Getting Better? Hegemonic, Negotiated and Oppositional Uses of Instagram for Mental Health Support

By analysing 600 Instagram posts that use mental health related hashtags, this article investigates how mental health communication and support practices are enacted on Instagram, and how such practices relate to the perceptible affordances and hegemonic uses of the service. The article demonstrates how Instagram tends to privilege casual snapshots of individual recovery, in line with broader discourses of positive thinking and individual responsibility. Whereas this hegemonic way of using the service may be functional for many users, three examples of negotiated and oppositional use are also discussed in the article: motivational picture quotes, text-rich posts, and non-recovery oriented posts. It is suggested that different ways of imagining and approaching the affordances of the service engender different patterns of support practices.


Social Media and Peer Support in the Context of Mental Health
According to international surveys, common mental disorders-including depressive disorders and anxiety disorders-are a public health concern in large parts of the world (Harris et al., 2020;World Health Organization, 2017). Many affected individuals turn to online environments for support and information. Hitherto, however, social science and humanities research on online mental health communication has largely focused on the interactional processes of text-based support groups, discussion forums and blogs, leaving other types of media unexplored. Questions have concerned, for example, the significance of support forums for identity work, self-advocacy, and diagnosis of self and others (Giles, 2014;Giles & Newbold, 2011, 2013, for stigma management (Yeshua-Katz & Martins, 2013), and for sharing illness narratives (Kotliar, 2016).
Such studies have demonstrated that one significant feature of online support is the display of empathy through reciprocal self-disclosure (Kaufman & Whitehead, 2017), or the telling of "second stories" that echo and elaborate on personal stories previously posted in the forum (Giles & Newbold, 2013). Furthermore, online support has been linked to issues of legitimacy, as the perceived right to seek support is dependent on the user performing illness in a certain way or subscribing to the explanatory framework dominant in a forum (Smithson et al., 2011;Stommel & Koole, 2010;Vayreda & Antaki, 2009). A range of mental health conditions have been addressed in research, but online forums for eating disorders have gained special attention, especially in terms of "negative enabling support groups" or so-called pro-ana communities, which undermine recovery by validating pro-anorexic attitudes (Haas et al., 2010;Chang & Bazarova, 2016). Thereby, they also construct alternative subject positions for girls and women who restrict their food intake (Day & Keys, 2008).
Overall, in addition to focusing on text-based media, previous research on online mental health communication has typically concentrated on relatively confined and enclosed support communities. But less is known about how coping and support are shaped by open platforms, such as Instagram, with their emphasis on publicness, visuality and multimodality.
There are a small number of studies that discuss self-injury photo-sharing in terms of nuanced identity performances through which stigmatizing discourses are negotiated (Seko, 2013), and as a way of giving and receiving support by visually sharing personal experiences. Looking specifically at Instagram, Andalibi et al. (2015Andalibi et al. ( , 2017 claimed that visual imagery may be more important than text to communicate experiences of depression, and that comments on depression-related images tend to provide positive feedback and emotional support. Ging and Garvey suggested that the aestheticisation and formulisation of pro-ana content, enabled by the creative affordances of Instagram, cause a mainstreaming of pro-ana imagery as it shares "instantly recognisable qualities with both mainstream fashion and art photography and Internet meme culture" (Ging & Garvey 2017, p. 13). This is similar to the trend noted by Seko and Lewis (2016) in their study of Tumblr representations of self-injury, where the posting and reposting of popular media memes allow users to share intimate feelings but in a selective way (Seko & Lewis, 2016, p. 14). However, as also noted, this repetitive circulation of generic representations may reproduce gendered and racialized notions of self-harm (Ging & Garvey, 2017;Seko & Lewis, 2016).
These previous studies suggest that the design and features of any social media service channel peer-to-peer interaction in particular ways, and thus are of key importance for how users conceive of their mental health issues. More detailed knowledge is still needed of how specific platforms and affordances may facilitate distinctive forms of mental health communication, coping and support.
In this way, the article expands on and contributes to emerging studies of mental health communication in visual-based digital media, as well as to a broader discussion on the normative effects of media-specific affordances. In a recent study, Mackson et al. (2019) analysed the relationship between using Instagram and mental health more generally. While they concluded that using the platform is associated with psychological well-being, except for in those cases when "Instagram anxiety" or elements of social comparison enter into the equation. This present study, however, is focused on how affordances of Instagram are used by individuals that take to the platform as a means of coping with, or otherwise express themselves about experiences of mental ill health.

Theorising Social Media Affordances
In the context of human-computer-interaction, the concept of affordances (Gibson, 1979;Norman, 1988) has long been in use as a way to describe the relation between technology and sociality, or, more precisely, the action possibilities of an object. According to this line of thought, any specific interactive media technology is suggestive-although not determinative-of particular practices and forms of usage.
To provide a more detailed understanding of how users relate to the "productive constraints" of social media services, and to emphasize how such activities are bound up in certain power dynamics, we concur with Shaw (2017) that the notion of affordances can be fruitfully combined with Stuart Hall's (1993) encoding/decoding communication model. Hall's model-originally developed for understanding reader interpretations of mass media texts-suggests that texts, while being polysemous, are encoded with a privileged or preferred meaning which typically embodies the dominant ideology. Depending on their social context, audiences may interpret a text in ways which are not necessarily in line with such a hegemonic reading. Hall suggests that readers can take up a negotiated position, where they acknowledge the preferred meaning of a text while simultaneously modifying it in line with their own experiences, or an oppositional position, where the privileged meaning is rejected in favor of an alternative interpretative framework.
Adrienne Shaw (2017) has developed this model to account for the ways in which particular forms of usage are encoded into interactive digital media platforms. Building on Gaver's (1991) threefold conceptualization of affordances as either perceptible, hidden or false, Shaw suggested that hegemonic use of a platform involves deploying its perceptible affordances in such a way that "the designers' and users' imagined affordances align" (Shaw, 2017, p. 7). Oppositional use, in contrast, includes using a platform in unexpected ways, by taking advantage of its hidden and false affordances-that is, affordances which are not immediately apparent and not meant to be deployed, or apparent affordances that do not have a real function. Negotiated use, then, falls in between these two categories.
In this article, we draw on Shaw's typology to identify how the taking up of different "using positions" is also bound up with different support practices and ways of conceptualizing mental health issues. Importantly, then, some ways of reaching out are implicitly privileged by the Instagram platform, and some are even explicitly construed as more appropriate than others. Before detailing the methods and results of the study, it is thus instructive to provide a brief overview of the perceptible affordances of Instagram.

Perceptible Affordances of Instagram
Previous research on Instagram points to its use for phatic communication (Niemel€ a-Nyrhinen & Sepp€ anen, 2020), and its focus on aesthetics (Kofoed & Larsen, 2016). Photo-sharing practices on the platform are described as polished and well thought out (Larsen & Kofoed, 2015), and the platform's snapshot character as well as its focus on 'likes' have been suggested to prevent problematic content and instead suit the reproduction of dominant ideologies (Tiidenberg, 2015).
Whereas visual content is the intended focus of the platform, captions can also be added to an image. The extent of such text-based elements seems to be limited; Instagram does not explicitly state a maximum length of captions, but only three lines of text, or 125 characters, are displayed in the feed. Any exceeding characters will show after clicking "More", but this configuration suggests that users have a limited chance to arouse readers' interest in a longer caption. Similarly, comments have a maximum length of 300 characters. The easiestand most common-way of responding to a photo on Instagram is to use the "like" feature, materialized in a heart shaped button.
Besides these perceptible affordances, built around notions of visuality, aesthetics and social recognition, the service's Community Guidelines provide clues to its imagined and intended use. In the context of this article, the policy against "glorification" of self-injury is perhaps the most relevant: Encouraging or urging people to embrace self-injury is counter to [Instagram's] environment of support, and we'll remove it or disable accounts if it's reported to us. We may also remove content identifying victims or survivors of self-injury if the content targets them for attack or humor (https://help.instagram.com/ 477434105621119/). Content that does not comply with this policy can be anonymously reported, so that Instagram can "reach out to this person with information that may be helpful to them"-and, presumably, remove their posts. As shown in previous studies, however, getting content removed from Instagram is not as easy in practice as it may appear from their policy. It may take repeated reporting (Olszanowski, 2014, p. 88), and there are instances where the platform's moderation practices have not only proven ineffective and non-transparent (Feuston & Piper, 2019, p. 9), but also to contribute to hurting already vulnerable communities (Chancellor et al., 2016(Chancellor et al., , p. 1202). Instagram's censoring also appliessometimes temporarily-to selected hashtags, meaning that they return no results in searches although they can still be used to tag images (Moreno et al., 2016;Suzor, 2016). Other hashtags are apparently less controversial, as search results are displayed with a click-through public service announcement and optional mental health guidance framed by the question "Can we help?".
Hence, based on what can be known of Instagram's content restrictions, issues relating to suicidality, self-injury and eating disorders are of major concern to the service. The formulations in the guidelines suggest that the general ethos of the service is recovery-oriented and focused on peer responsibility and support, which also has a bearing on mental health support in a broader sense. It is in line with the service wanting to "foster a positive, diverse community" (https://help.instagram.com/477434105621119/) and promoting itself as a "fun" way of sharing precious moments (e.g. Olszanowski, 2014). This ethos, and the affordances in which it is inscribed, is likely to facilitate some support practices rather than others, as will be further discussed below.

