Facial averageness: why "average" faces are rated attractive

The averageness effect is the well-replicated finding that a composite face — made by mathematically blending the shapes of many individual faces — is reliably rated more attractive than most of the individual faces used to create it. Here, "average" means mathematically averaged, not plain or ordinary.

What the averageness effect is

Take photographs of many different faces, align them on shared landmarks, and average their shapes and textures together into a single composite. The result is a face that looks like no one in particular but is judged, again and again across studies, as more attractive than the typical individual face that fed into it. Add more faces to the blend and the composite usually becomes still more attractive, up to a point. This is the averageness effect, and it is one of the most robust findings in the science of facial attractiveness.

The word "average" is doing precise mathematical work here and it routinely trips people up. It does not mean mediocre, forgettable, or middling on a scale of beauty. It means the arithmetic mean of many faces' geometry. A mathematically average face is unusually smooth, unusually symmetric, and unusually free of the small idiosyncrasies that individual faces carry — which is exactly why it tends to score highly rather than low.

The classic composite-photography history

The effect has a surprisingly old origin. In the 1870s and 1880s, the Victorian polymath Francis Galton created "composite portraits" by exposing multiple photographic negatives of different people onto a single plate, each for a fraction of the total exposure. He was trying to find the "typical" face of a category — criminals, or people with a particular illness — expecting the blend to reveal a characteristic type. Instead, observers noticed something Galton hadn't set out to find: the composites often looked better, more regular and more pleasing, than the individual faces that composed them.

That accidental observation sat mostly dormant until the late twentieth century, when researchers with digital morphing tools revisited it systematically. Landmark studies in the early 1990s digitally averaged faces and confirmed with controlled ratings what Galton had glimpsed: composites are rated as more attractive than their constituents, and the effect strengthens as more faces are blended in. Later work refined the picture — showing, for instance, that while averageness helps, the single most attractive face is usually not the maximally average one.

The leading explanations

Why should a blend of many faces be attractive? There are three main, non-exclusive accounts, and the honest position is that they probably work together rather than one being the sole cause.

Koinophilia

Koinophilia is the idea, drawn from evolutionary biology, that organisms may have evolved to prefer common, typical traits over unusual ones, because extreme deviations from the population average are more likely to reflect harmful mutations or developmental problems. On this view, an average face signals a lack of unusual, potentially disadvantageous features, and preferring it was historically a reasonable bet about health and genetic fitness. It's a plausible ultimate explanation, though like all evolutionary-psychology accounts it is difficult to test directly.

Processing fluency

The fluency account is cognitive rather than evolutionary. The brain builds a mental prototype of "a face" from all the faces it has ever seen, and images closer to that prototype are easier and faster to process. That ease of processing is itself experienced as mildly pleasant, and we tend to attribute the pleasant feeling to the face rather than to our own smooth processing of it. An average face is, almost by definition, close to the prototype, so it processes fluently and reads as attractive. This account elegantly explains why averageness is appealing even without invoking mate choice.

Correlation with symmetry

Averaging many faces doesn't only regularize shape — it also increases symmetry, because left-right idiosyncrasies in individual faces cancel out in the blend. Since symmetry is independently associated with attractiveness ratings (see our detailed guide to facial symmetry), part of the averageness effect may simply be symmetry riding along. Blending also smooths skin texture, removing blemishes and evening tone, which further lifts ratings. Disentangling how much of the effect is "averageness itself" versus these correlated byproducts is an active methodological question.

The limits of the effect

Averageness is real and robust, but a few caveats keep it from being the whole theory of facial attractiveness:

• Average is not the maximum. Studies that let people tune a face away from the average find that certain non-average, often youthful or sex-typical features can raise ratings beyond the pure average. The most attractive face is highly average, but not perfectly average.
• Composites are confounded. Blending improves symmetry and skin texture at the same time as shape, so a composite's appeal isn't attributable to shape-averageness alone.
• Context and culture modulate ratings. Preferences shift with familiarity, culture, and even the set of faces a viewer has recently seen, so no composite is universally "the" attractive face.
• It's about perception, not worth. The effect describes average ratings from groups of observers. It says nothing about any individual's value, and it is not a target to measure yourself against.

What it means for how we judge our own faces

The averageness literature carries a quietly reassuring message. Because composites blend away the very features that make a face distinctive, the "attractive average" is essentially a face with no strong particularities. Real, memorable, striking faces — including most faces people find genuinely captivating — are not maximally average; they carry distinctive traits. So the research absolutely does not imply that you should file down whatever makes your face specific in pursuit of an average ideal.

It also reframes self-criticism. When you fixate on a feature that deviates from "typical," you're often reacting to distinctiveness, not to a flaw. Distinctiveness reduces measured averageness but is also the raw material of a recognizable, individual face. Combined with the fact that your everyday view of yourself is distorted by mirrors and close selfies (our article on why faces look different in selfies covers the optics), the practical lesson is to distrust harsh self-judgments built on a narrow, distorted sample of your own face.

How averageness relates to AI face analysis

AI can measure how far a given feature sits from population norms, which sounds like it should produce an "averageness grade." Refrakt deliberately treats it differently. Averageness research is a finding about how groups of people rate composites; it is not a scoring rubric for an individual, and turning it into a personal ranking would misuse the science. What measurement can honestly offer is context: showing where a proportion falls relative to typical ranges, so a number becomes interpretable rather than mysterious.

• Norms as context, not verdicts. Refrakt's face analysis can show whether a measurement is typical or distinctive, without translating that into better or worse.
• Related but separate ideas. Averageness overlaps with symmetry and with proportion frameworks like the golden ratio, but each is its own measurement with its own caveats.
• Scores are tools, not judgments. Where a summary figure like a facial harmony score exists, it's a transparent aggregate you can trace back to individual measurements — not a claim about your worth.

The consistent Refrakt position is that the averageness effect is fascinating science about perception, and a poor instrument for grading yourself. Measure to understand; don't measure to rank.

Frequently asked questions