Research Methods & Procedure

Procedure
           This study took a multi-stage selection of pro-Ana individual blog entries, pro-Ana poetry, medical blog entries pertaining to eating disorders, and poetry with subject matter pertaining to mental illness. It incorporated the three most popular search engines: Google, Yahoo, and Bing (GlobalStats, 2011). Before coding any of the website groups, we cut out all pictures, link lists, website menus, and website disclaimers and changed the text into a plain text document. We also recorded the date, search engine used, site generation, web address, author of each blog and poem, the credentials of that author (if necessary) and the linked sites that came with each sample. We also looked for duplicates among each searches and discarded the extra content. Moreover, if we noticed an author’s same passage or poem in multiple blogs or pro-Ana sites, we threw out the duplicate passage.

Pro-Ana Blog Entries Sampling
           To sample for the pro-Ana individual blog entries, the researchers searched the term “pro-ana blog” under the three search engines (terms from Borzekowski et al 2010; Harshbarger at 2009). According to Borzekowski and colleagues (2010), a pro-Ana website is part of a group of pro-eating disorder websites. These pro-eating disorder websites “encourage knowledge, attitudes, and behaviors to achieve low body weights” and emphasize “achieving extremely thin or skeletal appearances” (Borzekowski et al, 2010, p. 1526). Therefore, for this study, we only took sites which referred to themselves as “pro-Ana” and emphasized restrictive dieting behaviors. Similar to Borzekowski and colleagues’ (2010) content analysis of pro-eating disorder websites, this experiment sampled from sites found on the first 10 pages for each term search, under each search engine. We took the first 500 words from each blog and started with the most recent entries. The researchers then obtained links to other similar sites (“second generation” sites) through the first blogs found. Sites that included additional topics outside of pro-ana content or did not have 500 or more words we excluded. In addition, we omitted professional medical organizational websites, journals, or forums on pro-Anorexia from the pro-Ana individual blog sample.

Medical Blog Samples
           The researchers then searched medical blog entries sample by using the term “anorexia nervosa medical blog.” Unfortunately, there were not enough medical blog samples simply pertaining to anorexia nervosa. Therefore, the researchers expanded the search to “eating disorders medical blog.” For the purpose of having similar content to control for the pro-Ana websites, we excluded medical blogs whose subject matter specifically pertained to eating disorders outside of anorexia nervosa such as bulimia nervosa, binge disorder, eating disorders not otherwise specified (EDNOS), and other rare eating disorders (e.g. Pica disorder). We primarily based our decisions on whether the blog pertained to one disorder through the descriptors in its title. For each sample, we took the first 500 words, going systematically for the latest entry to the next latest. We then obtained the first ten pages of results from each search engine and then found second generation blogs through the first ten pages of results. Medical blogs that were not written or endorsed by professional medical personnel (RD, MS, MA, LMSW, DMSW, MD, LP, LMFT, PsyD, and Ph.D), experts in eating disorders (e.g. published authors who wrote on eating disorders), accredited medical organizations (eating disorder clinics, hospitals, small mental health private practices), medical research facilities (e.g. universities) or well known newspapers (e.g. The New York Times) we excluded. We also excluded any medical blogs which discussed eating disorders and another illness or issue (e.g. eating disorders and medical marijuana).

Pro-Ana and Mental Health Poetry Samples
            For the pro-Ana poetry samples, the researchers have searched “pro-Ana,” along with the term “poetry” on the search engines of Google, Bing, and Yahoo. This study has considered that works which were listed by their authors as “poetry” or a “poem” under a website which fits under our definition of a pro-Ana site constituted as “pro-Ana poetry.” Similar to the blog sampling, the researchers took samples from the first ten pages from each search engine, and if any site led to other pro-ana poetry, the researchers will include such second generation sites. We then took the first 300 words of poetry listed on the site. We did not find enough samples from the first ten pages, so we took other poems from the existing sites, and ensured that no poem was repeated twice in our analysis. The maximum amount of samples we took from one site was six.
As for the mental health poetry, the researchers obtained these samples by searching the term “mental health poetry” under the same three search engines and taking the first 300 words. The poems’ authors had to either state explicitly that the poem pertained to a mental illness and/or health—or be categorized by the website’s author/s as pertaining to mental illness and/ or health. We excluded poetry that had been selectively posted (e.g. poems shown due to high approval ratings) from the sample in order to make the sample as similar to the freely written pro-Ana poetry. Likewise, we looked at the first ten pages of results from the search engines, and also looked for second generation results from those first ten pages. 

LIWC Dictionary
           We then ran the samples under the LIWC program and obtained word percentages of language pertaining to Graham and colleague’s five moral dimensions. The five moral dimensions of purity, authority, harm/care, ingroup loyalty, and fairness each have two descriptive categories that indicate negative language surrounding the moral dimension (vice), and positive language (virtue). For instance, for the category of purity/vice, the language refers to descriptors of moral failings within purity (e.g. disgust*, deprav*, disease*, and  unclean*).