Methods and Materials
Data for this study was collected at two different occasions that were a year apart (in March 2016 and March 2017 respectively). This allowed us to confirm that while individual users did come and go, the communicative patterns were fairly consistent over time. The collection included screenshots of the 100 most recent Instagram posts tagged with one of the Swedish hashtags #psykiskoh€ alsa (mental ill-health), #deprimerad (depressed) or #ångest (anxiety). This meant collecting 300 posts on each occasion, adding up to a total of 600 posts.
Clearly, this dataset is not current, and it was collected in relation to a now finished research project analysing online peer support and social media more broadly. To further validate the results of this study, however, we also collected the 100 most recent Instagram posts matching the hashtags at the time of writing in 2021 and found patterns that were highly consistent with what we saw in the original datasets. But there have still been some changes in the technological functionalities of Instagram and similar platforms in recent years. One such change is the broad adoption of so-called Stories. This format, consisting of short video clips or stills, sometimes set to music backings and customised using styled or animated text overlays or virtual stickers, is marked by its ephemerality (Vazquez-Herrero et al., 2019), as they are removed from view after 24 hours (unless the user chooses otherwise). Stories -which have rapidly risen in popularity since 2016 (Carah & Shaul, 2016) -have no doubt impacted on practices of self-disclosure, first on Snapchat (Bayer et al., 2016), but increasingly also on Instagram. It is therefore important to note that this present study is looking at the particular subset of user practices and affordances that relate to regular Instagram posts. This is a limitation of the data that we have at hand and, furthermore, while looking at Stories would be interesting in its own terms it demands other methodological and ethical considerations (see e.g. Bainotti et al., 2020). As a preliminary investigation into this, we performed searches for our three hashtags in focus on Instagram at the time of writing in 2021 and found no public Stories for #deprimerad (depressed) among the first 1000 search results, one (1) for #psykiskoh€ alsa (mental ill-health), and none for #˚angest (anxiety). As Stories only live for 24 hours on the platform, this reflects -at least -that at the time of this search, only one public story within the last 1000 relevant posts and/or the last 24 hours, matched our sampling terms.
In addition to the two above-mentioned datasets that were made the subject of close analysis, we used the Instagram Scraper (https://tools.digitalmethods. net/beta/instagram/#), and the Instagram Hashtag Explorer/Visual Tagnet Explorer (Rieder, 2015) to retrieve a larger set of metadata (relating to 1800 posts) which was used to get an overview of the general amount of response for different types of posts. While these posts were used to get an overview of the amount of response for different types of posts, the 600 manually collected screenshots were analyzed in detail and thematically coded with assistance of qualitative data analysis software ATLAS.ti.
While this article provides a qualitative analysis of the Instagram posts and therefore pulls its examples from a small and select number of posts we did, as clarified earlier, manually code the entire dataset of 600 posts. A post consists both of one or more images, and of a caption. The captions were coded based on their style (long, short, narrative, descriptive, etc.), while the images were coded based on what they depicted (selfie, nature, motivational text quote, etc.). We also coded the general message that was conveyed through the post (caption þimage) as a whole (recovery-oriented, help-seeking, etc.). When it comes to the latter, five user approaches emerged from reading and re-reading the data: (a) a personal, recovery-oriented approach; (b) a generic, recovery-oriented mode; (c) a personal, non-recovery-oriented mode; (d) a generic, non-recovery-oriented mode; and e) other. The latter category was excluded from further analysis. Additionally, and informed by Shaw's (2017) framework, the interpretative process focused on the ways in which specific Instagram affordances-i.e. photo sharing, captions, comments, likes and hashtags-were put to use. Each post was therefore coded with respect to the presence, form and content of these features. This meant that each post had four codes: one for what the image depicted, one for the caption style, one for the overall message of the post, and one stating which affordances were used.
To provide some quantitative contextualisation around the qualitative analyses, it may be relevant to share some general numbers. Approach (a) corresponded to 41% of post messages, while approach (b) amounted to 20%, meaning that 61% of posts were recovery oriented. Approaches (c) and (d) corresponded to 24% and 2% respectively, meaning that 26% of posts were non-recovery oriented. In the analysis below, we also discuss the category of motivational picture quotes (10% of coded images), and text-rich posts (18% of coded caption styles).
The three specific hashtags (Swedish tags for mental ill-health, depression, and anxiety) were selected for several reasons. First, they are in line with the study's intention to capture a broad spectrum of common forms of mental distress, rather than focusing on one specific condition. After an initial survey of English and Swedish hashtags related to such broad notions, the three selected hashtags also proved to be the most productive ones for the purpose of the study: they were frequently used, but they did not include too many off-topic posts (i.e. posts that did not in any way refer to mental health issues, apart from including one of the hashtags).
