Fig 1.
Sketch plot of estimated BMI (y-axis) as a function of actual BMI (x-axis).
The perceptual effect of contraction bias is apparent as a consistent regression slope of < 1 across all three regression lines: Actual BMIs below the population average are overestimated, actual BMIs above the population average (i.e.,~25) are underestimated, and actual BMIs close to the population average tend to be accurately estimated. Meanwhile, the attitudinal effect, captured by global EDEQ scores, is reflected in the systematic upward shift of the three regression lines as the EDEQ scores increase, reflecting increasing attitudinal body image concerns. This suggests that for any given actual BMI, the self-estimated BMI increases by a fixed amount according to the EDEQ score.
Table 1.
Eating disorder histories over time.
Table 2.
Participant characteristics.
Table 3.
Pearson correlations between psychometric subscales, together with PCA factor loadings.
Loadings <0.4 are not shown.
Fig 2.
(a) The relationship between participants’ BMI (x-axis) and their subjective estimate of body size (y-axis) for women with a history of Anorexia, Bulimia, and OSFED (red) and healthy controls (blue). The effects of age, PSYCH, and FAMPEER are statistically controlled. (b) The relationship between participants’ BMI (x-axis) and fitted values of estimated body size computed from the linear mixed effects model at ±1.5 SD for the value of PSYCH for each group. In both (a) and (b), the black dashed line represents the veridical relationship between estimated and actual BMI, with a slope of 1, indicating fully accurate estimation of own body size.
Fig 3.
A scatterplot of the relationship between participants’ BMI (x-axis) and their subjective estimate of body size (y-axis) for women with a history of Anorexia and OSFED (red) and healthy controls (blue).
The effects of age, PSYCH, and FAMPEER are statistically controlled. The black dashed line represents the veridical relationship between estimated and actual BMI, with a slope of 1.
Fig 4.
a) Cartoons illustrate bodies in a range of BMIs (blue) that people are exposed to in the real world and the mapping of these bodies onto a mental representation with a compressed range (red). This gives rise to the wider (blue) and narrower (red) distributions centred on the population average. b) Shows how the stimulus (S) response (R) relationship has a slope < 1 when participants estimate their body size. The dashed black line in the graph in b) represents veridical performance (i.e. perfect accuracy), where the slope = 1.
Fig 5.
a) Cartoon illustrates how individuals with eating disorders might learn to associate narrow BMI ranges into rank orderable categories. b) Shows increasing BMIs across the stimulus range (S) are correctly mapped onto the appropriate response (R) categories, leading to accurate performance, where the slope of the regression between the two is 1.
Fig 6.
a) The lower half shows uncertainty around possible response choices (red arrows) to a lower BMI stimulus, with a small shift to the right to ensure that the final choice (green arrow) is unlikely to be incorrect. The upper half shows wider uncertainty in the possible response choices (red arrows) to a higher BMI stimulus, with a larger shift to the right to ensure that the final choice (green arrow) is unlikely to be incorrect. b) Accurate performance at lowest BMIs for stimuli in the ‘expert range’, with increasing overestimation as the right shift (green arrows) in a) systematically increases as a function of stimulus BMI, leading to a slope > 1.