Facial femininity and perceptions of eating disorders: A reverse-correlation study

Eating disorders are prevalent in college students but college students are not accurate in identifying the presence of eating disorders (ED) especially when race is involved. Much has been researched about diagnostic ability in vignette form, but little outside of this. For example, it is not known how facial features, such as perceived femininity, may affect observers’ beliefs about the likelihood of disordered eating depending on race. In the present study, we examined how biases regarding facial appearance and disordered eating may differ depending on the race of face images. Using a technique called reverse correlation, we estimated the image templates associated with perceived likelihood of disordered eating using both White and Black Faces. Specifically, we recruited 28 college students who categorized White and Black faces according to perceived likelihood of an eating disorder diagnosis in the presence of image noise. Subsequently, we asked Amazon Mechanical Turk participants to categorize the resulting race-specific face templates according to perceived ED likelihood and femininity. The templates corresponding to a high likelihood of an ED diagnosis were distinguished from low-likelihood images by this second independent participant sample at above-chance levels. For Black faces, the templates corresponding to a high likelihood of an ED diagnosis were also selected as more feminine than low-likelihood templates at an above-chance level, whereas there was no such effect found for White faces. These results suggest that stereotyped beliefs about both femininity and the likelihood of disordered eating may interact with perceptual processes.

manipulated in facial femininity, both the Black and White race stimuli that were feminized, were more likely classified ad ED+.
The authors highlight the importance of their study as a "cornerstone" of a promising research field. The field would study whether facial appearance may affect the probability ("likelihood") of a person being perceived (and eventually mistreated) as likely or unlikely "victim" of an eating disorder, based solely on their appearance. They show that being of certain origin may affect the probability of being perceived as suffering from a certain illness (ED, in this very case). Authors also took the first step in identifying the traits that serve in such "folk classification" (sex-typicality).

Review of revisions
The authors took all my suggestions and objections into consideration and either extended/shortened the manuscript to address them, or replied why I might have been wrong. I appreciate their patience. Therefore, I recommend the paper for acceptance because, after the revisions, I can more clearly see its potential impact. The authors also state their willingness to proceed on this topic.
They added or extended the following sections within the manuscript: They re-wrote the abstract to emphasize that they studied "perceptual biases regarding facial appearance" and if they affect the diagnosis based on the race of the stimuli face. They also extended the introduction to stress that they studied (racially-specific) bias itself and did not identify the alleged "kernel of truth" in such a classification.
Throughout the description of methods, the authors also extended the section on the FaceGen database. I appreciate that. The extended description substantially improved my image on the utility of the database.
Thank you -we're glad to hear that this provided useful context.
The authors also edited the section on the image analysis. While they kept the section on the effect of image background/foreground colouration, they shortened the section on the effect of root mean square contrast.
In experiment 2, the authors added another short description of the stimuli processing within FaceGen. Authors also added the ethics statement and changed the term "Eating disorder" to ED (and of course, they explained the abbreviation).
Authors solved the technical issue of data-inaccessibility (regardless of whether it was a failure of mine or the OSF itself). They also cited the statistic software they use so that I could check the data and analyses. However, in my opinion, they only provide the JASP files for the Image statistical analysis. For the rest of the analyses, I could find only the raw MTurk data and classification images.
We apologize -these files should be available in our OSF repository at present.
Coming to the Response to Reviews section, authors replied in detail to my notions and suggestions, bringing further justification to their changes (or their decisions not to change the text according to the requests).
I accept their reply on my primary objections: I would like to thank them for explaining the study's actual goal and demonstrating that I was wrong in my suggestions. They explain what it is not a good idea to link the paper to evolutionary psychology. The revised version is much clearer regarding studying perceptual stereotypes based on facial traits and race (see also above).
I also accept that it would be extraordinarily complicated and ethically questionable to use photos of actual patients who suffer from eating disorders. In light of the manuscript's update, my suggestions on the use of different stimuli and study design are no more relevant. Also, the authors are right that reverse correlation (RC) is not a rare technique. My notion regarded just its usage within the field of face perception. Now, it is much clearer that both the goals and the whole "paradigm" of the current research report are much further from evolutionary psychology that I've been thinking. RC, therefore, seems a useful approach within the study of perceptual biases.
We appreciate the reviewer's thoughtful commentary about these issues and are glad to hear that our revisions helped clarify the key goals of our study.
They also claim that my suggestion to identify which features of the classification image affect the classification is partly resolved by the Image statistical analyses. Otherwise, they argue that capturing the aspects of texture that led to stimuli classification is above the article's scope of the article and exceeds the capacity and abilities of image analysis.

Suggestions:
Therefore, I have only several minor objections, which could, in my opinion, be solved easily: Thank you for these additional comments -below we describe how we have edited the text to address the remaining suggestions.
Page 2: "...their femininity may affect observers' beliefs about their eating behaviour" I would add: "only under specific circumstances (manipulation of stimuli)". -> The abstract has been modified upon my request. My request, however, did not address the difference in results between the first and second experiment-my apologies.

Thank you. We have included this in the revised references.
Page 8: "That is, given this particular sample of faces described in terms of their 3D shape and 2D surface properties, what is a compact dimensional model of the data that captures the correlational structure of the data? The resulting model makes it possible to assign spatial coordinates to each face such that groups of faces can be characterized in terms of their distribution within this space." The second sentence, in my opinion, do not answer the first question. Consider writing something like: "That is, given this particular sample of faces described in terms of their 3D shape and 2D surface properties, the compact dimensional model of the data based on the faces makes it possible to assign spatial coordinates to each face such that faces can be characterized in terms of their distribution within this space." Of course, if I miss the important message of the sentences, keep the original version.
Thank you. We have adopted this suggestion in the revised text.
Page 12: "...covaried with the appearance characteristics associated with disordered eating by our first group or observers." Did you mean "first group of observers"?

We did. Thank you! This has been corrected in the revised text.
Page 12/References: Authors should add the reference to the following citation: JASP, 2018 (currently missing in References).

We have added this to the references.
Page 20/References: Authors should add the reference to the following citations: Fink et al., 2006;Matts et al., 2007;Mitteroecker et al., 2015 (currently missing in References).
We have added these to the references as well.
Throughout the text, authors should unify the citation of the American Psychological Association data (either 2015 or 2013).

We have corrected these inconsistencies.
Reviewer #2: The authors did an excellent job in addressing reviewer comments. I only have a few minor additional comments.

Introduction
It is good that you explained the reverse correlation technique, but I would do it before you introduce the term. Perhaps refrain from introducing until it is explained.
We have edited the text to minimize references to the technique before we explain its use.

I found the explanation of the reverse correlation technique to be confusing
We apologize for the lack of clarity. We have attempted to re-work this part of the manuscript to make it more clear how reverse correlation paradigms work and why it is useful in this case.
Please elaborate why peer-identification of EDs is important and biases may be problematice.g., that peers could refer individuals to treatment, We have attempted to expand upon this point in the revised draft.

Methods:
I would minimize the use of acronyms -e.g., say MTurk instead of AMT workers, reverse correlation instead of RC We have adopted this suggestion throughout the text.

Discussion:
Please discuss in more detail the prevalence in men vs. women with EDs (i.e., what percentage of EDs is in men vs. women? Does differ by the ED?) This does vary by eating disorder -we have briefly included some of these statistics in the revised text.
Reviewer #3: Thank you for the opportunity to review this manuscript. This is an interesting topic that can be considered by readers. The Introduction section was very well prepared and comprehensively provides what is already known on this subject. The results are very interesting. However, after reviewing the manuscript, I have some minor comments: