The self—harmed,visualized,and reblogged: Remaking of self-injury narratives on Tumblr

Introduction

Over the past few decades, self-injury (SI), the non-suicidal and intentional damaging of one’s body tissue by cutting and burning, among other methods, has emerged worldwide as a cardinal mental health concern. Studies report that people self-injure to cope with negative emotions (Klonsky, 2011), punish oneself (Rodham et al., 2004), or solicit help from others (Nock and Prinstein, 2005). Some clinical experts have argued for conceptualizing SI independently from other mental illnesses and proposed non-suicidal self-injury (NSSI) disorder as a distinct diagnostic category (American Psychiatric Association, 2013). ¹

While SI affects all age groups, estimated rates are consistently higher among young people than among adults, with 13–17% of adolescents and young adults reporting at least one SI episode in their lifetime (Swannell et al., 2014). Researchers have also suggested that many youth with SI histories prefer private settings to self-injure and seldom seek professional help due to the fear of stigma and shame associated with the act (Rowe et al., 2014). It is thus not surprising that representations of SI have been flourishing in cyberspace, as the Internet’s relative anonymity has long afforded marginalized communities a refuge to express themselves with little fear of stigmatization (e.g. Boyd and Marwick, 2009). Indeed, one research suggests that youth who self-injure may also use online media to a greater degree than those who do not (Mitchell and Ybarra, 2007).

To date, several studies have explored SI-related online activities, focusing primarily on text-based interactions on discussion forums (Franzen and Gottzen, 2011; Rodham et al., 2007) and personal websites (Lewis and Baker, 2011). Yet, the exploration into visual and multimedia aspects of online SI activities still lags behind language-based inquiry. The few studies examining visual SI content on YouTube (Lewis et al., 2011; Lewis and Knoll, 2015) have been concerned mainly with the potential risks and benefits of graphic imagery from a psychomedical point of view while leaving aside how such imagery may enable people who self-injure to share their experience over the Internet.

In this article we explore images themed SI posted on Tumblr.com, an image-rich social media site that enjoys high popularity among users 13–24 years old (Pater et al., 2016), the age group at high risk of SI (Swannell et al., 2014). Since its inception in 2007, Tumblr has served as a hotbed of blogs featuring SI, eating disorders, suicide, and other self-harming behaviors. The public concern about the potentially deleterious effect these blogs may have resulted in the site’s 2012 implementation of a policy against “pro-self-harm blogs” that “actively promote or glorify eating disorders, cutting and suicide” while encouraging the blogs dedicated to “recovery and support.” ² Tumblr’s policy against these “pro-self-harm” blogs has been incorporated into its community guidelines, which declared non-compliant blogs to be removed from the site. ³ Along with this policy implementation, Tumblr began showing an automatic alert when users searched for SI-related keywords on the site (e.g. SI, self-harm). The message contains a crisis line telephone number and links to self-harm treatment and crisis intervention services. ⁴ Despite these efforts, SI-related blogs are still abundant on Tumblr. Unlike Instragram that has been actively censoring self-harm content by removing relevant hashtags from its search results, ⁵ Tumblr takes a blog-by-blog basis approach relying primarily on reports from users and staff members. Ostensibly, this renders the line between “pro-self-harm” and “pro-recovery” blogs less clear-cut.

Our analysis involves a close examination of visual, textual, and compositional attributes of photo-based images on Tumblr, exploring how these multimedia assemblages represent SI on Tumblr’s network. We argue that unique affordances of the Tumblr platform shape distinct ways in which SI is told and shared. These affordances also imply the broader sociocultural context drawn by the particular audiences in making sense of the shared story. In what follows, we first map out some of the critical attributes of the Tumblr platform that constitute a unique narrative milieu and social dynamic based on re-appropriation and collective curation. We then report findings of our visual narrative analysis that explored content and visual structures of the images, the associated texts, and the pattern of content dissemination. We discuss how SI is visually shaped on Tumblr and how people engage in meaning-making around SI by using heterogeneous cultural artifacts available to them. To the best of our knowledge, this is the first study investigating multiple narrative elements of SI-themed images on Tumblr, along with the relations between image characteristics and the diffusion of those images across the network. While our primary focus was on Tumblr, it is our hope that our multimodal approach will be applicable to other social media platforms.

