Last updated
Facial averaging
Facial averaging means blending many faces into one composite and then asking how that composite is judged. The idea is more than a century old, and one of its findings is famous: averaged faces tend to be rated as attractive. The later work is more careful about why. This guide covers the history of facial averaging, the best-known studies, the corrections that came after, and what an average face can and cannot tell you about your own.
How facial averaging works
Facial averaging works by lining up the same landmarks on each face, such as the corners of the eyes and the edges of the mouth, and then averaging the shape and the pixel values. The result has the typical proportions of the group and almost none of its individual quirks. Blemishes, scars and unusual features shrink as more faces are added, because few faces share them.
The tiny drawing below shows the shape side of it. The pale outlines stand for different faces, and the highlighted outline is their mean. Facial averaging smooths the extremes, so a very wide face and a very narrow one end up near the middle.
Galton and the first composite portraits
Francis Galton described the method in 1878, in a paper on composite portraits. He exposed photographs of different people on the same plate, so each contributed a little of the final image. He used it to look for a common type within groups, and he noted that the blended portraits looked smoother and better than the individuals. His reasons for the work were tied to the ideas of his era about heredity, and those ideas have aged badly. The technique, however, was the first form of facial averaging.
Langlois and Roggman, 1990
The study most people mean is "Attractive faces are only average," by Judith Langlois and Lori Roggman, published in Psychological Science in 1990. They digitized photographs of faces, averaged them in groups, and asked people to rate both the composites and the original faces. Composites were rated as more attractive than most of the single faces, and ratings rose as more faces went into the blend. Their reading was that people prefer faces close to the population mean, and that facial averaging produces such faces.
The study mattered because it gave facial averaging a testable claim, and because it fit work on how people like typical examples of other categories. It did not say that an average face is the most attractive possible face.
Corrections: very attractive faces are not average
Quick criticism followed. In 1991, Alley and Cunningham argued in Psychological Science that averaged faces are attractive, but that very attractive faces are not average. In 1994, Langlois, Roggman and Musselman replied in the same journal that none of the alternative explanations offered so far, such as the smoother look of blended images, fully accounted for their results. The same year, Perrett, May and Yoshikawa (Nature, 1994) found that a blend of faces rated attractive was preferred over the blend of all faces, and that pushing the difference further raised the ratings again, so the mean is not the peak. These results say that facial averaging describes a floor, not a ceiling.
A second correction concerns what facial averaging does to a face. Averaging smooths skin and tends to make faces more symmetrical, which both raise attractiveness on their own. Rhodes, Sumich and Byatt (Psychological Science, 1999) looked at this directly and concluded that facial averaging still had an effect after symmetry was controlled. Komori and colleagues (Acta Psychologica, 2009) set averageness against symmetry in a single study. Taken together, the studies suggest that facial averaging, smooth skin and symmetry are tangled together, and that no one of them explains the result alone.
Why would facial averaging make a face look good?
Researchers have proposed several explanations. One is evolutionary: a face near the mean might signal healthy development. Another is mental: Winkielman and colleagues (Psychological Science, 2006) argued that prototypes are attractive partly because they are easy for the mind to process. Gillian Rhodes (Annual Review of Psychology, 2006) and Little, Jones and DeBruine (Philosophical Transactions of the Royal Society B, 2011) summarize the evidence: facial averaging is one factor among several, with no single agreed cause. A statistical model by Said and Todorov (Psychological Science, 2011) points the same way: highly attractive faces depart from the mean along the dimensions that drive the ratings.
What facial averaging cannot say about you
An average face is a statistical object that depends on which faces went in: which country, age, and sex, and how many. There is no universal mean. Facial averaging also removes character: a face is judged in life by expression, movement, voice and context, and none of those survives blending.
A rating of your own photo is therefore not a distance from the mean. Measures such as symmetrical face scores and divine proportion face ratios describe a picture, and neither proves how anyone will react to you. The face proportions tool shows several ratios side by side. Where attractiveness research meets looksmaxxing, the psl scale explains how forums turn opinion into numbers. A PSL rating gives written notes on one photo, and it is not a verdict.
If comparing your face with composites makes you feel worse, stop. Studies of faces and attractiveness say little about your value, your health or how you will be treated by people who know you.
Facial averaging FAQ
What is facial averaging?
It is the process of blending several faces into one by averaging their shape and image values. The composite shows the typical features of the group.
Are average faces the most attractive?
Not exactly. Composites are rated as more attractive than most individual faces, but later studies found that blends of attractive faces beat the average, so the mean is not the peak.
Who invented facial averaging?
Francis Galton, who described blended photographs in 1878.
Does symmetry explain the facial averaging effect?
Partly. Symmetry and smooth skin raise ratings, and some studies found a separate effect of averageness after controlling for symmetry.
Does facial averaging make skin look better?
Yes, in composites. Blending smooths blemishes and texture, which raises ratings by itself.
Does this page use real faces?
No. The drawing is a simple outline diagram, not a photograph.