Expressions: A Look at Facial Symmetry and Its Implications on the Perception of Beauty

People actually prefer slightly asymmetrical facial features

"Beauty is truth, truth beauty,—that is all 

Ye know on earth, and all ye need to know."

John Keats (1795–1821) Ode on a Grecian Urn

While symmetrical faces are perceived to be attractive,
completely symmetric faces are disconcerting and are not perceived as normal.

Recently, while investigating the finger length ratio, I learned that the length of a person's ring fingers is determined by the amount of testosterone present in the womb during pre-natal development. The more testosterone, the longer the ring finger. This is true for males as well as females, but the correlated behavior traits are more pronounced in men. Other characteristics that are influenced by androgens in utero include a person's facial symmetry. The more testosterone in the womb, the more symmetry in adult facial features. 

Right after calculating my 2D:4D ratio (and trying to interpret it), I set out to investigate the degree of symmetry in the face of my closest available subject, moi. For readers who share my curiosity, here is a quick tutorial on how to select and manipulate a photo to investigate facial symmetry. I'll conclude the post with some thoughts on why this matters.

Original

Select a photo that is evenly lit and as nearly straight on as possible. Stoic expressions are fine, but I think it makes sense to find an expression that most people would recognize as natural.

Construct a grid and rotate the picture so that it is as nearly vertical as possible. You may need to focus on getting the eyes on the same plane, particularly if there is no distinct vertical axis. 

Modified original, aligned and cropped

Crop the photo to provide a small border around the area of interest (face). 

Now you are ready to make the vertical cut that divides the face into two halves. Try to slice down the axis of symmetry, splitting the forehead, nose, teeth, and chin as evenly as possible in one straight vertical line. You may have to minimize errors by splitting differences. That is, if the nose is off-center a little to the right and the teeth are off center a little to the left, the best vertical line would minimize the errors by splitting the difference between the nose center and the teeth center.

Next you are going to flip the images by rotating them around the vertical axis you have just described. As you create the four halves, give them distinct names like Right (i.e., the subject's right) and Flipped Right (i.e., the result of rotating the right image vertically 180 degrees). You'll also have Left and Flipped Left halves.

Right Symmetry

Once you create and identify the four halves, you can start mixing them up to create four different combinations. Two right halves, the right and flipped right, placed together side-by-side form the Right Symmetry view (shown above).

Left Symmetry

The left and flipped left halves create the Left Symmetry view (shown above).

Mirror

Put the two flipped sides together to create the Mirror view (shown above). At this point, you are ready to construct the complete set of four views.

Once you have the four essential views, you will need some software for consistent, reproducible evaluation of the degree of symmetry in each of them. Obviously, the constructed right and left symmetry views should have a very high degree of symmetry, and the original and mirror views should have similar, if lesser degrees of symmetry.  

Symmeter (http://www.symmeter.com/index.html) is a web-based system that provides a simple way to measure the symmetry of any person, place or thing using a digital image. Symmeter requires a small file, 100K or less, in JPEG format. If you want to convert files without downloading software, I had good luck with Zamzar (http://www.zamzar.com/). They have a fast and easy tool for converting your large and rich .png files to simpler .jpg files. As for Symmeter, it uses some older technology that does not mesh well with Google or Firefox, but I was able to get it to work with Explorer. 

Upload your first file to Symmeter and choose bi-lateral symmetry. Don't choose radial symmetry unless you are evaluating a wheel or a pizza. Define your area of interest with a rectangle or an ellipse. I found that I got better results, i.e., a higher symmeter value, when I used an ellipse instead if a rectangle. The square may be better for folks with wider faces. The system returns a number called the Symmeter Value, which is the percentage of symmetry in the area defined. The Symmeter value for most faces will be in the mid to high 90's, which is why the precision of the decimal places is important.  I rounded the fourth decimal to create the table below.

Notice that the Original and the Mirror are fairly close in value, as expected, but the Mirror is higher. I am not sure why they would not be exactly the same. Perhaps the photo was not taken dead-on, or perhaps the slightly uneven lighting casts some shadows that throw the results off. Notice that both Right and Left Symmetry views are very close to 100% symmetrical, as expected, with slight improvements using the ellipse.

Notice, too, that I added the results from a stock photo of George Clooney to the results table. I concede that George is an exceptionally handsome man. Notice that the image of George Clooney has a Symmeter Value of 94.792, noticeably less than the primary subject. The statement is not that I have a more symmetrical face than George. Even if that is true, my point was more to the fact that perceptions of beauty are only partially based on symmetry. Collecting the symmeter value is an exercise in the pursuit of truth. The truth has a beauty all its own. There is no positive correlation between Symmeter value and aesthetic appeal (or if there is, that is not my point). Case in point: the image labeled Left Symmetry is more appealing than the one labeled Right Symmetry, even though they have similar and very high bi-lateral symmetry.    

There are plenty of web sites that discuss the issue of facial symmetry and beauty. The primary subjects seem to be "Beautiful People" celebrities like Tom Cruise, Angelina Jolie, Brad Pitt, and the aforementioned George Clooney. Although symmetry is theoretically desirable, plenty of folks with asymmetric faces have adorned the cover of People magazine. 

Gorgeous George, charter member of Jimmy Kimmel's
Handsome Men's Club

Symmetry is not necessary for aesthetic appeal. Few things in nature are symmetrical, and as stated at the outset of this post, completely symmetric faces are disconcerting and are not perceived as normal. Since I know that my Left Symmetry is more visually appealing than my Right Symmetry, I will use that to my advantage. But I will also be suspicious of superficial beauty, now that I understand how easily my eyes can be fooled.  

This brings us to the concluding thoughts about my motives for this investigation and related inquiries into physical and behavioral characteristics that, to some extent, are determined genetically. I do not care about finger length because a certain 2D:4D ratio is "desirable." Nor am I interested in facial symmetry because a certain Symmeter Value is "superior." I do not think one MBTI profile is "better" than the others.  But I do care about understanding and appreciating the truth inside of biological diversity.  There is real beauty in that pure truth.

Next steps in the inquiry would probably involve relating the degree of symmetry to the finger length ratio an a large number of subjects. The hypothesis would be that if a developing fetus is exposed to a certain level of androgens in utero, and if the effects of  this exposure are increased facial symmetry and longer ring fingers (among other effects) then these effects should be correlated. Perhaps someone has already started this investigation.

Is aesthetic appeal the same thing as beauty? If we believe Keats, aesthetic appeal can be a mirage, but the truth is beautiful.  

"Beauty is truth, truth beauty,—that is all 

Ye know on earth, and all ye need to know."