercommunal exchanges and in-group/out-group delineations. In using an ethnographic approach, we explore this from the perspective of a novice Tumblelogger who self-injures and is learning to engage with others who use this space; in other words: we are trying to understand how someone may enter and become entangled in an imagined community of NSSI practice. Note that being a novice Tumblelogger who self-injures does not necessarily equal being new to NSSI as a practice. Studies have shown that people who engage in online spaces about NSSI are likely to self-injure already (Lavis and Winter, 2020; Shanahan et al., 2019).

Our hope for this contribution is that it may help NSSI researchers, health practitioners, families and loved ones of people who self-injure identify potential signs of NSSI activity so that they may intervene, but also intervene in a more empathetic way as they learn how participating in an online community of people who share the same experiences may help individuals feel supported, understood, and less isolated.

2. NSSI blogging on Tumblr: imagined communities of practice

We tend to think of communities as place-based (e.g. towns, neighborhoods) or attribute-based (e.g. sports teams, academics), but not all communities are based on shared space or demographics of their members. Some instead are based on shared actions or interests, especially in online “interest-driven communities” (boyd, 2014, p. 39), where users can easily curate the content with which they wish to engage. We use the term ‘communities of practice’ (Cochrane, 2017; Eckert, 2006; Eckert and McConnell-Ginet, 1992; Gruzd et al., 2016) or ‘CoP’ to describe these group formations. In NSSI contexts, centering practices rather than identities can help relieve some societal stigma. A practice-centered approach fits the recommendations of the International Society for the Study of Self-Injury (ISSS) (International Society for the Study of Self-Injury, 2021). We should speak of “people who self-injure” (Lewis, 2017, p. 204), rather than ‘self-harmer’ or ‘self-injurer’, to avoid equating the person with the practice. This is not to say that people who self-injure do not sometimes self-identify with these terms; it rather cautions us as researchers to challenge notions of fixed or pejorative identities often associated with NSSI and avoid using terms that perpetuate such notions.

Applying a practice-centered approach to the wider community is similarly helpful. Rather than treating communities of NSSI practice as a homogenous group, we suggest paying attention to the “commitment to shared understanding” and “mutual sense making” (Eckert, 2006, p. 683) that people who share NSSI content online display through their shared practices. These practices include not just NSSI itself, but also wound care, coping mechanisms such as creative expression and humor, replacement behaviors, practices of (and advice for) avoiding detection or hospitalization, the language used to talk about NSSI, as well as modes of communication and content sharing (e.g. platform choice, content curation, unspoken and spoken rules in engaging with or hashtagging NSSI content). Practices and experiences thereof can vary. A vital asset of the CoP model is its tolerance for “the fluidity of social space and the diversity of experience” (Eckert, 2006, p. 684) and the challenge it poses to generalizations and assumptions of group homogeneity. As such, we believe that the CoP model is suitable for NSSI research because it challenges us to understand the roles that self-injury, related practices, and senses of community and belonging play in the lives of people who self-injure.

Another important aspect for NSSI studies is “the place of the community of practice in the larger social order” (Eckert, 2006, p. 683). CoP exist in relation to other communities and within the context of societal discourse. As such, the place of communities of NSSI practice in the social order may, again, be affected by stigma. Community members may develop various attitudes and beliefs about themselves and others based on experiences within and outside the community (Staniland et al., 2021). Therefore, the community of practice is a place where identities can be shaped and re-configured. This also may involve the “development of a style – including a linguistic style – that embodies these interpretations” (Eckert, 2006, p. 683). Given the stigma and the monitoring of NSSI content on social media, the language used in NSSI spaces online is in constant flux and involves a somewhat covert vocabulary to keep the community out of the moderation radar.

The internet, and social media especially, can be helpful tools to trace networks and language use within the vast existence of NSSI content. On the other hand, community practices involve some level of secrecy, as well as external restrictions and limitations of content-sharing that differ from social media platform to platform. Additionally, social interactions between members are either infrequent or hidden from public view (e.g. direct messaging, private groups). This requires some adjustment in the theoretical framing. In a study on online discourse among wheelchair users, Leslie Cochrane finds that “bloggers demonstrate remarkably similar sense-making around disability practices” (2017, p. 160) despite (presumably) not interacting with one another. Cochrane uses the framework of imagined communities of practice to address this lack of interaction, combining the CoP concept with Benedict Anderson’s (2006) imagined communities. Members of imagined CoP “may not interact with other members of the community or use identical practices, but they may still recognize themselves and others as belonging to such a community and sharing common experiences” (Cochrane, 2017, p. 155). This approach allows us to speak of a community of people who engage in and discuss similar practices but whose levels of interactions in the community spaces vary. Whether their posts are socially oriented or self-motivated (Seko et al., 2015), what unites them is the practices they engage in and the communal norms and patterns that they develop in expressing, sharing, and curating their thoughts on the matter. In other words, imagined communities of NSSI practice are “constructed through discourse” (Bareiss, 2021a, p. 156).

