New normal for body size perception: Participants consistently mis-categorized overweight male bodies as normal weight, while accurately categorizing normal weight

Misalignment between perceptual boundaries and weight categories reflects a new normal for body size perception. Annie W. Y. Chan, Danielle L. Noles, Nathan Utkov, Oguz Akbilgic & Webb Smith. Scientific Reports volume 11, Article number: 10442. May 17 2021. https://www.nature.com/articles/s41598-021-89533-5

Abstract: Combatting the current global epidemic of obesity requires that people have a realistic understanding of what a healthy body size looks like. This is a particular issue in different population sub-groups, where there may be increased susceptibility to obesity-related diseases. Prior research has been unable to systematically assess body size judgement due to a lack of attention to gender and race; our study aimed to identify the contribution of these factors. Using a data-driven multi-variate decision tree approach, we varied the gender and race of image stimuli used, and included the same diversity among participants. We adopted a condition-rich categorization visual task and presented participants with 120 unique body images. We show that gender and weight categories of the stimuli affect accuracy of body size perception. The decision pattern reveals biases for male bodies, in which participants showed an increasing number of errors from leaner to bigger bodies, particularly under-estimation errors. Participants consistently mis-categorized overweight male bodies as normal weight, while accurately categorizing normal weight. Overweight male bodies are now perceived as part of an expanded normal: the perceptual boundary of normal weight has become wider than the recognized BMI category. For female bodies, another intriguing pattern emerged, in which participants consistently mis-categorized underweight bodies as normal, whilst still accurately categorizing normal female bodies. Underweight female bodies are now in an expanded normal, in opposite direction to that of males. Furthermore, an impact of race type and gender of participants was also observed. Our results demonstrate that perceptual weight categorization is multi-dimensional, such that categorization decisions can be driven by ultiple factors.

Discussion

By providing corroborating evidence from univariate and multi-variate analyses to investigate body size perception, we are able to identify the complex relationship between gender and race types of the stimuli and of the participants, and the impact of these factors on body size categorization. In particular, we have revealed that performance (percent accuracy) for body stimuli is not uniform, whereby participants performed best for Normal weight and worst for Obese stimuli. We also revealed evidence of an interaction between weight category and gender of the stimuli; participants were more accurate for Underweight and Normal male stimuli (leaner size) relative to the same weight category of female bodies, but they were more accurate for overweight and obese female stimuli compared to male bodies of the same size.

Multi-variate decision tree analysis provided not only consistent results but has pinpointed the direction of estimation errors. Specifically, it revealed that while our participants were reliably making more under-estimation errors for Overweight and Obese male stimuli, they were quite accurate when categorizing Overweight and Obese female stimuli. The overall decision tree pattern (Fig. 2 and Supplemental Figure 2) suggests a strong bias for male stimuli where participants showed an increasing number of errors from leaner to bigger bodies, particularly under-estimation errors. Importantly, there was an expansion of Normal weight category, such that for male stimuli, while a high percent accuracy was found for Normal weight (Fig. 2B), there were also substantial under-estimating errors for identification of Overweight male bodies (Fig. 2C), where participants consistently mis-identified Overweight as Normal weight. Thus, it ndicated that the perceptual BMI for Normal male bodies is now higher than the recognized BMI. For female stimuli however, an expanded boundary for Normal size was found in the opposite direction. Participants have categorized Underweight as Normal (Fig. 2A) as well as accurately categorized Normal bodies (Fig. 2B), suggesting that the averaged perceptual BMI for ormal female bodies is now lower than the recognized BMI. The fact that both male and female participants shared many of the same biases also suggests that visual learning plays a critical role in developing these specific biases. Part of these results was consistent with previous findings; for example, it has been reported5,6 that participants (predominately Caucasian women) make more under-estimation errors when judging Caucasian Overweight or Obese male bodies. It is also worth highlighting that due to the diverse range of our stimuli and equal sampling across both gender and race of participants in the current study, we are able to capture opposite response patterns when judging male vs female bodies, thus, a more complex pattern than previously reported.

As mentioned earlier, prior work has primarily reported perceptual biases for own body weight that might associate with race or gender types, while others have reported biases when judging others’ bodies, but theyhave primarily focused on a particular gender, race, and/or weight category of the stimuli or participants. Our current study focused on assessing others’ body weight as observers, accounting for race and gender of both stimuli and participants. We found that perceptual errors could be associated with characteristics of the participants and the stimuli. For example, all participants, regardless of their race and gender, showed more under-estimating errors by mis-categorizing AA female overweight bodies as normal (Fig. 3). Intriguingly, (Fig. 2A) when judging female underweight bodies (all race types), AA participants (both genders), showed a stronger over-estimating bias for female underweight bodies, mis-categorizing those images as normal weight. CA participants, however, did not exhibit such bias.

It has been well-established in the face perception literature that people are more accurate in recognizing and identifying faces of their own race compared to other race groups. This discrepancy in performance is known as the “other-race effect”32,33,34,35,36,37,38,39,40. In the context of body weight perception, our current results did not show any other-race effect; no interaction between stimulus and participants’ race types was found in the univariate analysis or multi-variate analysis. However, we identified various participant-specific race effects from the multi-variate results. For example, we found that AA participants were slightly better at categorizing Obese male bodies relative to CA participants. AA participants also over-estimated Underweight female bodies, as they had consistently mis-categorized Underweight female bodies as Normal size (and this effect was not present in CA participants). For normal male bodies, AA performed better than CA, but AA made more over-estimation errors than CA. Interestingly, some stimuli-specific race effects were also identified. Overall, categorization performance was slightly (but significantly) better for Avatar stimuli in terms of percent accuracy. Multi-variate analysis further revealed that during categorization of female overweight stimuli, participants showed higher accuracy for Avatar and CA bodies than for AA (Fig. 3). A recent study20 had used visual adaptation to study the after-effect following repeated exposure of Asian or Caucasian female bodies, and their results seemed to be consistent to our findings. They also reported a lack of “other-race effect” at the stimuli level, but they reported that Asian participants seemed to show a weaker adaptation effect relative to Caucasians; however, the effect was not specific to Asian or Caucasian stimuli.

“Own-gender biases” have also been reported in face perception literature41. People are better at recalling or recognizing faces of their own gender relative to faces of the opposite gender41,42. Limited work has been conducted regarding gender-biases in body perception. Our recent study43 investigated gaze-pattern during perception of upright vs inverted bodies, but observed no differences in eye-movement patterns between male and female participants during a same/different categorization task of male body images. Multi-variate analysis in the current study has identified significant differences in performance between viewing female and male body images. There is a stimuli- and participant-specific gender effect that is particularly prominent for Obese bodies. Specifically, a marked difference was found between male and female participants, where male participants showed significantly higher accuracy for female Obese bodies. For male Overweight bodies, male participants performed better than female, while female participants showed more under-estimation errors. This suggests that under-estimation bias for Overweight male bodies was primarily driven by female participants. A stimuli-specific gender effect was also observed whereby, consistent with the univariate analysis, participants performed more accurately for Underweight male bodies than female Underweight bodies. Overall performance for Normal weight images was also better for male than female bodies. For Obese bodies, performance was better for female bodies, and there were also more under-estimation errors for male Obese bodies. These findings demonstrated that, by increasing the diversity in the stimuli and participants tested and by adopting a multi-variate approach, a more complex categorization pattern can be revealed. Furthermore, our observations of behavioural biases for higher BMI male stimuli and for lower BMI female stimuli seem to be consistent with the idea that partial overlapping or multiple gender-specific neural mechanisms may be at play during body size perception24,25.

Two major theories have been adopted to elucidate perceptual weight biases: the Weber’s law and contraction bias12,13,44. Specifically, the Weber’s law would predict that since detection of change of one’s body size is in constant proportion with one’s own weight, it is more diffcult to notice the change when one is overweight/obese. Alternatively, contradiction bias predicts that one’s perceived own BMI is inversely correlated with their own actual BMI. It has been reported that such correlation was only found during size estimation of participants’ own avatar, but did not generalize to estimating others’ body size12. While these theories may be helpful for explaining error in estimating one’s own weight, it is rather difficult to apply them to explain errors/biases during identification of others’ weight, especially when there are a lot more variables (race, gender, body weight, etc.) when dealing with “other bodies”. As we have shown here, estimation accuracies and errors interact with the type of stimuli presented in the experiment, thus illustrating that with increasing diversity in the stimuli, it might not be possible to show an “one-to-one mapping” using the above theories, as estimation decisions might be more complex than previously thought. While it is important to recognize that people have different body sizes, shapes, and other physical characteristics19, and that even BMI cut-off points may not capture variations in physiological measurements across cultures45, our current approach aims demonstrated that it is possible to capture and quantify some of the multi-dimensional visual characteristics, and it is critical that future work should also harness similar approaches.

Our findings here certainly do not attempt to capture categorization patterns for all types of bodies, and despite the constraints in our well-controlled paradigm (in real life, people with the same BMI may have different body shapes, and we see bodies from many different viewpoints other than straight-on), we have taken an important first step to quantify complex patterns in body weight perception. Finally, we believe that providing a careful characterization of perceptual biases in body weight here may lead to better diagnostic decision-making and development of personalized intervention programmes in both clinical and non-clinical settings.