Insights / Measurement

← Back to insights

Measurement

Facial symmetry

What it measures and what it doesn't

Author
Santiago Grandas Forero MSc Psychological Research, University of Oxford
Reviewed by
Macken Murphy MSc Cognitive & Evolutionary Anthropology, University of Oxford
Published
Reading time
7 minutes

Abstract

Symmetry is the most-measured aesthetic signal because it’s the easiest to define mathematically: take landmark positions on one side of the face, mirror them, compare to the other side. The literature broadly supports the claim that more symmetric faces are perceived as more attractive — but the effect size is smaller than popular accounts suggest, and the metric has well-known limitations.

How we measure symmetry

VERASOMA uses normalised left-right deltas at standardised facial landmarks. Each landmark on the right is mirrored across the vertical midline; the distance from its mirror to the corresponding left landmark contributes to a symmetry score. The score is averaged across landmark groups (eyes, nose, mouth, jawline) so a localised asymmetry doesn’t dominate.

Why symmetry signals something

Developmental noise — minor disruptions during foetal and adolescent growth — produces small left-right asymmetries. The size of those asymmetries is a faint but measurable signal of developmental robustness. Observers appear to read symmetry implicitly as a quality cue, though they can’t usually articulate why.

Why symmetry is over-emphasised

Most faces are mildly asymmetric and most observers don’t notice. The effect size of symmetry on attractiveness ratings is real but modest (Cohen’s d typically 0.3-0.5). Headlines that present symmetry as the driver of attractiveness oversimplify a noisy literature.

Practical takeaways

A symmetry score in the 80-90 range is typical and unremarkable. Scores below 70 may reflect a localised asymmetry worth noting (and worth a clinician’s eye if dramatic). Scores in the 90s are notable but not particularly actionable.

FAQ

Are perfectly symmetric faces more attractive? Composites of perfectly symmetric faces look slightly off — they violate the small-asymmetry baseline observers expect. Real faces in the high-symmetry range tend to be most-preferred.

Can I improve my symmetry? The structural component is largely fixed; soft-tissue asymmetries (smile, brow position) are partly trainable.