Can a Writing Task Reduce Anti-Fat Bias? Small Trial, Small Effects

Eine junge Studentin steht lachend auf einem sonnigen Campusweg in Colorado im Gespräch mit einer Kommilitonin. KI-generiertes Motiv, keine realen Personen.

Short, cheap interventions are attractive in research on weight stigma: a questionnaire, a writing task, ten minutes. A 2024 randomized trial from the University of Colorado Boulder tested exactly that. Participants wrote texts that were meant to make their own anti-fat attitudes feel inconsistent, and the researchers measured bias toward others and toward oneself before and after. This article is about how that trial was built and how far its numbers carry. It does not rate the writing tasks as a method, and it does not recommend them.

The paper is “Reducing anti-fat bias toward the self and others: a randomized controlled trial” by Emily B. Kramer, Evava S. Pietri and Angela D. Bryan, published on 18 April 2024 in the Journal of Eating Disorders (12:46, open access). The authors declare no competing interests. The open access fee was paid by the university library fund. The text I read contains no trial registration number.

How the trial was built

The participants were female-identifying undergraduates from a university subject pool, who received course credit. Everything happened in a single online session on Zoom, supervised by a member of the team, lasting 30 to 45 minutes. After a baseline questionnaire, participants were randomly assigned to one of three writing conditions, each with five prompts and a minimum of 15 minutes on task. Then the same measures were taken again.

Dot chart of the change in four bias measures before and after one writing task, by group; every group moved, the control group too.
Chart: Fatosphere Editorial. Source: Kramer et al., J Eat Disord 2024;12:46, Tabelle 3. Read on 09.10.2026.
ConditionWhat participants wrote
Dissonance about the selfAdvice to five fictional girls aged 9 to 11 who ask about weight and body image
Dissonance about othersArguments against five legal policies that disadvantage fat people, for example size-based hiring or social media algorithms that favour thin bodies, “regardless of your actual beliefs”
Control (“bias-consistent”)The same five policies, but arguing in favour of them

The idea comes from cognitive dissonance theory: people who argue against their own attitude may feel discomfort and then adjust the attitude. The self-directed task is modelled on the Body Project, an eating disorder prevention programme. The authors expected the self-directed task to work best. That expectation did not hold.

Four outcomes were measured: anti-fat attitudes toward others (AFAT, 1 to 5), fear of fat (GFFS, 10 to 40), weight bias internalization (WBIS-M, 1 to 7), and an Implicit Association Test (IAT) as the implicit measure.

What the authors report

Two of the four outcomes differed between the dissonance groups and the control group: attitudes toward others (AFAT) and fear of fat (GFFS). Weight bias internalization and the implicit measure did not differ by condition. The two dissonance conditions did not differ from each other. All three groups, the control group included, improved on all four measures from before to after. The authors name a few possible reasons for that overall improvement, among them sympathy triggered by any of the prompts and, for the IAT, a practice effect. They do not test between these.

Reading it carefully

The sample is narrow, and the authors say so. The participants were 18 to 24 years old, 89 percent white, with a mean BMI of 22.7. Seventy percent were in the “normal weight” range and 5.5 percent (11 people) in the obese range. Most said they wanted to lose weight (70 percent), and nearly half were trying to. That describes who was studied. It says nothing about how fat people would respond to these tasks, and the authors write that the results may not generalize to more diverse body sizes, to men, or beyond one session without follow-up.

The head count changes between sections. The abstract says N = 197. The methods say 200 were enrolled. Table 1 shows 198. The before and after table (Table 3) has 189 people with data at both times (63, 64 and 62 per group), which fits the degrees of freedom of a test reported later. I found no flow diagram and no stated reason for the missing people.

The effect sizes are small, and the text and the table seem to disagree. Attitudes toward others (AFAT) sit near the floor of the scale at baseline: 1.49 on a scale from 1 to 5. The text says the dissonance groups improved by 0.02 points more than the control group. From the means in Table 3, I get a difference of roughly 0.035. For fear of fat (GFFS) the text gives 0.45 points, while Table 3 implies about 0.8. The text values are about half of what I compute from the table in both cases, which fits a contrast-coded model coefficient. That is my reading. The supplementary model tables that would settle it (S1 to S4) were not part of what I read. Standardized effect sizes are not reported in the main text.

MeasureDissonance, self (before to after)Dissonance, othersControl
AFAT (1 to 5)1.50 to 1.461.45 to 1.381.54 to 1.52
GFFS (10 to 40)19.30 to 18.4018.92 to 17.9119.69 to 19.53
WBIS-M (1 to 7)3.12 to 2.903.11 to 2.883.33 to 3.20
IAT D-score0.41 to 0.260.50 to 0.380.41 to 0.30

The control group also moved. Because everyone improved, the before and after change of a single group cannot be credited to the writing task. Only the difference between groups can, and that difference exists on two of four measures.

One exploratory detail. In the subscales of the AFAT, the group difference appeared only for “physical and romantic unattractiveness”, not for “social and character disparagement”. The authors describe this analysis as exploratory.

Smaller inconsistencies. The text reports that 140 participants wanted to lose weight, Table 1 says 138. The design section names a mixed ANOVA, the analysis section multilevel models. None of this changes the picture, but it shows how carefully one has to read a single small study.

A note on weight-loss framing

The trial sits in the field of eating disorder prevention, and its sample was largely occupied with losing weight. This article reports that as a feature of the sample. It does not take weight loss as a goal and does not assess any programme that pursues it.

Why it belongs here

A trial like this shows what “reducing anti-fat bias” can mean in practice: one session, students, questionnaires, a few minutes after the task. That is a first signal about how attitudes respond to a prompt, not evidence of lasting change or of changed behaviour toward fat people. The authors themselves list the missing follow-up as a limit. For the wider picture of how such trials look across the field, see our reading of a review of 56 trials, which maps who was studied and what was measured across the field. The trial was run in the US with a US sample, so it says nothing specific about Germany or the UK.

Source: Kramer EB, Pietri ES, Bryan AD. “Reducing anti-fat bias toward the self and others: a randomized controlled trial.” Journal of Eating Disorders 2024;12:46, published 18 April 2024, doi.org/10.1186/s40337-024-00994-1. Full text (open access, PMC11025213) read on 9 October 2026.