Tumblr: fragmented narrative space for self-curation

The emergence of social media that embrace “social behaviors that traverse and frequently converge a variety of behavioral and performative platforms” (Papacharissi, 2010: 312) has unsettled the earlier notion of “virtual communities” as singular, static social groups with strong psychological bond (Parks, 2010). Some media scholars turned to the notion of affordances to explore unique ecology and the emergent sociability of social media. Parks (2010), for example, examined social affordances of MySpace that “facilitate or ‘call forth’ the constitutive elements of community” (p. 109). Papacharissi (2010) further called attention to particular social, technical, and architectural affordances of a given platform that generate specific types of communication and unique patterns of sociability while excluding others.

Tumblr has a series of unique structural and social affordances that promote a particular mode of communication. On Tumblr, users can post texts, pictures, GIFs, and other multimedia content on their short-form blogs, known as Tumblelogs. Although known as a social media site, the “social” features on Tumblr are different from other social media platforms in several key ways. Unlike Facebook, Twitter, and other popular social networking sites, it is not mandatory for Tumblr users to provide basic profile information or publicize “friend lists” to showcase their social connections. Many users employ pseudonyms, create more than one Tumblelogs, or collaboratively manage their blogs with others to avoid linking their blogs with their real-life identities (Renninger, 2015). There is also no native public commenting feature for a Tumblelog, unless blog owners add a third-party plug-in to their blogs or activate “reply” function to allow post replies from people they follow. Such a relatively “asocial” affordance of Tumblr has made it particularly conducive to performing non-normative subjectivities. Some marginalized groups for whom Tumblr serves a prominent venue include NSFW (Not Safe/Suitable For Work) content uploaders (Tiidenberg and Gómez Cruz, 2015), LGBTQ (lesbian, gay, bisexual, trans, and queer) community (Cho, 2015; Fink and Miller, 2013), asexual community (Renninger, 2015), and alienated youth (Hart, 2015) who prefer to keep some aspects of their identities invisible from friends, families, co-workers, or employers.

Instead of user-centered interactions, sociality on Tumblr is, for the most part, expressed by content-based interaction, especially through “reblogging” and “liking.” While giving another post a “like” is a gesture of affirmation prevalent among social media platforms, the “reblog” feature contributes largely to Tumblr’s unique media ecosystem. By reblogging a post, bloggers can copy and repost the material made by others onto their own dashboards (i.e. homepage). The record of interactions with a post is immediately attached to the post through “notes” that list the original poster and each user who has reblogged or liked the post. Reblogged content, original uploads, and entries posted on the blogs a user follows are all congregated into the user’s dashboard, displayed together in juxtaposition. This function has facilitated unique dynamics of re-appropriation and re-sharing, to the extent that over 93% of the Tumblr posts are reblogs rather than the original posts (Xu et al., 2014). In this ecosystem, where original and re-appropriated materials harmoniously coexist, the authenticity and ownership matter less than the speed and scale of sharing; what Tumblr embraces instead is, as Cho (2015) posits, the continuous act of curation—the social practice of finding, sorting, and assembling digital content for the purpose of public display.

By offering users personal gallery for social curation, Tumblr embraces what Hogan (2010) calls the “exhibition” of the self through a bricolage of digital artifacts. While for Hogan the “curators” of identities are non-human agents such as algorithms that manage digital content beyond user’s control, Tumbleloggers seem to play an active role in their self-curation in collaboration with the algorithm. Through an eclectic mix of images and texts users haphazardly represent their interests, memories, and feelings, while metadata generated by Tumblr algorithm (e.g. the record of reblogging and liking) weave dispersed narrative fragments into shared story clusters. What seems intriguing here is the technical affordance that fosters instant feedback loop between the Tumbleloggers and the network: as soon as a user assigns tags on his or her blog post, or likes/reblogs other’s post, the system immediately links the user to existing story clusters and people with similar interests. With an aid of the algorithm that constructs “relational juxtapositions” (Hayles, 2007: 1603), users are able to recognize how their self-presentations are connected to others (perhaps in serendipitous ways) and express their interests by using specific tags or reblogging particular posts accordingly. This user-algorithm symbiosis illustrates what Hayles (2007) coins “the dance” of narratives and databases, in which fragmentary narratives assembled by databases spawn new narratives of shared experience beyond conventional causal and conceptual structures.