The coding procedure facilitated a detailed understanding of how these different modes co-occurred with particular uses of Instagram's features. By relating such patterns to Shaw's theoretical framework (2017) as well as to the previous research on Instagram's features, we were able to identify dominant, negotiated and oppositional uses in the material-each one associated with specific support practices. Dominant use of the platform, i.e use which conformed to the platform's designer-imagined affordances and which was also most frequent in the data, included posts best described as individual snapshots of recovery. Negotiated uses of the platform's perceptible affordances included two sub-types of posts: motivational picture quotes and text-rich posts. Lastly, oppositional use positions were taken up in posts showing support for nonrecovery oriented approaches to mental health issues.
It should be noted that this typology is not clear-cut and many posts exhibited traits from several categories, but the broad classification nevertheless represents significant trends in the material at large, and the extracts provided in this article are selected to illustrate the differences between posts.
While the collected data is publicly available-only public posts appear in Instagram search results-the sensitive subject matter requires consideration of ethical issues (cf. Markham & Buchanan, 2012). We have therefore chosen not to include any photos in the article. Furthermore, usernames are de-identified, and published quotes from captions and comments have been translated from Swedish to English. This procedure does not prevent the identification of individual posts and users, but it might complicate attempts at tracing them.

Individual Recovery Snapshots as Hegemonic Use
A large part of the posts in the material consists of selfies or other snapshots from everyday life, accompanied by short captions and a range of hashtags. While some of them have mental health as their topical focus, others are reports on routine events-such as meals, training and meeting friends-in which the poster's relation to mental health issues is only briefly indicated. However, a shared characteristic, in addition to the type of images and use of text elements, is their emphasis on individual recovery.
Post 1, illustrative of this category, includes a photo of two small plates with semlor-sweet rolls with whipped cream-on what looks like a kitchen counter. It has a minimal caption stating: "Go me h" and is accompanied by 30 hashtags. Many of them refer to eating disorders, e.g. #fuckanorexia and #anorexiarecovery, and others work to further emphasize the poster's relation to the photo, such as #iamproud and #challenge. Together with the very short caption, these hashtags provide a context for interpreting the visual content-in this case by constituting the poster as a courageous person battling a difficult condition. The post had received 86 likes at the time of data collection, and all comments applauded the poster's decision to eat a semla bun: The image of the buns echoes the "casual photograph" described by Manovich (2016) as the single most common type of image on Instagram; it serves as a documentation record rather than an aesthetic object. It appears to share a visual story and, in this sense, it is consistent with the platform's perceptible affordances and its focus on visual identity work through displays of everyday moments.
A similar example, albeit slightly more aestheticized, is a black-and-white selfie of a young woman with her eyes closed and her hair covering the right side of her face. Her hand is held in front of her mouth as if trying to keep herself composed (Post 2). The image is accompanied by four hashtags (#mentalillhealth, #anxiety, #gad and #neveralone) and the caption reads: "The ugly face of panic and anxiety. In any case, I'm going to the gym today, I'll cry my way there. I will fucking not let you control my life. Minute by minute, hour by hour, day by day." This post had received 41 likes and a number of boosting and affectionate comments: In both of these examples, captions are kept short and concise, with hashtags used to provide additional information. This can be understood as instances of hegemonic use, where posts and responses accommodate to the "productive constraints" of the platform (Stanfill, 2015) which puts the image center stage and downplays the importance of verbal communication. Notably, however, the textual features are deployed by posters to position themselves in relation to the visual content. By explaining the intended meaning of the image, they also clarify what type of response that could be expected from other users. While the images share a visual story in themselves, then, captions and hashtags seem required for the posts to be recognized as calls for support.
First, "likes" are one example of how support may be provided. As a paralinguistic affordance, the like has been described as a phatic element, or social media small-talk: it enables seemingly trivial communication that serves a social purpose without offering any substantive information (Hayes et al., 2016;Miller, 2008). Thus, it is a versatile feature precisely because it can be attributed a multitude of both content-based and relation-based meanings-in addition to literally signifying enjoyment of a specific media object, it may also serve as maintenance of friendship and expression of pity or reciprocity (Eranti & Lonkila, 2015).
Likes are in many ways similar to emoji, the extensive use of which is a main characteristic of comments. The emoji-rich language can be understood as another example of how Instagram's perceptible affordances are used in line with hegemonic conventions on the platform. It is arguably related to the platform's privileging of imagery over text, and emoji can thus be read as low-cost, space-saving signifiers that are performative of social bonds rather than conveyors of information-a form of easy-to-use resources for creating and maintaining support relationships (Miller, 2008). However the emoji vernacular also has normative dimensions by way of its actual content (Stark & Crawford, 2015). A closer look at the emoji used in comments on recovery snapshots shows that they are limited to symbols of affection (e.g. kissing lips and kissing face, hearts of different shapes and colors), approval (e.g. clapping hands, OK hand gesture, thumbs up, stars) and boosting (e.g. flexed biceps, fisted hand, raised fist). In other words, they form a seemingly standardized vocabulary for expressing emotions that aligns with the hopefulness and fighting spirit of the original posts.
The content as well as the response of Post 1 and Post 2 are illustrative of the dominant mode of relating to mental health issues found in this study. The posts are constitutive of a recovery-oriented subject position, where recovery is understood in terms of personal responsibility and individual emotion work-as seen in the first poster daring herself into recovery as well as the second poster's description of her ongoing battle against panic and anxiety. Notions of recovery as a matter of governing one's inner life resonate well with contemporary discourses on happiness and positive psychology, according to which positive emotions "come with the embrace of one's power to change one's emotional well-being, and with the assumption of responsibility for those emotions" (Binkley, 2011, p. 385). These discourses-and the emotional configurations invoked by them-have been conceptualized as modes of neoliberal governmentality, through which the therapeutic task is delegated to the individual (Binkley, 2011;Scharff, 2016).
The recovery-oriented position adheres to the platform's explicit policies and content restrictions, which seek to preclude the sharing of thoughts and behaviors defined as potentially destructive. Moreover, this position is facilitated by the perceptible affordances' highlighting of "mediated phatic sociability", where networking and connections are more important than the content of the dialog (Miller, 2008, p. 395). Instagram's built-in focus on aesthetics, together with its emphasis on publicly sharing visual stories from everyday life, appear to promote the furnishing and display of positive messages and identities that are likely to be appreciated by other users. For instance, the multitude of potential meanings of a like may evoke uncertainty and cause misunderstandings when someone likes a post with a clearly negative message (Sumner et al., 2017). Against this backdrop, posts with an orientation towards recovery and an investment in discourses of positive psychology lend themselves well to use of the like feature, as liking seems less ambiguous here.
Furthermore, recovery-oriented posts are privileged by the seemingly limited possibilities of textual communication, which prompt straightforward statements rather than nuanced and elaborated accounts-in captions and comments alike. Instead of explicitly asking for help, the hegemonic using position involves reaching out to gain recognition for one's efforts by visually communicating the battle against mental health issues and deploying textual features to inscribe it in a discourse of individual recovery. Interactions are largely centered on the poster, although users may engage in mutual exchanges of support on each other's posts.
At first, this type of communication may seem fleeting and shallow, but the phatic and emotive interactions can nevertheless be understood as enactments of social support. Judging from the number of posts following this pattern, they serve a purpose for a large part of Instagram users, and perhaps mainly for those who are already invested in a positive way of thinking about their current situation-depending on type of condition, illness trajectory and coping strategies.

Negotiated Uses
While many users take up a hegemonic position in posting on the platform, others tend to tweak its functions to better suit their own ideas and needs for social support. This includes adjusting its perceptible affordances, as well as making use of affordances that are less obvious or even hidden. Using Shaw's (2017) typology, such posts can be described as instances of negotiated use, and two examples are particularly prominent in the material-what we call motivational picture quotes and text-rich posts. Both conform to the same underlying ideology of recovery and positive thinking, but each involves a specific pattern of support practices which differ from the individual and phatically mediated support afforded by the platform's perceptible features.

Motivational Picture Quotes: Collectively Supportive Resources
Motivational picture quotes constitute a widespread means of communicating about mental health issues on Instagram. Such multimodal compositions where text is overlaid on an image-often called image macros-have previously been associated with humorous memes (Miltner, 2014;Shifman, 2015), but in this context, they are typically superimposed aphorisms or messages of hope on a stock photo or monochrome background. Examples include, for instance, a photo of a pink diamond against a white background, superimposed with the encouraging message "You're a diamond, dear. They can't break you" (Post 3), or a sepia-toned photo of a vintage bus on a winding road with the overlaid text (Post 4): I stopped telling myself that I am lost. I'm not. I'm on a road with no destination, I'm just driving with hope that I'll find a place that I like and I'll stay there. I'm not lost, I'm on my way.
Both of these images are widely circulated in online contexts beyond Instagram. For example, the image of Post 3 can be found under the tag "Strong women quotes" on websites such as www.picturequotes.com and www.quotegram.com, and the image of Post 2 can be traced back to the website Higher Perspective (http://www.higherperspectives.com/st/About). In addition to re-circulating such already existing picture quotes, users can also create and distribute their own pictures. However, Instagram's editing features do not allow users to add text on top of images within the application; for this purpose, users must instead turn to third party apps. In this sense, motivational picture quotes can be seen as results of negotiated uses of the platforms.
The generic character of motivational picture quotes provides a contrast to the hegemonic type of personal snapshots. While the image is the distinct focus, it is only important to the extent that it carries a verbal message, and captions are usually deployed only to emphasize the ideas already put forth in the text overlay. For example, Post 3 is accompanied by a short statement that situates the image in a health context, thus underscoring the need for continuous work towards recovery: "Every day is a battle against the illness. It is always present. Like diamonds we're unbreakable. We are diamonds. We can do anything!" Similarly, Post 4 has a caption that copies the picture quote and adds: "Anyone else that's just on their way? Stay strong my fluffy unicorns. You'll get there. I believe in you." In both posts, the hashtags used are broad and refer to physical and mental conditions (such as #chronicpain, #depression and #bipolar) as well as to ways of approaching them (#mentalillnessrecovery, #staystrong).
In contrast to the focus on individual recovery, seen in hegemonic use of the platform, first-person statements are rarely drawn on here. Rather than seeking support and recognition, posts including motivational picture quotes constitute generic supportive resources in themselves, targeting a broad audience of likeminded people as seen in the nonspecific use of "you" and "we" in the captions. The text overlay on the bus image indeed speaks of an "I", but the context downplays the significance of the poster's personal experience, instead focusing on reaching out to the "fluffy unicorns" in need of encouragement. In this sense, while motivational picture quotes are clearly recovery-oriented and inscribed in discourses of positive psychology, their use is also bound up with a strong emphasis on peer support and community-building.
The provision of support through this type of posts is recognized by others, and comments are generally appreciative and thankful. Furthermore, other users express identification with the posted content, as seen in the comments of Post 3:

User1:
Poster: Thanks sweetheart @User1 User1: I'm the one who's thankful that you keep spreading awareness and at the same time boosting and strengthening us.
User2: Nice and good pictures! At first glance, these comments appear similar to comments on hegemonic snapshots: in both cases, they are kept short and make use of emoji to signify enthusiasm. However, comments on motivational picture quotes are generally less person-centered, focusing on dialogue in ways that seem to reinforce the notion of a "we" put forth by the original poster. Such support practices that reach out beyond the poster, thus constituting a sense of shared experience and struggle, seem closely tied to a negotiated using position that extends the platform's functionality by deploying partly hidden affordances in the form of additional editing tools and ways of creating and sharing images.

Text-Rich Posts: Support Through Reciprocal Self-Disclosure
A number of posts are characterized by their emphasis on captions, which are text-rich and oftentimes with a clear narrative structure. They tend to give rise to a large amount of reponse, not least in the form of substantial comments and discussions. For example, Post 5 includes an image of a pencil-drawn mandala with the following caption: I told a colleague today that I'm having an evaluation and that they're leaning towards bipolar disorder. Got the reply, "That's great! Then you'll learn to bring down the highs and get up when you're low." I reply that I will probably need medication for that. Then I'm met with the opinion that medication is wrong. Why should that be wrong? If I can have a smoother and stable life with more balance, why would it be seen as bad that this is accomplished with the help of medicines? Wouldn't happen with many other types of illnesses.
A few hashtags accompany the image, for example, #neveralone, #sickinthehead, #challenge and #bipolar. While they serve to emphasize the status and attitude of the poster, they are not required for interpretation as the caption largely speaks for itself. The specific event accounted for, and the rhetorical questions posed in the caption, elicit a number of extensive comments which-taken together-form something similar to a comment thread in a discussion forum: User1: I had an evaluation. Thought bipolar. But no, I'm like this without having a diagnosis. But the medicine I get is also given to people who are bipolar. Actually, helps really well.
User2: I agree! I'm taking antidepressants myself and feel very good! Not planning to quit. Hugs and don't care what others might think.
Poster: @User1 I'm glad to hear! They were going to start giving me medicine last week but they wanted to do an evaluation first. My brother is not evaluated but takes medicine for it so in some way I feel some hope that things might become a bit more stable. Doesn't really matter whatever name they choose for the thing I have, but I need to try something new now.
Poster: @User2 Yes I'm taking two different antidepressants and I'm not ashamed of that. You don't take it for fun but because you have to User3: Sounds like a really stupid and unconsidered reply from your colleague. I think it would be good for you to get the "right" medicine for the "right" illness. The most important thing is that you feel well. Period.

User2: Word! Hugs to you
Poster: @User3 Yes I hope it's about lack of knowledge. But I think one comes across stupid comments about medicines and mental ill-health in various places. It makes me so sad and angry at the same time.
User4: I'm also being "gone through" because of suspicions of bipolar disorder, there are many different types of bipolar. If it turns out that I have it, I'm willing to try medication. It's incredibly naı¨ve to believe that you can control your ups and downs yourself, that's exactly the problem, that the controls are loose. It's as stupid as saying you can regulate your production of insulin if you have diabetes. For me it's a relief to have a thorough assessment regardless of the result. Hugs.
Poster: @User4 That's how I feel too! I hope to find a fairly right way of treating it. Regulating ups and downs by myself hasn't worked so far and I guess a possible diagnosis doesn't make any difference there. Hugs! User5: Like it would dissolve by itself, or be something you can learn to master. No one would get the idea of saying that to a person with diabetes or some other chronic physical diagnosis. You are the one who should be feeling well, not the others.
