The Science of Sex: Computing Beauty, Part One

Pin it

Computing beauty — Part One: Morphing the Perfect Face.

“My dear, I just have to say how average-looking you are tonight!” said Fred the statistician to his date, only to receive a glass of 1990 Chateauneuf-du-Pape full in the face. “What did I say wrong?” he moaned, dabbing the wine from his cheek and jacket. “It was intended as a compliment. Numerical analysis has proven that average looks are the most attractive.”


Fred had a point. In common parlance, “average-looking” means midway between beautiful and ugly, or worse. Fred, though, was talking about a mathematical procedure that, according to some scientists, can extract the Platonic ideal of facial beauty from a streetful of real faces.


The basic observation was made by Judith Langlois and Lori Roggman of the University of Texas in 1990. They took photographs of the faces of a large number of people of the same sex and ethnicity, digitized the images and identified numerous landmarks — such as the outer corners of the eyes. Then a computer produced a composite image in which the layout and spacing of every feature represented the average of all the individual measures. Let’s call this image “Average Jane.” The individual faces and Average Jane were then shown to a bunch of men and women, who were asked to rate how attractive they looked. Average Jane was consistently judged more attractive than any of the real faces.

Langlois and Roggman suggested that computer-averaged faces are attractive because they come close to some internal image of the “ideal” or “prototypical” face — an image that is either genetically programmed into our minds or is based on our experience of all the faces we have encountered. According to this idea, we judge a face’s attractiveness based on how close it comes to that prototype.


But beauty is not that simple. A group at the University of St. Andrews, Scotland, led by psychologist David Perrett, took the averaging routine a step further. From the initial set of individual photographs, they selected a subset of individuals who were rated as more attractive that the others, and computer-averaged this subset. The resulting image — call it “Super-Jane” — was rated as more attractive than Average Jane. This result would not be expected if “averageness” were the sole criterion of beauty.


Finally, Perrett’s group programmed the computer to morph Super-Jane in a direction away from Average Jane. So, for example, if Super-Jane had wider spaced eyes than Average Jane, the computer moved the eyes even further apart. The resulting composite — let’s call it “Ultra-Jane” — was rated as even more attractive than Super-Jane. Again, this undercuts the notion that pure averageness is the sole criterion of beauty, for Ultra-Jane was far from average in every respect.


This finding is puzzling. You’d think that, after thousands of human generations, evolutionary pressure would have induced every fetus to develop a face as similar as possible to Ultra-Jane (or “Ultra-Joe”), but instead real faces cluster around Average Jane and Average Joe. There must be something that penalizes the quest for “ultra”-level beauty and that balances out the benefits that this degree of attractiveness would confer.


Here are some possibilities. One is that extreme beauty involves sacrificing useful anatomical structures. In women, for example, a small jaw is attractive, but small jaws are weak jaws, and a weak jaw means difficulty in eating — or did mean that during most of human evolution. With the introduction of cooking and eating utensils, however, strong jaws have become less necessary, so we may expect future generations of women to become smaller-jawed and more attractive.


Another possibility, suggested by Randy Thornhill of the University of New Mexico, is that ultra-beautiful people would be socially handicapped because they would be hard to recognize. Counter to one’s intuition, very attractive faces are indeed harder to recognize and distinguish than less attractive faces. This may be because they are too similar to our inner “prototype,” so that they have few features by which we can distinguish them.


In spite of the existence of “prototypes,” the judgement of beauty is clearly influenced by culture. For example, a person will generally rate a face as more attractive if it belongs to his or her own ethnic group than if it belongs to another group. Also, the criteria of beauty can change over time — a well-known example is the shift toward thinness as a criterion of beauty in our own culture. It’s thought that this change has come about because food is now so abundant that thinness no longer indicates an inability to obtain it. Thus, while evolution has predisposed us to find certain people beautiful, it has also given us the flexibility to make judgements appropriate to our circumstances.


So is there a message for the Freds of this world? Maybe it’s this: make yourself look as similar as possible to the way everyone else is trying to look, but add a tattoo or some other distinguishing mark so you don’t blend in and get lost among the beautiful people. Oh, and choose your words carefully when complimenting your dinner-date.