Consequently, the practice of self-curation on Tumblr has added a new dimension to the way personal lived experience of pain and suffering is narrated and shared. Although SI is, like other pain experiences, originally confined to one’s body, Tumblr’s affordances that patronize digital curation and depersonalized, fragmented self-expression generate new narrativity of how to comprehend and represent SI beyond immediate physical contexts. In their analysis of chronic pain narratives on Tumblr, Gonzalez-Polledo and Tarr (2016) contend that pain experiences no longer belong to one person on Tumblr; instead, by virtue of traveling around the site through reblogging, personal pain experiences are transformed into “symptomatic communication” that “no longer aims to represent or show reality” bound to individual bodies (p. 7). The popularity of “memes” on Tumblr—namely, images, videos, and spoofs under constant circulation and modification by users (Vickery, 2014)—is particularly suggestive in this regard, as this emergent narrative format enables users to re-appropriate cultural artifacts around them to express their unique stories—often in a sarcastic and provocative manner. ⁶ Tumblr thus captures and capitalizes on what Jenkins (2006) dubbed the tendency of media convergence wherein not only media content but also “lives, relationships, memories, fantasies, desires” flow across media channels (p. 17).

Method

Results

Among the 294 images sampled, still photographs accounted for 255 (86.7% of the total images), of which 200 were single-frame photographs and 55 were “photosets,” multiple photographs arranged in one frame. The rest of 39 images were GIF animations (13.3% of the total images), most of which were single GIFs with only one image comprising two GIFs in one frame. We observed several differences between images directly depicting self-inflicted bodies (Direct SI images) and those without such content (Indirect SI images). In what follows, we report the findings focusing on the differences between these two image types.

Visual grammar

Table 1 presents the visual grammar of Direct SI images versus Indirect SI images. A chi-square analysis indicated several significant differences regarding how these two types of images were visually structured. The vast majority of Direct SI images (n = 95, 77.2%) were close-up shots of body parts, followed by zoomed-in shots of injured skin (n = 26, 21.1%), whereas Indirect SI images consisted of close-up shots (n = 125, 73.1%) and images taken from medium distance (n = 38, 22.2%) with no zoomed-in shots, χ²(3, N = 294) = 64.37, p < .001, phi = .468.

OPEN IN VIEWER

Table 1.

Visual grammar of the images sampled (N = 294).

	Direct SI images (n = 123)	Indirect SI images (n = 171)	Total (N = 294)	χ2	Phi
Distance
Close-up shot	95	125	220	64.37*	.468
Zoomed-in shot	26	0	26
Medium shot	2	38	40
Long shot	0	8	8
Point of view
First person	109	56	165	93.93*	.565
Second person	5	83	88
Selfie/mirror	9	31	40
Other (iPad screenshot)	0	1	1
Color
Color	90	78	170	26.23*	.299
Black and white	30	92	122
Monotone (sepia)	3	1	4

SI: self-injury.

*

p < .001.

Most Direct SI images (n = 109, 88.6%) were taken from a first-person perspective, capturing the scene from a body owner’s point-of-view, in contrast to Indirect SI images that depicted more second-person (n = 83, 48.5%) than first-person perspectives (n = 56, 32.7%). This indicates that Direct SI images were twice more likely to be taken from the subjective angle than Indirect SI images, replicating how the self-injured bodies would appear to the eyes of body owners, χ²(3, N = 294) = 93.93, p < .001, phi = .565. Nine Direct SI images and 31 Indirect SI images were identified as “selfies,” which reflects the emergent genre of amateur self-portraits taken in a mirror or from arm’s length with a camera-equipped phone (Tiidenberg and Cruz, 2015).

With regard to color schemes, nearly three-quarters (73.1%, n = 90) of the Direct SI images were color, whereas over a half of the Indirect SI images (53.8%, n = 92) were black and white, χ²(2, N = 294) = 26.23, p < .001, phi = .299. Many Indirect SI images, especially snippets from popular media content, appeared to be converted to black and white, whereas the vast majority of Direct SI images appeared to represent ordinary color snapshots involving little visual manipulations.

Narrative themes

A total of nine narrative themes were identified across the 294 images: hopeful/pro-recovery (n = 86), hopeless (n = 86), denotative (n = 42), SI normalization (n = 20), sarcasm (n = 18), eating disorder (n = 16), suicide (n = 10), help-seeking (n = 8), and depression (n = 8). Rather than running through all those themes, in what follows we provide detailed descriptions of the four most prominent themes with representative examples (see Figure 1 for the entire themes).

OPEN IN VIEWER

Figure 1.

