New Study Asks the Question: Should This Book Be Published?

In the first part of this series, I described how I ended up accidentally writing an anorexia recovery memoir. Once it was finished, I realized that despite all my efforts to make it something that would be helpful not harmful to readers, I had no idea whether I’d succeeded.

The realization came when I was out gardening. I was peer-reviewing a journal article about representations of eating disorders in recent French literature, and the article included a line about the disingenuousness of authors who publish books about their own eating disorder experience despite saying that reading about other people’s eating disorders had only ever been unhelpful for them. I didn’t want to be one of those disingenuous authors, and I definitely didn’t want to publish something that made recovery harder rather than easier for anyone. The sentence that really woke me up, in that article I was reviewing, included a citation of my own previous research (Troscianko, 2018):

However, in the light of Troscianko’s study, such claims may appear misguided, or even disingenuous; in some cases, authors claim that their goal is to help or inspire others and include details of how their own reading of ED memoirs and novels fuelled their disordered eating. (Jones, 2020)

Seeing my own work enlisted to get this point across made the message hit home even harder: I must not ignore the implications of my own research. Not long after I’d come in from my weeding, it became pretty rapidly clear to me that an experiment was needed. The research referenced in the article was a survey study: asking people about their experiences of reading narratives about eating disorders and their preferred other genre. What was needed here was something experimental not observational: something with a control group and randomization that let you come to conclusions about causality. And in this experiment to supplement the efforts and hopes with evidence, there would need to be stakes: specifically, it would need to be completely clear whether the book had passed or failed the test—whatever I decided the test should be.

This was not a conventional experiment; I know of no other that has ever aimed to determine a publish or don’t-publish decision for a work of memoir or fiction. With no close precedents to draw on, it took me a long time to work out the design, including what exactly to measure, what the control condition should be to compare against reading the book, and who the participants should be. (For some of the methodological questions that needed resolving, see my discussion here.) Then it took a while to get ethics clearance, and a lot longer to recruit enough participants, analyse the data, write the paper, through the peer review process, and published: four years in total, from that Covid summer afternoon pulling up brambles.

Part of the ethics approval process included demonstrating that we had structures in place to guard against experimenter bias since this was me leading a study on my own book. Once it was done and I was writing up the results for publication, I had a disagreeable conversation with a senior medical humanities professor whose response to my one-line description of the experiment was: “Well, you’re the last person in the world who should be doing that project, aren’t you?” Her point was that the author of the book should not be the lead experimenter. But I think she was wrong, as well as profoundly uncurious. I made sure the design was robust, of course, and employed a colleague to take care of participant recruitment and day-to-day management of the study, while another colleague did the quantitative data analysis. All this helped run the experiment and that the results made sense as rigorously and seamlessly as possible.

But ultimately, it didn’t feel like there was a serious risk of bias, since pass or fail for the book were both wins for me. As with any well-designed experiment, the great thing about this one is that we would learn something useful whatever the outcome. And more pointedly, in either case, I would get to avoid the scenario where I publish something and it’s horribly triggering or otherwise damaging for people. Of course, no experimental results can provide guarantees for any individual in the future. But at least performing this one would give me a confident sense of whether the book was likely to do harm or good or neither.

In the final part of this series, I’ll cover what we found and what the broader implications are.