Data study

Facial measurement norms: percentiles for 14 facial ratios across 1,404 faces

The median and spread of canthal tilt, fWHR, midface ratio, eye spacing and ten more measurements, taken from 1,404 front-facing photos. With the share of faces that actually meet each commonly quoted target.

Published
Last reviewed
Version
1.2
Reading time
10 min
Author
LooksMeter Research

Key findings

  1. 01The median canthal tilt across 1,404 faces was +4.2°. 91.2% of faces had a positive tilt.
  2. 02The median facial width-to-height ratio was 1.63. 11.8% of faces fell inside the band of 1.8 to 2.1 that is often quoted for it.
  3. 03The median midface ratio was 1.00. 74.9% of faces were within 10% of 1.0.
  4. 04Eyes sat wider apart than the rule of fifths says. The median spacing was 1.23 eye widths and 8.2% of faces were within 10% of one eye width.
  5. 05Of three ratios with a golden ratio target, one had its median near 1.618: lower lip to upper lip at 1.63. Mouth width to nose width had a median of 1.28.
  6. 06Every figure is an aggregate. The dataset holds no photo, no account and no single person's measurement.

Section 1

What this document is

A facial measurement means little without something to compare it with. Most comparisons on offer are targets: a tilt that is said to be ideal, a ratio that is said to be perfect. This document gives the other kind of comparison. It reports how 14 facial measurements are actually distributed across 1,404 faces, so that a single reading can be placed among real ones.

For each measurement it gives the 5th, 25th, 50th, 75th and 95th percentiles, the mean and the standard deviation. It then reports how many faces meet each commonly quoted target.

LooksMeter also publishes live facial ratio statistics that update daily from recent scans. This document is different in kind. It is a fixed sample with a stated method, so that it can be cited and checked. Its figures and the live figures will differ slightly because they cover different scans.

Section 2

Sample

The sample is drawn from completed LooksMeter scans made between 23 July 2026 and 17 November 2026. