Past visual experiences weigh in on body size estimation

Abstract

Body size is a salient marker of physical health, with extremes implicated in various mental and physical health issues. It is therefore important to understand the mechanisms of perception of body size of self and others. We report a novel technique we term the bodyline, based on the numberline technique in numerosity studies. One hundred and three young women judged the size of sequentially presented female body images by positioning a marker on a line, delineated with images of extreme sizes. Participants performed this task easily and well, with average standard deviations less than 6% of the total scale. Critically, judgments of size were biased towards the previously viewed body, demonstrating that serial dependencies occur in the judgment of body size. The magnitude of serial dependence was well predicted by a simple Kalman-filter ideal-observer model, suggesting that serial dependence occurs in an optimal, adaptive way to improve performance in size judgments.

Figure 1

Average performance in the bodyline task. (A) Mean size judgments given to each of the seven categories of body, which varied from very thin to very overweight. Error bars represent ± 1 s.e.m. The solid line represents the best fitting linear regression (slope 0.68, R² = 0.99). The dotted line represents linear use of the bodyline, without scaling. (B) Average precision thresholds, given by standard deviation of bodyline judgements, as a function of body category. Bars show 95% confidence intervals, almost all of which span the mean, suggesting that precision varied little with body size.

Figure 2

Individual data for the bodyline task. (A) Histogram showing the distribution of coefficients of determination (R²) for the linear fit. Most are above 0.95, suggesting that the categories were perceived as equidistant, and mapped accurately, save for a scaling constant. (B) Regression indexes of the individual subjects (1 minus the slope of best fitting linear regression to their bodyline data) as a function of precision thresholds, defined as average standard deviations for judgments at each category. There is no significant correlation between the two variables. (C) Magnitude of serial dependence of individual subjects (defined as the slope of the regression line for similar previous body sizes, illustrated in Fig. 3) as a function of regression index. Again there is no significant correlation, indicating that the two processes are independent. (D) Magnitude of serial dependence as a function of precision thresholds. There is a strong and significant correlation, with higher thresholds leading to greater dependency, as predicted by the Kalman filter model (eqn. 8). The top right data point in (D) is not an outlier but nevertheless we re-ran the analysis without this individual. The correlation remained highly significant: r₍₁₀₂₎ = 0.56, p < 0.0001.

Figure 3

Serial dependencies in body size estimation. Data show the average biases in the perceived size (difference between perceived and physical size), as a function of the difference in size of the body on the preceding trial. Data are averaged over all observers and body categories. Error bars represent ± 1 s.e.m. The continuous curve shows the predictions of the parameter-free Kalman filter model (eqn. 8). The horizontal dotted line plots the average bias, which is slightly negative.

Figure 4

Visual depiction of the bodyline task, in which a female body image was presented for 250 ms, immediately followed by a visual noise mask for 500 ms. Participants indicated the perceived size of the image by clicking on the bodyline delineated with extreme female bodies as anchors presented a further unit of scale beyond the bounds of the numberline. For illustration purposes the females are represented by synthetic body images created in Poser®.