Hashtagging, hashtag creation and modification are skills and practices central to communities of NSSI practice. The community of practice is a site for “process[es] of identity and linguistic construction” (Eckert, 2006, p. 685). While this is true for any community of practice, linguistic construction in communities of NSSI practice is facilitated by rigorous hashtag and content moderation. Researching eating disorder (ED) content ‘beyond the hashtag’, Ysabel Gerrard argues that “these communities can only occupy a certain kind of space on social media: one that is located at the margins” (2018, p. 4493; emphasis in original). Posts and blogs in popular hashtags may be deleted, leading bloggers to invent creative alternative solutions to maintain the community and its content.

3. Methodology

We use online ethnography to study imagined communities of NSSI practice on Tumblr. Online ethnography is “suitable for engaging with online interactions and communities” (Lavis and Winter, 2020, p. 843). While online community studies were popular in the early 2000s, their conceptualization pre-dated the advent of social media (Gruzd et al., 2016) and specific sites’ unique affordances that may allow for different community formations. We consider Tumblr to be suitable for our study because it restricts sensitive hashtags only minimally, which facilitated our access to NSSI-related posts. Moreover, Tumblr’s anonymity options enable people who post about socially unsanctioned content to do so without revealing identifiable information.

3.1. The field site: Tumblr

Referred to as the “forgotten” social media site, Tumblr began operation in 2007 as a popular blogging platform (Anderson, 2015; Attu and Terras, 2017). Although Tumblr’s popularity in the social media landscape has decreased in recent years, following its history of company changes and content bans, the site has by no means disappeared: As of 2021, Tumblr hosts over 530 million blogs on every conceivable topic (Tumblr, n.d.). Blogs hosted on Tumblr are called “Tumblelogs” and the bloggers behind them are referred to as “Tumbleloggers”, respectively. Offering a high degree of anonymity, all that is needed to create a Tumblr account is an email address, which remains private, and a unique username, which functions as the public blog handle. While users are not required to disclose any identifiable information at all on their blogs, studies show that it has become common practice among Tumbleloggers to disclose aspects of one’s identity in the blog title, blog description, optional “about” pages, pinned posts or hashtags (Oakley, 2016). Despite the anonymity options, Tumblr encourages community formation and aims to let “your interests connect you with your people […] in millions of communities across millions of #tags” (Tumblr, n.d.) . Hashtags provide a commons for users to create, access and engage with content on Tumblr. Most commonly, such engagement takes place on a user’s ‘dashboard’, the homepage where they would find posts from Tumblelogs they follow as well as Tumblr-generated recommended posts, blogs, and hashtags. Users can like, reblog (with or without comment) and reply to posts. These interactions will be publicly visible to others as so-called ‘notes’ that are counted at the bottom of the post. Invisible interactions include following blogs or hashtags of interest, and direct messaging between users. A mixed feature is the ‘ask box’: If enabled by the blogger, other users can access their ask box to submit questions or comments identifiably (using their username) or anonymously, and the reply can similarly either be posted publicly on the user’s Tumblelog or be returned privately to the person who asked. Because Tumbleloggers can stay anonymous, the “relatively ‘asocial’ affordance of Tumblr has made it particularly conducive to performing non-normative subjectivities” (Seko and Lewis, 2018, p. 182), including NSSI. Adding to the attraction of Tumblr as a host for NSSI content, the site is most popular among young people between 13 and 24, which mirrors the age range at which the risk of engaging in NSSI is the highest (Klonsky et al., 2014; Lewis and Heath, 2015; Seko and Lewis, 2018).

Tumblr officially bans any activity which “actively promotes or glorifies self-harm”, defined as “urg[ing] or encourag[ing] others to: cut or injure themselves; embrace anorexia, bulimia, or other eating disorders; or commit suicide rather than, e.g., seeking counseling or treatment, or joining together in supportive conversation with those suffering or recovering from depression or other conditions” (Tumblr, 2020). The community guidelines provide a link for others to report NSSI activity, though they offer no transparency as to how decisions are made on the matter or whether this is the only means through which Tumblr removes NSSI content. This user-centric approach to content moderation primarily via reports differs from other social media sites like Instagram, where searches for known NSSI hashtags return the most popular posts but not any recent ones (Seko and Lewis, 2018). On Tumblr, search results will be shown in full and can be ordered either by most popular or by most recent. This provides a wealth of data for research purposes but means that Tumbleloggers will be exposed to the entirety of the content returned for their searches, including graphic imagery and descriptions of NSSI, unless and until posts or blogs are deleted by Tumblr.

3.2. Ethical considerations

Prior to commencing our research—which, apart from the social media research discussed in this paper, includes traditional research with human participants—we submitted a protocol (number 118114) to our institution’s Research Ethics Board. At that time, we were notified that we did not require board approval for the social media-based research, as all gathered information was in the public domain. However, we believe that there is an important distinction to be made between research ethics approval and ethical research. NSSI is a sensitive topic and therefore we followed precautions to not harm the people whose posts we collected. Central to ethical social media research is the belief that research should respond to four main areas: private vs. public data, anonymity of participants, informed consent, and risk of harm (Townsend and Wallace, 2016; Hennell et al., 2020). The Tri-Council Policy Statement 2 (Canadian Institutes of Health Research, 2018), the guiding document for research ethics in Canada, does not mention social media research specifically, but distinguishes between the public domain with or without reasonable expectation of privacy. We reviewed Tumblr’s privacy policy (Tumblr, 2021), which informs its users that “anything you publish is publicly accessible unless you have explicitly selected otherwise.” The British Psychology Society’s (2021) Ethics Guidelines for Internet-Mediated Research advise that users may nevertheless consider their accessible content private, which blurs the lines of ‘reasonable’ expectations of privacy. We took the view that users know their posts to be public because a blog is an intentional act of connecting with people and sharing information (boyd, 2014, p. 7) and the use of hashtags publicizes their posts to a wider audience (Townsend and Wallace, 2018).

Anonymity of participants was ensured both by Tumbleloggers and by us. Tumbleloggers who self-injure did not typically disclose identifiable information on their accounts, often actively de-identifying their blog sites by choosing pseudonyms for themselves and other people they would mention in posts. Only a few would post identifying images or data. Additionally, following the example of similar studies (Gerrard, 2018; Lavis and Winter, 2020; Seko and Lewis, 2018), we do not disclose any usernames/blog URLs, specific hashtags, images, or otherwise identifiable information in this article. Passages that cite directly from Tumblelog entries are altered slightly to prevent possible identification through text searches.

As we did not engage in any direct communication with Tumbleloggers, we did not seek traditional informed consent. We sought guidance on this issue through the framework for ethics on social media research developed by Townsend and Wallace (2018) by first reviewing Tumblr’s terms and conditions. Here, we found no specific policy on informed consent and research. As a result, we opted to connect the informed consent with the larger issue of risk of harm. Because the posts represent actions that have already happened and are being documented in a public sphere, the process of seeking informed consent may increase risk to bloggers, as the process of contacting them, engaging with them, and discussing risks could lead to a triggering event. Admittedly, this is an area that requires greater overall research that is beyond the scope of this paper.

Along with triggering risk, potential harms include reputational damage, increased risk of targeted negative response, and changes to how the individual is treated (Hennell et al., 2020). We believe these risks are largely mitigated by the blogger-initiated efforts to de-identify themselves and our additional anonymization measures described above.

3.3. Data collection

Data was collected over two months in the summer of 2021. By using a mix of systematic search and snowball sampling (NSSI content recommended by Tumblr), we manually compiled 182 popular hashtags that show the range of possibilities in becoming exposed to NSSI-related content. However, it is important to note that “a hashtag cannot tell the whole story about a given phenomenon” (Gerrard, 2018, p. 4494), especially in marginalized communities where hashtag use may be closely monitored. Indeed, some of the hashtags we collected were explicitly related to NSSI, others used covert language, likely to avoid detection (Seko and Lewis, 2018). Drawing from Gerrard’s (2018) methodology, we followed our collected hashtags to allow Tumblr to generate relevant data without ever following any blogs or liking their posts. Tumblr recommended posts, blogs and hashtags related to the followed hashtags, creating an experience that might resemble a Tumblelogger’s first venture into NSSI content. We kept fieldnotes in the form of observation logs to document this experience, which allowed us to understand how someone new to NSSI may start engaging with it, and how Tumbleloggers connect with others, learn unspoken rules, keywords, and language practices.

We compiled 70 Tumblr accounts, including blog descriptions and recent hashtag use, if any, to get a feel for commonly self-disclosed demographic info, hashtag popularity, as well as keywords that could signal NSSI content to other Tumbleloggers. We chose blogs that featured in the NSSI-related hashtags we searched and blogs that were recommended by Tumblr. At least one of the most recent 10 posts had to be related to NSSI for the blog to be included in our data. We revisited each of these 70 blogs three months after initial collection to see whether they were still active/online. Roughly one third of the compiled blogs had been deleted, which suggests high blog fluctuation within the NSSI community on Tumblr.

Other than hashtags and blogs, we collected data from 241 posts on Tumblr. 106 of them were systematically collected in four NSSI-related hashtags. We used a function on Tumblr whereby posts are sorted starting with the most recent upload, rather than the most popular post, so that our data would reflect a wide range of post types rather than widely shared, popular entries. In more frequently used hashtags, we limited our post collection to the past day or two to keep the sample sizes comparable. The four hashtags in which we collected data were not ‘flagged’ by Tumblr, meaning that users can follow the tags to keep up to date with them. Flagged tags are not followable and amounted to only 19 out of an overall 182 collected hashtags. Another 135 posts were collected randomly as part of logging observations, navigating Tumblr like a user newly curious about NSSI content. Some of these posts were recommended to us on our dashboard, generated by Tumblr according to the hashtags that we followed, others we spotted on the blogs or in the hashtags also recommended by Tumblr. As such, this part of the collection process resembled a snowball sampling approach.

3.4. Analysis

We sorted early emerging data on an Excel spreadsheet into categories (e.g. advice for practice, memes, photos of injuries) using thematic analysis. We screenshot posts that we found and compared them to other collected posts. We used this categorization to make sense of the variety of content before developing a 5-level data analysis scheme based on the emerging data. Five levels were systematically analyzed for each post: (1) whether the post was mainly self-oriented or socially motivated (Seko et al., 2015), (2) position or attitude towards NSSI and/or recovery, (3) visual content, if any, (4) textual content, if any, and (5) hashtags used to classify the posted content, if any. Visual content on Tumblr can include photography or videography, and we classified textual content as either the blog entry text itself or any words and phrases occurring within the visual content. As such, our 5-level analysis easily encompassed multimedia posts. To ensure interpretative validity throughout the analysis process, we shared our findings within the team and cross-checked individual interpretations.

4. Results

4.1. The Tumblr algorithm: entering new communities of practice

Our ethnographic approach entailed following NSSI-related hashtags and keywords on Tumblr, as Tumbleloggers would do if they wanted to keep up to date with new content relevant to their interests. In an observation logged only three days after following our curated hashtags, we noted that our dashboard now suggested blogs for us to follow, the majority of which posted NSSI content. The recommended blogs that did not (recently) post any NSSI content were at the very least related in content, for instance posting about trauma, gore or eating disorders (ED). Tumblr also suggested more related hashtags, which we then followed to add to our collection and to prompt the algorithm to generate more data of the sort.

Eventually, our dashboard started showing us “recommended posts” that were more attuned to the content we had been searching. Posts with relevant content and hashtags appeared, most of them about ED but a small portion about NSSI as well. We took the oversaturation with ED as an indication that there is a lot of overlap between the communities of ED practice and NSSI practice, and that many bloggers were affected by both. Some of the NSSI posts that Tumblr recommended to us were not labelled with any hashtags, making it abundantly clear that someone who would even just tentatively start engaging with NSSI Tumblr may experience a snowball effect of related content pouring into their suggestions.

An observation that we found troubling was that the Tumblr-generated support messages for popular hashtags relating to NSSI stopped showing up about a month into our data collection and continued to not be shown as frequently. Support messages still popped up consistently for keywords (e.g. searching for selfharm instead of #selfharm) but not for hashtags. It seemed as if Tumblr recognized our continuous search for these topics and accordingly adjusted its pop-up messages. This may have implications for Tumbleloggers’ awareness of and access to support and crisis hotlines.

4.2. Blog description and self-disclosure as indicators of belonging

While Tumblr does not ask its users to disclose any identifiable biographic or demographic data on their Tumblelogs, many do choose to tell readers something about themselves. This information is typically found in the blog description at the top (or occasionally to the left) of the blog. Some Tumbleloggers also ‘pin’ an about me type of post to the top of their Tumblelog. The information on display varies from blog to blog, but often includes age, personal pronouns and/or gender identity, and romantic and/or sexual orientation. We noted a tendency toward young age (disclosed ages ranged from 15 to 25 in our blog compilation) and a variety of personal pronouns and sexualities that pointed toward occurrence of NSSI across gender and sexuality spectrums.

Another recurring piece of information in blog descriptions were keywords related to mental illness, traumatizing events, or disordered behavior. The keywords often come with trigger warnings. A blog showing the information “TW for BPD, PTSD, CSA, ED & SH”, for instance, would be run by someone who has (diagnosed or undiagnosed/self-diagnosed) borderline personality disorder (BPD), post-traumatic stress disorder (PTSD), experienced childhood sexual abuse (CSA) and struggles with an eating disorder (ED) and self-harm/NSSI (SH). They issue a trigger warning (“TW”) so that other Tumbleloggers can proceed with caution when reading or following their blog. However, this self-disclosure may serve multiple functions: Not only do Tumbleloggers warn potential readers about the content of the blog; they can also signal belonging to CoP on Tumblr. When Tumblr recommended Tumblelogs for us to follow, we could quickly determine whether they posted NSSI content by scanning the blog description and pinned post, if available. If the Tumblelogger mentioned “sh”, “s/h” or fully spelled out “self harm” on their blog, they would likely post NSSI content. Relevant keywords that pointed to potential NSSI as well were “tw” or “cw” (trigger/content warning), “please block, don’t report” and “vent acc(ount)” (see next section). Since we, as newcomers to the imagined community of NSSI practice, picked up these cues quickly, it is fair to assume that Tumbleloggers who self-injure would also be able to recognize like-minded others and build their network accordingly.

Some blog names were also indicative of NSSI activity. Of the 70 accounts we compiled, eight names showed thematic connections to self-harm, blood, cuts, scars, or blades, and twelve to death and decay. Nine alluded to physical exhaustion or illness, another nine to emotional distress, and five to mental health obstacles. Finally, eight names included pejorative language toward the Self. Many of these overlapped, resulting in names to which more than one of these categories apply. Just as the keywords in blog descriptions, these creative monikers would signal to others that this Tumblelogger participates in an imagined community of NSSI practice.