As can be seen in this extract, the text-heavy caption engenders reciprocal self-disclosure where three of the five commenting users relate back to their own experiences of medication in a form of "second stories" which validate the poster's standpoint that pharmacological treatment can be useful. In this sense, this post is productive of the type of support that has been identified in text-based online discussion forums (Kaufman & Whitehead, 2017;Giles & Newbold, 2013). Two commenting users do not explicitly draw on their own experiences, but they confirm the poster's interpretation of the recounted event, thus enacting a form of affirmational support. The poster also returns to the discussion, elaborating on her opinions and acknowledging points of identification between the users.
Because of its caption and comments, this type of post provides a stark contrast to the seemingly phatic communication that is the dominant mode of support on the platform-note, for example, that only one emoji is used in the comments of Post 5. Furthermore, while the image in this type of post can be either a selfie, an everyday snapshot, a work of art, or-more rarely-a motivational image quote, what distinguishes it from other forms of posts is that the image is entirely peripheral; it is the caption that communicates a message. Hence, by drawing on the partially hidden affordance of lengthy captions and comments, users negotiate the intended meaning of Instagram as a service dedicated to sharing images and transform it instead into a text-based social space more reminiscent of an online discussion forum. So, while the commenting function as such, is indeed part of the intended and thus hegemonic forms of communication on Instagram, these practices of appropriating them in ways that border on repurposing them can be seen as negotiated use of the platform.
Notably, this negotiated position seems to be taken up by clusters of users who post and respond to each other in similar ways, which might suggest that the text-rich format forges closer networks between users than the brief interactions seen in hegemonic uses of the platform. It is somewhat in line with Andalibi et al.'s observation that caption-heavy posts about depression include opportunities for potential coping (Andalibi et al., 2017(Andalibi et al., , p. 1490. Moreover, while the text-rich posts are constitutive of recovery-oriented subjectivities, similar to hegemonic use of the platform, their emphasis on textual communication opens up for more nuanced reflections and critical discussions-as seen, for example, in how users negotiate the social meaning and management of a mental health condition in Post 5.

Oppositional Uses: Support for Non-Recovery
Some posts are more clearly distinguished from the rest of the material in that they do not take a recovery-oriented stance and even seem to defy Instagram policies on "glorification" of self-destructive behaviors. Such posts are typically centered around a picture quote with a negative message, similar to the generic self-injury memes identified by Seko and Lewis (2016) on Tumblr. For example, Post 6 contains a picture quote in black and white, depicting a handwritten statement saying "I hate who I am". The post has received 33 likes but no comments. It has no caption but is accompanied by a number of hashtags. Unlike the hegemonic use of hashtags to communicate a desire for recovery, these tags refer exclusively to negatively charged concepts, such as #suicide, #deepcuts, and #worthless. Moreover, the post contains two hashtags which were originally invented to escape Instagram's ban on content that glorifies self-harm: #selfharmmm and #secretsociety123. Clearly, such posts oppose the discourse of recovery, pointing out more nihilistic and resigned positions.
A similar example is seen in Post 7, which includes a thinspiration-type of image: a sepia-toned photo of the lower part of a presumably female body, in which most of the legs are covered by a short skirt and stockings. The legs are extremely skinny. A superimposed text reads: "Eating feels wrong"-a message also emphasized in the brief caption: "Eating feels so fucking wrong! ". The image has 27 hashtags, including #stopeating, #skipdinnerwakeupthinner, #suicidalthoughts, #anaroxia, and the previously mentioned #selfharmmm. The post has received 31 likes and one comment, the latter being a mention of another user for the purpose of sharing the image.
The opposition expressed does not only have to do with the content of the posts, but also with how the users try to circumvent platform functionalites and policies. Both of these posts described above include visual and textual elements that could be seen as violating the service's policy; thus, they could be reported for removal. Yet, their mere existence stands as an example of how some users take up oppositional positions in using the platform (and get away with it, at least temporarily). The hashtags, in particular, highlight the power differentials embedded in the platform and the service's role in regulating communication. But they also illustrate how users tamper with such regulations by continuously introducing new terms and alternate spellings of keywords-note, for instance, the spelling of anorexia in post 7. In this way, users can still post and find content deemed controversial by the service, and the tweaked hashtags may also work to strengthen the sense of community between users with a similar outlook (cf. Ging & Garvey, 2017;Moreno et al., 2016).
At the time of writing this article, both #secretsociety123 and #selfharmmm appears to have been banned by Instagram, but they are most likely already replaced by other, slightly modified hashtags. These constant attempts at circumventing censorship can be seen as a way to take advantage of the hidden affordances of the hashtag feature by deploying it in ways contrary to its intended use. In fact, banned hashtags can even be understood as a type of false affordance: because they can still be added to posts, they look like they should be doing something that they cannot do-that is, return results in hashtag searches. By insisting on using them despite their banned status, users seem to reappropriate the specific hashtags as tools for identity work and signifiers of a certain mode of relating to mental distress, rather than as means of facilitating searches and public sharing of posts.