Narrative themes identified across the sampled images (N = 294).

Hopeful/pro-recovery

In all, 86 images (38 Direct SI images and 48 Indirect SI images) were coded to convey hopeful, recovery-oriented narratives, delivering supportive message to viewers who self-injure through visual and textual means. These images were often accompanied with personal stories of recovery and resilience or encouraging messages to others with a SI history, such as “With time, patience, tears and love, all your wounds will heal. Physical and Emotional. I promise.”

Thirty-two of these hopeful/pro-recovery images depicted scars or healing wounds, rather than new, bleeding injuries. Among them, eight images showed a tattoo inscribed over people’s scars, with the uploaders sharing their story of recovery from SI or stating their support for pro-recovery movements such as the Project Semicolon. ⁹ One representative example was the image depicting a semicolon tattoo with butterfly wings over a number of scars on the arm. The associated caption read,

Absolutely adore my first tattoo. I chose a semicolon with a butterfly because “an author uses a semi colon when he could have ended a sentence but chose not to, the author is you and the sentence is your life.”

By referring to the Semicolon Project’s tagline, the blogger expressed support for this initiative, while at the same time the scars underneath visually narrated a history of personal struggle and the pain endured during this time. This blending of personal and communal narratives was also observed in selfies that conveyed hopeful messages, wherein individuals were smiling at the camera while showing visible scars on their bodies. Some of these images were photosets that juxtaposed before-and-after selfies, in which the “before” pictures depicted individuals with facial expressions symbolizing struggle, distress, and discomfort, while in the “after” pictures they appeared happy and self-confident.

Hopeless

Totally 86 images (21 Direct SI images and 65 Indirect SI images) communicated a sad and hopeless mood. The majority of Direct SI images coded into this category were close-up shots of fresh, bleeding injuries on arms, thighs, or other identifiable body parts, with an accompanied caption or superimposed message emphasizing negative feelings associated with SI (e.g. shame, guilt, and loneliness) such as “Can’t say a word … no one cares anyway so whatever.” or “Darker, deeper six feet underground. I’m screaming but no one hears a sound.” Compared to the hopeful narratives described above, the Direct SI images communicating hopeless tone were less likely to document people’s face.

A hopeless mood was much more prevalent among Indirect SI images than Direct SI images. Most often, this was represented as black-and-white character photographs or GIFs. In these character images, people seldom gazed directly at the camera; rather, they turned their heads away or hid faces with hands, crying, trembling, anxiously smiling, or performing similar gestures that signaled a negative emotional state. A representative example was a black-and-white GIF taken from the TV series My Mad Fat Diary that showed a side profile of a young woman lying in bed, looking up at the ceiling with a vacant, numb face, letting tears running down her face; the text overlay read, “I sit here crying tears no one will ever see.” The character images as such did not directly depict SI, but instead could be read to figuratively communicate feelings behind SI. Coupled with personal taglines indicative of sad, pent-up emotions, these images may represent a sense of loneliness, hopelessness, and vulnerability associated with SI.

Denotative

In all, 42 images (31 Direct SI images and 11 Indirect SI images) were coded as denotative, as they represented subjects (e.g. self-inflicted wounds, objects, scenery) in a straightforward manner. The majority of these images were close-up shots of body parts (e.g. arms, thighs) and zoomed-in shots of injured skin with little or no textual explanations. The hashtags associated with these images tended to literally describe the content (e.g. #cuts, #blood), as though the purpose of uploading these images was to simply depict the subject matter. This lack of context was most conspicuous among zoomed-in shots of bleeding wounds, in which body parts were not identifiable.

SI normalization

In 20 images (15 Direct SI images and 5 Indirect SI images) showing self-inflicted wounds (mostly open, bleeding wounds), razor blades, or other materials that symbolize SI, the act of SI was presented as an effective coping mechanism. Many of these images accompanied provocative text justifying or normalizing their actions. For instance, the caption to a photograph of bandaged arm with blood seeping through reads, “you think that when i take a blade to my skin—you’d think i’d be hyperventilating crying sobbing uncontrollably. but really, when i feel my skin split in two is when i feel the most calm.” Arguing against “you”—assumingly a person who does not self-injure—the blogger seemed to frame SI as a deliberate act under his or her control. The bloody bandage in the image also implied that the person took self-care of the outcome of SI, perhaps without seeking help from others.

Pattern of reblogging

Following the content and thematic analysis, we examined the patterns of content diffusion by looking at the number of reblogs attached to the sampled images. Overall, Indirect SI images as a whole received almost 10 times more reblogs than SI images; 68 out of 123 Direct SI images received a total of 1024 reblogs ranging from 1 to 338 per image (mean [M] = 15.52; standard deviation [SD] = 49.39), whereas 130 out of 171 Indirect SI images received a total of 10,238 reblogs ranging from 1 to 889 per image (M = 79.98; SD = 139.25). An independent-samples t-test was performed and indicated a significant difference in the number of reblogs that was observed between the two types of images, t(192) = 3.64, p = <.0001. Cohen’s d was .62 indicating a medium effect size.

To further ascertain the pattern of reblogging, we looked into which narrative theme was more frequently reblogged than the other. As presented in Table 2, there was a notable difference in the number of reblogs between different narrative themes. The most reblogged among Direct SI images were denotative images (n = 13, 590 reblogs), followed by SI normalization (n = 12, 137 reblogs), and hopeful/pro-recovery images (n = 21, 132 reblogs). Among Indirect SI images, the hopeless theme was most commonly reblogged (n = 56, 5214 reblogs), followed by images pertaining to suicide (n = 9, 1494 reblogs), and then those with a hopeful/pro-recovery theme (n = 29, 1353 reblogs). Of the 19 images that received more than 200 reblogs, 15 were identified as film and television memes featuring characters crying, anxiously smiling, or yelling at others. In all, 11 out of 19 communicated a hopeless theme, followed by 2 denotative, 2 sarcastic, and 2 suicide themes. Hopeful/pro-recovery message was conveyed by only one image, a black-and-white screenshot from a popular TV show Grey’s Anatomy.

OPEN IN VIEWER

Table 2.

Number of reblogs attached to the images sampled (N = 198) and corresponding narrative themes.

Narrative theme	Direct SI images (68 had at least one reblog)				Indirect SI images (130 had at least one reblog)
	Number of images	Number of reblogs	Mean (SD)	Range	Number of images	Number of reblogs	Mean (SD)	Range
Hopeless	11	96	8.73 (12.17)	38	56	5214	94.80 (121.35)	367
Suicide	0	0	0 (-)	0	9	1494	166.00 (272.01)	671
Hopeful/pro-recovery	21	132	6.29 (9.57)	33	29	1353	46.66 (165.75)	888
Sarcastic	7	47	7.83 (9.06)	0	6	744	124.00 (137.13)	340
Denotative	13	590	45.38 (106.96)	337	7	150	21.43 (19.66)	54
Eating disorder	0	0	0 (–)	0	11	529	52.9 (42.28)	109
Depression	0	0	0 (–)	0	6	462	77.00 (63.34)	156
SI normalization	12	137	12.45 (16.36)	49	5	157	31.4 (35.32)	91
Help-seeking	4	22	5.50 (7.05)	15	1	135	135 (–)	–
Total	68	1024	15.52 (49.39)	337	130	10238	79.98 (139.25)	888

SI: self-injury; SD: standard deviation.

Discussion

Our visual narrative analysis revealed how different narrative elements—content and composition of images, associated texts and hashtags, and the patterns of reblogging—together generate particular ways in which SI is narrated and shared on Tumblr. To our surprise, over half of the images studied (58.2%) did not directly depict self-inflicted wounds, despite the accompanied SI-related tags and captions. A number of differences were identified between the Direct SI images featuring wounds and Indirect SI images without injury depictions. The content of Direct SI images was largely congruent with previous studies on visual representations of SI online (e.g. Lewis et al., 2011; Seko, 2013), in which cutting on the arms or legs was the most prevalent practice; this also aligns with clinical research regarding the most common SI methods used by those who self-injure (Klonsky, 2011). The Direct SI images predominantly communicated a first-person vantage point, in contrast to the Indirect SI images that tended to show a second-person, objective perspective. Taken from a first-person perspective, these Direct SI images appeared to invite viewers to see what the body owners see, to vicariously feel the pain as their own, while constantly reminding that the wounds were intentionally inflicted on the body. Many of the Direct SI images appeared to be made and uploaded by the owners of the wounded bodies—although they might not be a user’s making but taken from elsewhere on the Internet. As implied in a previous study (Seko, 2013), these graphic self-portraits may constitute a form of visual monologue that makes pain recognizable and sharable for asynchronous viewers.