4.3. Self vs. social: motivations, attitudes, and interactions in NSSI Tumblr posts

Though we could not always make a clear-cut distinction between self-oriented and socially motivated content (Seko et al., 2015), identifying inward or outward elements of a post helped us understand interactions between members of an imagined community of NSSI practice, and potential rules for or shared interpretations of such interaction as well. Most posts that we categorized as self-oriented resembled entries you would find in someone’s diary. Their prevalence among our recorded posts means that they were either the most posted type of entry or the least likely to be reported and removed by Tumblr, or a combination of both. Diary-type posts would include descriptions of someone’s day, emotional state or relationships with other people and typically be written from a first-person perspective either without a clear audience or addressed to a specific person in the poster’s life. A great deal of emotional turmoil, pejorative self-perceptions, or resentment toward others (said to be responsible for the poster’s pain) was present in the content of these posts, exemplifying how NSSI is used as a tool for emotional regulation (Klonsky et al., 2014).

The practice of sharing posts of this nature on Tumblr is commonly described as “venting”. Common preambles such as “alright so vent” or “vent in tags” indicated that the following post (or its tags) contained personal, distressing content. Most of these personal story posts had few notes compared to other NSSI content. Under the occasional post, however, people would leave replies to the effect of “same”, “please stay safe” or “I’m here if you need to talk”, suggesting that even initially self-oriented posts bear the potential for social interaction because they are shared with others. Most people extended kindness and offers of support in their comments, exemplifying why NSSI bloggers may turn to imagined communities of NSSI practice to express themselves without judgment and to feel cared for and welcomed.

Other self-oriented posts showed pictures or videos of open wounds, bandaged injuries, or scars. This type of content was among the quickest to be reported and deleted by Tumblr, which we confirmed by reviewing collected data periodically. These posts also created the most tension when tagged or labelled in a way that was deemed inappropriate: When correcting others or asking for trigger warnings, Tumbleloggers would hint at certain informal community norms, for instance being transparent about posting sensitive content or avoiding the cross-tagging of NSSI and ED content so as to not expose people outside the CoP to in-group content. Evidently, participating in various CoP on Tumblr entails familiarizing oneself with the proper hashtags for each community and using them responsibly, that is: avoiding (re-)traumatizing others with potentially harmful content. Properly indicating such content via trigger warnings is seen as a courtesy to those who filter out content to which they may be sensitive. Hashtagging thus does not only represent a labelling practice, but a practice of community care and safe space creation.

Self-motivated posts in our sample outnumbered those that were socially oriented; however, socially motivated posts were more varied in type. This includes advice and inquiry-and-reply posts that are meant to help others with various NSSI-related practices. They included the purchase, storage, use and maintenance of tools for self-injury (e.g. where to hide blades or how to clean them), self-injury techniques and safety procedures (where or how deep to cut), wound care (how to clean and bandage wounds), the acquisition of what one could call a repertoire of detection avoidance (how to hide injuries or how to come up with believable excuses if detected), and alternate coping mechanisms or recovery strategies (putting band aids on uninjured skin, snapping rubber bands against one’s wrist, not seeing relapses as a recovery failure).

We also identified a few informative posts meant to raise awareness about NSSI. Posts in this category educate people who do not self-injure about what NSSI means or does not mean (“we do not just seek attention”, “it is addictive, like alcohol”), what it does for the people who engage in it (“it makes us feel better in the moment”, “this is how I cope”), or identities of people who self-injure (“people of every gender self harm”, “Black people self harm too, it’s not just ‘that white girl shit’”). Utterances like these give us insights into attitudes or beliefs that people who self-injure have about themselves versus those that they are faced with outside the community, highlighting the need for non-judgmental spaces in which people who engage in NSSI can express themselves without fear of social repercussions.

Another type of socially motivated posts was geared toward community-building and networking on Tumblr. Some Tumbleloggers asked people to interact with their post so that they could identify others interested in NSSI and follow them. Other posters acknowledged the high fluctuation of blogs due to detection and deletion: People returning to Tumblr with a new account after being previously removed would disclose their former account name and ask others to amplify their post so that they could rebuild their network of ‘mutuals’ (people who follow each other). Engaging with an NSSI post thus means presenting one’s blog to other Tumbleloggers in the NSSI community.

Two other types of socially motivated posts were related to recovery from NSSI. General messages of support (“you can do it”, “stay strong”) were one type and sometimes included educational content, for instance information on what trauma responses look like, or how recovery and alternate coping mechanisms work. The other type of socially oriented recovery posts is what could be called ‘game-ified’ posts, in which Tumbleloggers who want to recover ask their followers to engage with the post (giving it a ‘note’) and aspire to not self-injure for as many days as the number of notes that the post gets. To maximize the note count, some ask other Tumbleloggers with a larger following to post the ‘game’ for them.

Lastly, there was a category of posts that we found hard to assign to one end on the self/social motivation spectrum. These were posts that featured creative expression, for instance in the form of poetry, creative writing, artwork (drawn, painted or photoshopped images) or memes related to NSSI. Artwork especially was sometimes categorized as “traumacore” (an aestheticization of trauma or traumatic events). Such creative posts drew on inner thoughts and feelings, but also typically generated the most notes, perhaps because Tumbleloggers found others’ artistic expression of their struggles relatable, aesthetically pleasing, or humorous (in the case of memes). As such, these posts represented self-oriented coping strategies to the same extent that they furthered social engagement and shared meaning-making.

4.4. Community and belonging: boundaries of inclusion and exclusion

“How deep do I have to cut so that it counts?” asked one Tumblelogger in a post alluding to an idea that there might be a measurable criterion for validity in the act of self-injury. Someone replied that anything skin-breaking, even a scratch, would count. This exchange shows that people who self-injure may turn to NSSI spaces on Tumblr to experience validation and a sense of belonging. We saw multiple instances of Tumbleloggers who self-injure seeking or making connections with others who self-injure by posting about wanting to follow more NSSI blogs, asking for and giving advice or expressing concern and support for other bloggers. One poster asked others to share NSSI pictures with them in exchange for their own, which bears the implication that connections between Tumbleloggers who self-injure may also develop in private, for instance by exchanging direct messages.

There were also contrary posts such as one Tumblelogger expressing their regret that Tumbleloggers who self-injure did not interact as much as “the ED community”. The poster did not add any details other than the prevalence of photos of injuries and “nothing more” in NSSI spaces, but this perspective shows us how NSSI and ED communities might be (imagined to be) different, despite an overlap in content, and that Tumbleloggers negotiate boundaries between them. Our findings regarding trigger warnings and cross-tagging rules (see previous section) confirm that some practices are in place to keep content separated, often for the sake of the collective well-being of the community.

A preoccupation with the well-being of others was also communicated in profile descriptions, where many Tumbleloggers who self-injure stated that those in recovery should not interact with or turn away from their blog so as to stay safe. Here, boundaries are drawn between groups of people who currently self-injure and those who are recovering. While recovery may happen in non-linear ways (and this was acknowledged in some instances), there was a tendency among Tumbleloggers who self-injure to view recovery as a status or achievement rather than a process.

Further community boundaries were communicated in a post in which the blogger claimed to know a way to confound others who did not self-injure by stating just two words: “Styrofoam” and “beans”.

4.5. “Spicy beans”: NSSI-Related argot and slang

“Styrofoam” and “beans” were but two of the ‘confusing’ words we came across on Tumblr. Many of them we only deciphered when we found a post on another social media site listing popular NSSI-related terms. We compiled a list of terms relevant to the content of this paper(Table 1):

Table 1.

A compilation of selected NSSI-specific terms and their meanings.

Term	Meaning
Barcode	Pattern of linear wounds and scars resembling a barcode; can be used as a verb (“to barcode” means “to self-injure”)
Beans	The hypodermis layer of skin consisting mostly of fat and resembling beans
Sewer slide	Sounds like and is used as a codeword for “suicide”
Slicey boï/boi	The tool used to self-injure (e.g. razor blades, scissors)
Spicy	The stinging sensation caused by recent injuries coming into contact with water
Styro(foam)	The white dermis layer of skin resembling Styrofoam, “styro” is also used as a term for wounds
To yeet (past tense: yeeted/yote)	To self-injure
Yeet	Self-injury

Terms like “sewer slide” and words for different skin layers seem to function as codewords, meant to disguise NSSI activity by using covert language or argot, a set of vocabulary that develops among “groups that operate on the periphery of respectable society” (Eble, 2006, p. 413). Other terms such as “yeet” likely arose as slang (popular non-standard expressions often short-lived in use [Eble, 2006]) outside the community of NSSI practice but with a different in-group meaning. While the majority of these terms seem to have originated outside of Tumblr – further exemplifying that CoP may not function the same across all social media – many of these terms found use on Tumblr as well, likely because people who self-injure use multiple social media. One Tumblelogger complained that the new vocabulary sounded embarrassing; other Tumbleloggers embraced this language, talking about doing “styros” or announcing their plans to “yeet” later. Some terms also form (part of) new hashtags which, while not nearly as populated as more explicitly NSSI-related hashtags, do represent strands of the community that may not be as monitored due to covert language use.

They also represent practices of shared meaning-making, which was evident in meme posts we collected. While some memes were obvious even to people who do not self-injure, others required more intricate knowledge of the language used in the community. One meme showed a picture of a product barcode and simply indicated that the poster’s self-injuries look similar. Another meme compared the act of showering to a picture of a can of spicy baked beans. Without the context of “spicy” meaning the sting of an injury under water and “beans” referring to the layer of fat of the hypodermis, the meaning of this meme would be lost. As such, becoming and being a member of a community of NSSI practice means familiarizing oneself with community-specific language. This became clear not only in our Tumblr data but also the vocabulary post outside of Tumblr that we consulted to learn the terminology: The poster admitted to initially finding it difficult to understand NSSI posts and wanted to share the vocabulary to assist others. This was an important viewpoint for us as we explored Tumblr with a similar approach: as someone new to the community and navigating unfamiliar norms and practices.

