Insights · Whole face · Measurement

Eye–mouth distance : face length (Pallett)

The only empirically derived "midface ratio": Pallett, Link & Lee (2010) found raters preferred an eye-to-mouth distance of ≈0.36 of face length (hairline to chin).

Confidence
Tier B · wide interval
Unit
ratio (dimensionless)
Landmarks
iris_R, iris_L, sto, tr, me
CI floor
± 0.02
dist(eyeLine, sto) / dist(tr, me)

Depends on the user-placed hairline (trichion) — no detector locates it and it moves with hairstyle.

Measure yours
Illustration · on your scan the same lines are drawn on your own photo

How to read it

Below the mean

the eye-to-mouth distance is short relative to face length (a compact midface).

Near the mean

this proportion sits near the value Pallett et al. identified as the preferred/average configuration.

Above the mean

the eye-to-mouth distance is long relative to face length (a long midface). Note the value depends on the user-placed hairline.

Caveat

Every consumer app computes a different thing and calls it "midface ratio". This is the one definition with a citation, and it was derived on one population and one sex.

Reference values by population

Reference populationFemale (mean ± SD)Male (mean ± SD)nSource
North American White0.36 ± 0.030.36 ± 0.03Pallett et al. 2010 — preferred value on Caucasian female faces (n=4 experiments); treated as the population-average configuration. Provisional for other groups. prov.
East Asian0.36 ± 0.030.36 ± 0.03Pallett et al. 2010 — preferred value on Caucasian female faces (n=4 experiments); treated as the population-average configuration. Provisional for other groups. prov.
South Asian0.36 ± 0.030.36 ± 0.03Pallett et al. 2010 — preferred value on Caucasian female faces (n=4 experiments); treated as the population-average configuration. Provisional for other groups. prov.
African / African diaspora0.36 ± 0.030.36 ± 0.03Pallett et al. 2010 — preferred value on Caucasian female faces (n=4 experiments); treated as the population-average configuration. Provisional for other groups. prov.
Middle Eastern0.36 ± 0.030.36 ± 0.03Pallett et al. 2010 — preferred value on Caucasian female faces (n=4 experiments); treated as the population-average configuration. Provisional for other groups. prov.
Hispanic / Latino0.36 ± 0.030.36 ± 0.03Pallett et al. 2010 — preferred value on Caucasian female faces (n=4 experiments); treated as the population-average configuration. Provisional for other groups. prov.
Not specified — show pooled0.36 ± 0.030.36 ± 0.03Pallett et al. 2010 — preferred value on Caucasian female faces (n=4 experiments); treated as the population-average configuration. Provisional for other groups. prov.
Where these numbers come from

Transcribed from the cited anthropometric literature; SDs of ratios are propagated conservatively (independence assumed). Entries marked prov. await line-by-line audit against the primary tables — see the norm provenance note.

Why every value has an error bar

Every value Canthion reports carries a 1σ uncertainty propagated from MediaPipe Face Mesh's published accuracy (3.96% of inter-ocular distance mean absolute error per landmark, Kartynnik 2019). For a typical 300 px inter-pupil photo that is ≈12 px per landmark, which is why small ratios and angles get wide intervals. A number without an interval is a number pretending to be more precise than it is.

References

Related: how bone-level measurements are taken from a photo · methodology · all insights