Ethics-testing an eating disorder recovery memoir: a pre-publication experiment

Abstract

Abstract Background Narratives (including memoirs and novels) about eating disorders (EDs) are typically published with the intention to benefit readers, but survey evidence suggests that reading such narratives with an active ED may more often be harmful than helpful. To reduce the probability of inadvertent harm and learn more about how narrative reading and EDs interact, a pre-publication study was designed to determine whether or not a recovery memoir should be published. Methods 64 participants with a self-reported ED read either the experimental text (The Hungry Anorexic [HA]) or a control text (Ten Zen Questions [TZ]) over a roughly two-week period. All participants completed the Eating Disorder Examination Questionnaire (EDE-Q) and the Anorexia Nervosa Stages of Change Questionnaire (ANSOCQ) one week before and two weeks after reading, and answered three recurring open-ended questions at regular timepoints during and after the reading. Computational analysis of the free-text responses assessed text/response similarity and response characteristics on emotional, sensory, and action-effector dimensions. Both rating-scale and free-text data were analysed using mixed ANOVAs to test for effects of time and condition, and the university ethics board was notified in advance of the quantitative threshold for harmful effects that would prohibit the ED memoir from being published. Results On the two quantitative measures, there was an effect of time but not of condition: Significant improvement was found in both groups on the EDE-Q (with a medium-to-large effect size) and the ANSOCQ (with a very large effect size). In an ANCOVA analysis, no significant mediating effects were found for age, education, duration of professional support for the ED, or pre/post-reading BMI change. For the free-text responses, linguistic similarity measures indicated that HA responses most closely matched the text of HA, with the same being true for TZ. In a word-norm analysis, text condition significantly affected six emotional, sensory, and action-effector variables (interoception, olfaction, gustatory, mouth, torso, and hand/arm), mean scores for all of which were higher in HA responses than TZ responses. Close reading of readers’ responses explored two potential mechanisms for the positive effects of time but not condition: engagement with the during-reading prompts as part of the experimental setup and engagement with the texts’ dialogical form. Conclusions The ED memoir was found not to yield measurably harmful effects for readers with an ED, and will therefore be published. The finding that significant improvement on both quantitative measures was observed irrespective of text condition suggests that positive effects may be attributable to linguistic characteristics shared by the two texts or to elements of the reading and/or reflective processes scaffolded by both. The quantitative results and the free-text testimony have implications for our understanding of bibliotherapy, “triggering”, and the practicalities of responsible publishing.