5. Discussion

Our paper positions practices and interactions at the center of communal formations between Tumbleloggers who self-injure. Membership of this imagined community of practice is informal and in flux. There are no demographic inclusion or exclusion criteria, though our findings show a tendency toward young age and broad representation across the gender and sexuality spectrums, based on self-disclosures of age, pronoun/gender, and romantic/sexual orientation on Tumblelogs. Racial identity was less often disclosed but we saw some discussions of NSSI stigma in Black communities in some posts and replies. Other studies contextualize these findings: Teenagers and young adults have long been documented to be the most at-risk population for NSSI (Klonsky et al., 2014; Lewis and Heath, 2015; Seko and Lewis, 2018), other researchers have reported elevated risks among sexual and gender minority youths (Rogers and Taliaferro, 2020), and NSSI among young adults of color is underresearched despite reports of high risk factors (Hinton, 2019; Rojas-Velasquez et al., 2021). Disclosure of demographic information is voluntary on Tumblr and in no way did our data show that imagined community membership is limited to or negotiated in terms of identity.

Rather, the only instance in which we observed discourse about membership validity was the one in which a Tumblelogger asked how deep they had to cut for it to count, and another quickly reassuring them that any self-injury counted. On the other end of the spectrum, we observed similar questions about the validity of recovery. While scholars regard recovery as a non-linear, open-ended process in which cessation of the practice is not the primary defining factor of the recovery journey (Lewis and Hasking, 2020, 2021), many Tumbleloggers considered their relapses a personal failure to stop self-injuring and ‘exit’ the NSSI community. It seems then that the factor most pertinent to imagined community membership is a practice: the act of self-injuring itself. Though NSSI is primarily considered a personal coping mechanism used “to regulate unwanted affective experiences” (Lewis and Heath, 2015, p. 527), studies have shown that it also has a communicative function: The “paradox of the social in self-injury” (Steggals et al., 2020, p. 158) lies in its potential to render visible and communicate one’s pain to others (Alvarez, 2021; Steggals et al., 2020). As such, we interpreted photographs, descriptions, or other depictions of self-injury as both a form of self-oriented coping and socially motivated communication.

Despite its centrality, self-injury is not the only practice that members of imagined communities of NSSI practice engage in; there are many adjacent practices that contribute to community formation. We found these practices to be centered around practical advice, collective coping mechanisms, social support, and language use. We interpreted these as forms of collective sense-making and communal care through consuming relatable content. The engagement and support one encounters can evolve into a “sense of community” and “sense of belonging” (Seko et al., 2015, p. 1339), which in turn may relieve feelings of being alone or misunderstood (Alvarez, 2021; Lavis and Winter, 2020; Seko et al., 2015). Accordingly, in its Community Guidelines, Tumblr recognizes that a “[d]ialogue about [self-injurious] behaviors is incredibly important and online communities can be extraordinarily helpful to people struggling with these difficult conditions” (Tumblr, 2020), which might explain the platform’s lack of hashtag restrictions in comparison with other social media (Seko and Lewis, 2018). Indeed, with few exceptions, the public interactions we observed between Tumbleloggers who self-injure showed commitment to emotional support, validation, and safe space creation. On the other hand, we noticed a troubling tendency in Tumblr’s user-specific recommendations to be too helpful in suggesting more and more blogs, hashtags, and content to us based on the keywords and tags we searched and followed, and less consistent over time in displaying support messages with crisis hotlines when searching sensitive content.

To be sure, the ever-popular belief that exposure to NSSI content could cause more social media users to self-injure has been discredited by many researchers, who also warn of the social contagion claim’s stigmatizing effects (Staniland et al., 2021). In fact, our data shows that many members of the Tumblr community of NSSI practice concern themselves with reducing potential harm done to others by labelling their blogs and posts with trigger warnings, and correcting others when they cross-tag sensitive content. Lavis and Winter note that “self-surveillance and self-censorship [are] undertaken to protect others within the community” (2020, p. 848) and describe how the definition of triggering content is expanded to include not just NSSI imagery but also discussions of NSSI, trauma and distress.

We did observe a greater overlap between NSSI and ED or mental illness content than instances of Tumbleloggers policing it. While this may have implications for the effectiveness of user-based harm reduction efforts, it certainly also gives us valuable insights about how Tumbleloggers who self-injure perceive their “orientation to other communities of practice” (Eckert, 2006, p. 683). Studies have shown that ED and mental illnesses do often co-occur with NSSI (Lewis and Hasking, 2021; Zinoviev et al., 2016), and the information we gathered from our Tumblr account samples reconfirm this. Even more prevalent were mentions of their “orientation […] to the world around them more generally” (Eckert, 2006, p. 683): Because NSSI is often practiced as a way to make sense of or navigate one’s personal relationships (Alvarez, 2021), but also commonly misunderstood and stigmatized by those who do not self-injure (Hilton, 2017; Lewis and Hasking, 2021), many people who engage in NSSI turn to online communities to express their feelings to others. Much of our collected textual data revolved around offline relationships with friends, romantic partners, and family. Personal stories or so-called ‘vent posts’ thus can give researchers valuable insights into what self-injury means to those who engage in it, and how they position themselves in relation to people online and offline.

