Classical face proportions vs real faces

Contents

Classical face proportions: three kinds of claim What the ancient texts say: Vitruvius and Polykleitos The painters’ rules: measuring aids Measured: classical face proportions mostly fail How FacialRate does it: your proportions, then close museum works Next in this series FAQ

Key points

Three kinds of claim: an ancient text's rule, a drawing convention, a measured finding. Only the last measured people.
Vitruvius's face rules are simple fractions: 1/10 of the height, in three equal parts.
Polykleitos's Canon is lost; what survives covers the body, not the face.
In studies from 1985 and 2012, nobody had three equal thirds.

Classical face proportions: three kinds of claim

Most classical face proportions were written or drawn, not measured. Ask which kind a rule is:

An ancient text’s claim. Vitruvius, a Roman architect of the 1st century BCE. He doesn’t say he measured anyone.
A drawing convention. A painter’s measuring aid.
A measured finding. Real faces measured to see how often a rule holds.

Pull them apart, then measure one photo yourself.

What the ancient texts say: Vitruvius and Polykleitos

Vitruvius used the body as a model of proportion for temples:

The face, chin to hair roots, is 1/10 of the height; the head, chin to crown, 1/8.
The face splits into three equal parts: chin to the underside of the nostrils, nostrils to between the brows, brows to the hair roots.
With arms and legs spread, the body fits a circle centred on the navel, and a square, since height equals arm span.

None of it is 1.618.

The Greek sculptor Polykleitos (5th century BCE) wrote a treatise on proportion, the Canon, and made a statue to demonstrate it. Both the treatise and the bronze are lost; we have Roman marble copies. The surviving accounts describe body proportions and stance, with no rule for the face.

The painters’ rules: measuring aids

Cennini (about 1400, workshop manual): a third of the face is the unit for the whole figure; ear to ear is one face; the figure is eight faces plus two thirds.
Leonardo (notebooks): the space between the eyes is one eye’s width; pupil to pupil is 1/3 of the face width.
Dürer (1528): figures 7, 8, 9 and 10 heads tall, many types rather than one. He wrote that no single person is a model for everything, and that stout and thin figures differ in every proportion.

Today’s drawing guides are simplified heirs of this line; see neoclassical facial canons.

Measured: classical face proportions mostly fail

Facial thirds and fifths diagram: the thirds (four horizontal lines) and the five eye-widths (six vertical lines) on the sample photo. From top to bottom the horizontal lines are the hairline, the brow line, the base of the nose and the chin; the vertical lines split the face width into five at eye level.

1985: 153 young adults, 9 canons. The most often valid: eye gap equals nose width (40%) and nose width equals 1/4 of face width (37%). The vertical thirds fit worst.
2012: 103 people, 3D photos, valid within 1.5 mm. Equal thirds: 0%. Eye gap equals eye length: 0%; everyone’s gap was wider. Mouth equals 1.5 × nose width: 8.7%. Eye gap equals nose width: 19.4%.

Sample photo (above), thirds: 219 / 284 / 286 px, middle ÷ lower 0.99; the hairline is estimated, so the upper third is least reliable. Five eyes, left to right: 131 · 120 · 144 · 110 · 77. The gap is wider than an eye; the outer parts differ because the head is turned a little.

Measure your own: facial thirds. The 1.618 version: golden ratio face.

How FacialRate does it: your proportions, then close museum works

The score comes from a machine-learning model that learned from real human ratings, not a fixed formula like equal thirds, calibrated and tested against real ratings.
The free face report: face type, strongest feature, three highlights and 11+ dimensions. Free, no limit. Tips cover photography, hair, makeup and skincare only.
Artwork match: on your device, 478 face landmarks give 6 face proportions (eye shape, face shape, spacing of the features and so on). The photo never leaves the device. Each work’s face was measured the same way; only works that measure reliably are included. One work sits next to your photo, a few more equally close, in no order.
The works: public-domain (CC0 or public-domain mark) sculptures and portraits from open collections: the Met, the Smithsonian American Art Museum, the Rijksmuseum, the Getty and others. Works that caricature or mock appearance or age are excluded.
“Close” means facial proportions only, not looks, and has nothing to do with the score. The same photo always gets the same works.

It’s in the full report, an optional one-time purchase: title, artist, date, one sentence on where you’re close (no numbers, no ranking) and a comparison image. Unlocking sends one request with a few proportion readings; it isn’t stored. The sample photo’s closest work: Robert Vanderburgh McKim by William Henry Rinehart, about 1867, Smithsonian American Art Museum (CC0). See artwork match.

Next in this series

Greek sculpture faces: idealised, copied, once painted.
Renaissance head proportions: how painters measured a head.

FAQ

Which artwork do I look like? Can a face analyzer tell?

An art match compares 6 face proportions with museum sculptures and portraits. It doesn't say who you look like or measure beauty.

Is it normal for my facial thirds to be unequal?

Yes. In two measured studies nobody had three exactly equal thirds. Equal thirds are a drawing convention.

Greek sculpture faces: idealised types, not portraitsFaces in art Renaissance head proportions: how painters measured a head, and how you canFaces in art Neoclassical facial canons: where the face-proportion rules come fromFacial proportions What are facial thirds and fifths? Measure your own faceFacial proportions