What these examples illustrate, then, is how some users reject a hegemonic position and instead find their own ways of interacting on the platform. This can be seen in the explicit transgression of rules and regulations, but also in how this type of post implicitly challenges the norms and ideologies that follow from Instagram's perceptible affordances. Rather than conceptualizing mental distress as something to be overcome through individual recovery and emotion work, the posts above inscribe the affordances in an interpretative framework where mental health issues are stylized, promoted and valued.
The examples also show how ideas about appropriate use of the platform are closely tied to what is considered to be acceptable strategies for coping, or for seeking and giving support. As Shaw (2017) noted, "misuses" of technology tend to be downplayed and understood as failures, especially when performed by people defined as marginal. Instead of seeing these non-recovery posts as misuses or failures, then, they could be understood as oppositional uses bound up with support practices that are not always recognized as such from a hegemonic perspective. By sharing non-recovery oriented content, the poster can gain recognition from others, as seen in the relatively large number of likes received by Posts 6 and 7. The lack of comments might be a consequence of other users not knowing what to say or perceiving comments as being too intimate, whereas liking is less intimidating (Sumner et al., 2017, p. 9). The likes may also indicate approval of the content of the post and identification with the stance of the poster.
In this respect, the sharing as well as liking of non-recovery posts can be understood as enactments of mutual support, where users express and affirm their feelings without revealing too much personal detail (cf. Seko & Lewis, 2016). Thus, while it must be noted that content endorsing self-harm and selfstarvation may pose a threat to some individuals, as stated in Instagram policies, the oppositional position may also involve forms of social support for users who are not willing to subscribe to dominant notions of recovery.

Concluding Remarks
In conclusion, our analysis of Swedish language Instagram posts relating to mental health support shows that the platform, by way of its perceptible features, tends to privilege one type of post: casual snapshots of individual recovery. Such posts tie in with dominant cultural notions of recovery through positive thinking and individual responsibility; thus, they are easy to respond to, and show support for, through the like button as well as the limited space and emoji-rich vernacular of comments. Because this is one of the most common type of posts in the material, it seems to involve a form of support that works well for many users. It might also indicate that people turn to Instagram when they are already on the path towards recovery, due to the communicative constraints of the platform.
However, the study also identified widespread negotiations of Instagram's perceptible affordances, of which two examples have been discussed: motivational picture quotes and text-rich posts. Both deploy partly hidden affordances of the service in ways that are performative of slightly different support practices. Whereas hegemonic use of the platform engenders support in the form of individual boosting, motivational picture quotes are recognized as collective supportive resources, and text-rich posts facilitate support through reciprocal self-disclosure and nuanced discussions. Still, both motivational picture quotes and text-rich posts are consistent with the platform's underlying ethos of recovery.
A third type of post, however, uses Instagram's affordances in oppositional ways, by defying user policies and providing mutual support and recognition for users who do not identify with platform-dominant notions of recovery-especially users suffering from self-harm and eating disorders. The oppositional using position is simultaneously marginalized and brought into being through the censorship practices of the platform. Notably, however, abusive or critical comments on such posts were not seen in this study. A similar tendency has been noted by Andalibi et al. (2017) with respect to depression-related imagery. This may be due to users editing the comment fields, or to Instagram removing controversial comments, but it is noteworthy as social media communication about self-harm and eating disorders has otherwise been associated with heated debates about authenticity and in-group boundaries (Boero & Pascoe, 2012;Horne & Wiggins, 2009;Lundstr€ om, 2018).
Taken together, the results of the study point to how platforms shape the way social support is enacted, but it also highlights how users tinker with the apparent limits of a service, imagining uses that are not necessarily in line with hegemonic perceptions and thus appropriating existing or hidden affordances to suit their own needs. It should be noted that most posters tend to attach themselves to one specific using position and type of post, rather than alternating between them. While it is beyond the scope of this study to map the characteristics of individual users, preliminary observations indicate that the social context of users, including gender, age, type of condition and illness trajectory, matter for how they find ways of navigating the platform. For example, the clusters of users that form around text-rich posts seem populated mainly by people presenting as adult women on their way back from exhaustion or burnout, whereas the posters of banned content-to the extent that they reveal their identities-are primarily people presenting as young women dealing with eating disorders and self-harm. This suggests that mental health communication and, more specifically, support practices are not only a matter of relating to the affordances of the specific platform, but also to the norms and vernaculars emerging in different groups of users. Exploring how hegemonic and nonhegemonic modes of technology use are bound up with performances of gender, age and particular mental health conditions would thus be an important question for future studies.