You are here
Are faces more attractive when they are closer to the average of their ethnic group?
Potter and Corneille came up with the following study:(1, pdf)
Abstract: Face attractiveness relates positively to the mathematical averageness of a face, but how close attractive faces of varying groups are to their own and to other-group prototypes in the face space remains unclear. In two studies, we modeled the locations of attractive and unattractive Caucasian, Asian, and African faces in participants’ face space using multidimensional scaling analysis. In all three sets of faces, facial attractiveness significantly increased with the absolute proximity of a face to its group prototype. In the case of Caucasian and African faces (Study 1), facial attractiveness also tended to increase with the absolute proximity of a face to the other-group prototype. However, this association was at best marginal, and it became clearly non-significant when distance to the own-group prototype was controlled for. Thus, the present research provides original evidence that average features of faces contribute to increasing their attractiveness, but only when these features are average to the group to which a face belongs. The present research also offers further support to face space models of people’s mental representations of faces.
It has been repeatedly shown that attractive faces tend to be closer to the average of their ethnic group. But it is also well-documented that attractive non-European faces are closer to European norms than the average of their respective ethnic groups whereas attractive European faces are very European-looking, and the authors do not mention this. It is hardly necessary to cite this literature. East Asian Manga is full of characters with face shapes shifted toward European norms; East Asian aesthetic cosmetic surgery procedures cluster in the mid-face, generally shifting the face toward European norms; the models in magazines catering to the African-American community have faces shifted toward European norms, and so on.
Potter and Corneille cited a study showing that the attractiveness of European faces increased with a shift toward Asian norms, but it is easily seen how flawed this study was by looking at their version of an attractive European woman:
This image is taken from Rhodes et al. (2005).(2, pdf) Methodological criticism of this study –
The authors obtained averages by adjusting for distance between the eyes, which is an incorrect way to address shape in shape comparisons. The reason they made averages of faces within ethnic groups was to obtain an attractive face given that averageness is a well-documented correlate of attractiveness, but this method leaves out of analysis a much more powerful correlate of beauty in women, namely the extent of femininity. In addition, the average of attractive faces is rated more attractive than the average of non-attractive faces, and hence the choice of models used to make the ethnic group averages comes into the picture. Furthermore, the offspring of individuals from different continental populations do not manifest faces in between those of their parents (has been discussed within this site). So this study is completely useless.
So how did Potter and Corneille manage to come up with their results?
Here is how. In a multi-ethnic sample of faces, we have the following:
Total shape variation = shape variation within groups + shape variation between groups
To address the hypotheses the authors tested, the two components must be separated, but the tool they used to generate 3D faces, FaceGen Modeler, doesn’t do a proper job of separating these components when it generates random faces. It is not even designed to. The tool is predominantly used by game developers to come up with custom faces, and these faces need only look male or female or of individuals of various ethnic backgrounds (see the following example).
The face shown above is easily recognizable as one that approximates Angelina Jolie’s face, generated using FaceGen modeler, yet a comparison with the actual Angelina Jolie will reveal a number of differences, which cannot be eliminated no matter how one fine tunes the face using the software. So, the basic idea behind FaceGen modeler is to get a rough approximation. Download the fg (facegen format) file of Angelina Jolie’s face (by strigoi) and play with the trial version of FaceGen to see for yourself.
We also have:
Total shape variation = shape variation due to varying sex hormone levels + shape variation due to factors other than sex hormone levels
We can expand the first equation to:
Total shape variation = (within groups shape variation due to varying sex hormone levels + within groups shape variation due to factors other than sex hormone levels) + (between groups shape variation due to varying sex hormone levels + between groups shape variation due to factors other than sex hormone levels)
We know that many shape variables are affected by how the shape variable has been moulded by sex hormones and how it has been moulded by other factors. Thus, for instance, the nose will become more projecting with both increasing Europeanization and increasing masculinization. We know that the extent of masculinity-femininity is related to attractiveness. Hence, we need all four components separated in equation 3, which, again, FaceGen Modeler doesn’t achieve well when it generates random faces.
Once the four components are separated, the authors’ hypotheses can easily be tested by regressing the attractiveness ratings of individual faces against the components of total shape variation. To address the issue under investigation, one has to control for the component of total shape variation comprising of variation within groups. See the pdfs of the articles on beauty as shape and a derived preference concerning jaw shape for how to use geometric morphometric tools to separate the components of face shape variation; this would require using actual faces and recording landmark coordinate data for each face.
So what did the authors do?
They computed the distances between each face and the group prototypes. So, the four components of total shape variation were not taken into account by the software they used, and they were not separated in computing the distances between individual faces and the group prototypes. How did the authors hope to address their hypothesis?
If a correlate of beauty is related to approaching an average but also deviating from this average in a specific manner, then each correlate negates to some extent the effect of the other. Since the majority of face shape variation is found within populations, how do the authors hope to be able to detect components of beauty that increase attractiveness on one count but decrease it on another if the component of total shape variation that exists within populations is not removed/controlled for? The study is seriously methodologically flawed and useless.
Another shortcoming of this study was that the analyses were run on ratings of the front view of the face only, whereas the side view is more relevant to the ethnicity question.
- Potter T, Corneille O. Locating attractiveness in the face space: faces are more attractive when closer to their group prototype. Psychon Bull Rev. Jun 2008;15(3):615-622.
- Rhodes G, Lee K, Palermo R, et al. Attractiveness of own-race, other-race, and mixed-race faces. Perception. 2005;34(3):319-340.