These text posts bore a wealth of linguistic data. Our findings show the use of in-group terminology that would likely not mean much to people outside the community of NSSI practice. Some memes and jokes were specifically based on the assumption that only those who self-injured would understand, which may create an in-group feeling of belonging, as argot and slang typically “bring about a social bond among their users through awareness of shared knowledge” (Eble, 2006, p. 412). This shared knowledge can also help Tumbleloggers identify blogs with NSSI content, or make their own NSSI content recognizable, using topical usernames or disclosing pertinent keywords in their blog descriptions.

While practice-specific vocabulary emerges in many communities, we also need to consider the added contexts of content moderation and fear of detection in NSSI communities. Many of these terms functioned as codewords, for instance in hashtags that were “less indicative of SI to avoid censorship” (Seko and Lewis, 2018, p. 195). Of course, the language we observed in our data was recorded at a specific point in time and will be subject to continuous change. We entered our data collection with the expectation of encountering certain covert terms and hashtags covered in literature on NSSI, such as #secretsociety123 and #ehtilb (spelling “blithe”, a codeword for self-injurious behavior, backwards) (Moreno et al., 2016), but realized quickly that these terms were no longer in frequent use, as Tumblr posts under these hashtags were rare in 2021. A non-Tumblr post explaining the vocabulary to others, as well as someone’s post on Tumblr complaining about new terms sounding awkward, contextualized and confirmed our documented terminology’s recent addition to the community lexicon. Learning the vocabulary of imagined communities of NSSI practice helps identifying NSSI activity and working out how community members engage in meaning-making by using specific in-group language relating to NSSI practices. It is our hope that this will lead to easier identification of (online) NSSI activity and an improved understanding of the importance of community among people who share self-injury experiences, and therefore more empathic interventions for care and support.

5.1. Limitations

In studying social media posts, we explore utterances or impressions whose context has collapsed. A context collapse is defined as a situation when actors are attempting to understand unrelated social contexts which are dependent on different social norms for their meaning (boyd, 2014, p. 31). Therefore, doing research in a situation of context collapse is challenging because we must recontextualize each post, a process which will impact interpretation of meanings. We can thus not claim accuracy for our interpretations and acknowledge that we cannot do justice to the complexity of the community or its members’ shared experiences. Rather, in documenting our observations of our first time navigating NSSI content on Tumblr, we used an (auto)ethnographic approach to put ourselves in the shoes of someone who is new to NSSI Tumblr and learns how to become and act as part of this imagined community of practice.

While we certainly emulated a user experience, we cannot guarantee that it is the typical user experience, as we were limited by a couple of factors. Out of ethical considerations, we did not interfere by liking posts or following blogger profiles. We cannot clearly say what Tumblr-generated recommendations for posts or blogs would look like for someone who does actively engage with NSSI content. An active user’s dashboard would mainly show content produced by blogs they follow, with some recommended posts interspersed. We did not block any hashtags or profiles in our settings, which Tumbleloggers can choose to do if they want to filter out content to which they may be sensitive. Finally, our data consists of publicly available information and did not grant us any insights to communal interactions between Tumbleloggers who self-injure that may happen in Tumblr’s direct messaging function or other forms of private chats. Further research and inclusion of interviews with Tumbleloggers may be necessary to get a more holistic understanding of the practices that form the pillars of Tumblr’s imagined community of NSSI practice.

6. Conclusion

In this paper, we approached NSSI content on Tumblr by centering the practice and consequently framing the social bonds between Tumbleloggers who self-injure as an imagined community of NSSI practice. Methodologically, we implemented this approach by using online ethnography, not only sampling posts but also logging our observations and thoughts. This gave us the opportunity to explore NSSI content on Tumblr through the eyes of someone who is new to the community and is learning the unspoken norms and meanings of its practices and interactions. Our findings showed that a variety of social and linguistic practices underpin the engagement between Tumbleloggers who self-injure, which may result in senses of community and belonging they may not find as easily with people who do not self-injure due to stigmatizing views. Conversely, distinctions between communities of NSSI practice and other CoP are enacted through social and linguistic factors as well, for instance in form of norms for hashtag and trigger warning use but also in terms of in-group language that may not be accessible to non-members of the community. Tumblr enforces content moderation but encourages supportive interactions among people who self-injure. The platform’s overall ambiguity results in NSSI community practices that are rich in both covertness and transparency. Future research would benefit from ethnographic approaches that observe community formations and underlying practices from an in-group perspective to understand the processes of shared meaning-making in communities of NSSI practice and their importance to people who self-injure.