While carrying a personal testimony of suffering, a significant number of Direct SI images communicated hopeful, pro-recovery messages. In many of these images, scars represented a key vehicle for storytelling, helping the uploaders to visually declare their authenticity and narrate personal pain endured. In line with recent research (Lewis and Mehrabkhani, 2016), SI scars depicted in these images appeared to symbolize resilience and inner strength, enabling some individuals to reintegrate the painful past into their self-narratives. This re-reading of scars as a proof of resilience can be seen as an adaptive process, whereby these persons work to re-construct coherent, meaningful narratives and re-establish their sense of self-worth to fit their changing circumstances. At the same time, personal wounds can also be used to claim membership to SI communities. For example, findings from an interview study with youth who self-injure indicated that images of self-inflicted wounds may help people with SI experiences to identify with similar others and feel a sense of belonging (Seko et al., 2015). In this study, photographs of semicolon tattoos and drawings over scars visually express belonging to a particular community and a sense of solidarity to similar others. Through the use of shared symbols (i.e. semicolon mark), personal SI scars may be resituated in the context of the pro-recovery movement, in which shared subcultural understanding and conventions make individual experience communal and relatable.

Contrary to Direct SI images, the majority of Indirect SI images conveyed hopeless narratives, particularly in the form of black-and-white character photographs or GIFs originated from popular media content. Many of these memes were based on movies and TV shows in which characters displayed gestures and facial expressions that symbolize feelings of despair, loneliness, and alienation. By re-appropriating popular media texts to express profoundly different messages (i.e. former/current experiences with SI or feelings prone to SI), these hopeless memes seem to represent SI as a form of life struggle virtually anyone can face, while complicating conventional readings of SI as an individual, pathological experience. Through the processes of modification, transmission, and re-circulation, the topic of SI seemingly manifests on Tumblr via a multitude of popular and subcultural materials, taking on a renewed, collective meaning that spans the multiplicity of personal experiences. In this sense, Tumblr appears to embrace what Jenkins (2006) has coined “collective intelligence,” wherein people harness their individual expertise and mobilize media artifacts as tokens of shared identity, tastes, and meaning—in case of SI narratives, what SI means to them beyond psychomedical understanding.

Given the shift in the iconography of SI—from the depiction of self-injured bodies to the figurative representation of mental suffering—our reblogging analysis is suggestive as to which SI images are more widely spread among Tumblr users. Results revealed that Direct SI images were reblogged less often than Indirect SI images. Within Direct SI images, the most frequently reblogged were denotative ones accompanied with little or no textual information. Within Indirect SI images, the most frequently reblogged were popular media memes conveying hopeless or suicidal messages, whereas hopeful, recovery-oriented images accompanying personal narratives diffused less widely. This trend toward less personalized, decontextualized content may indicate users’ preference for generic media content through which they could express and affirm their pent-up feelings without sharing too many details of their lives. Hopeless images could also raise empathy among viewers with various life struggles without limiting the scope to SI. Indeed, many hopeless memes found in this study accompanied a variety of hashtags indicative of mental and emotional issues in addition to SI, including depression, suicide, and eating disorders. The affordances of Tumblr may thus generate a particular way of expressing one’s history or interest in SI, which is relatively anonymous and depersonalized, but easily decipherable through a specific logic of “collective intelligence” (Jenkins, 2006).

This disproportionate reblogging also bears important implications for the efforts to bring support to people who self-injure. First, repetitive digital circulation of popular media memes as “the narrative of SI” may reproduce stereotypical representation across popular media, in which SI is portrayed predominantly as superficial cutting performed by young White females (Whitlock et al., 2009). Although such memes may enable people who identify with such portrayals to voice their feelings anonymously through a shared template, the gendered and racially skewed representation identified in this study may reproduce the discourse on young (White) women as vulnerable victims while excluding the experience of people who do not fall into the category of what many may view as a “typical self-injurer.” Johansson and Sternudd (2015) make a similar line of observation in their exploration into SI monologues on YouTube, in which a motif of a “hunched-over sitting girl” is becoming a new “visual trope” that reproduces a much-criticized perception of girlhood as frail, docile, and subordinated. Furthermore, the dominance of hopeless message may be detrimental to vulnerable viewers who strongly identify with such messages. The negative impact of narrative reinforcement through online SI materials on those who self-injure has been pointed out by several researchers (e.g. Lewis et al., 2011; Whitlock et al., 2009), as these materials tend to frame SI as unstoppable addiction beyond one’s control, induce the sense of powerlessness, and thus may contribute to the maintenance of the behavior.

Consequently, SI narratives on Tumblr seem to make an exemplary case of what Boyd and Marwick (2009) call the “conundrum of visibility” pertinent to youth-generated “problematic” online content. On one hand, the Internet’s ability to visualize “problematic” behaviors like SI evokes a reasonable fear against pernicious and contagious effects of digital materials on vulnerable youth. In particular, the dark, hopeless imagery circulated across Tumblr and elsewhere on the Internet may appear “worrisome” (Lewis et al., 2011) to the eyes of clinical and educational professionals working with young clients. Yet, on the other hand, as Boyd and Marwick (2009: 410) posit, such visual narratives bring to the light the otherwise hidden, larger social issues today’s youth are facing such as bullying, substance abuse, suicidal ideation, and other mental health difficulties and help viewers understand “the lives of youth, the risks and dangers they face, and the personal, social and cultural logic behind their practices.”

We understand the images of SI on Tumblr as fragmentary, yet collective narratives of those bloggers with an SI history or interest, exploring, adopting, and rejecting the various discourse and cultural artifacts available to them. With the unique affordances of Tumblr’s platform that facilitates social dynamics of self-curation, representations of SI have become polysemous and conveyed diverse and sometimes conflicting meanings beyond their immediate context. The act of reblogging may further reshape what SI means within Tumblr’s dispersed and ever-changing public, re-contextualizing SI in a broader discourse of suffering with other mental health and social issues. As such, people striving to develop a strategy to help youth facing social and mental health challenges may benefit from learning about the emergent iconography of SI and other mental health issues on social media.

While this study focused solely on photo-based visual content, there are a variety of non-photographic images such as drawings, cartoons, and animations that would give further insights into how SI experiences are told and understood by social media users. A deeper engagement with these vernacular narratives will allow a glimpse into how young people employ changing value systems and make sense of what these issues mean to them beyond conventional psychomedical discourse. In doing so, it deserves critical attention as to how such reframing enables particular SI experiences to be shared while silencing others online and offline. If, as this study implied, popular media memes circulated online reproduce a stereotypical image of “self-injurer,” it will be crucial for future research to trace the portrayal of “atypical self-injurers” across popular media and the Internet and critically assess how different discourses and modes of representations have been mobilized to frame their “atypical” bodies.

Although it was beyond the scope of this study to analyze the role individual users play in content dissemination, it would be informative for future research to explore who emerge as key social hubs for content diffusion (i.e. opinion leaders) through network analysis. Empirical studies including interviews with users, too, would further untangle what constitutes mutually meaningful content within the network and how Tumblr is important to those who self-injure. Moreover, the present research left room for hashtag analysis. It has been reported that hashtags are a strong predictor of content dissemination over social media platforms, as they connect individual content to a greater public discussion (Stefanone et al., 2015). Although this study used denotative query terms based on preliminary observation and a review of previous studies, it is likely that Tumbleloggers use other hashtags that are less indicative of SI to avoid censorship. Future research may thus merit from analyzing co-occurrence and semantic relatedness among hashtags associated with SI-related online activities. In their recent cross-platform analysis, Pater et al. (2016) identified a corpus of hashtags accompanied with eating disorder–related activities across Tumblr, Twitter, and Instagram. Following this step and identifying a corpus of SI-terminology shared between diverse platforms (e.g. Tumblr, Instagram, Twitter, Pinterest) would allow for more holistic understanding of online SI narratives beyond one platform. In doing so, a special attention should be paid to what Pater et al. (2016) call “social engineering” techniques, namely, users’ efforts to circumvent censorship, for instance, by using irrelevant words as hashtags.

Finally, informing health professionals of popular narrative types and the pattern of content dissemination carries valid implications in the development of a client-centered health promotion. Although this study identified the dominance of hopeless narratives, the reference Tumblr users employ to narrate SI is by no means static or unchanging. Given that there are a number of hopeful narratives by SI survivors and those in recovery stage, it may be beneficial for Tumblr to take a proactive stance to promote these pro-recovery narratives and provide recovery-focused online resources beyond its current content regulation policy. The positive affordances of peer-based network need to be maximized in such pro-recovery interventions in order to help youth develop healthy connections that could last beyond online interaction.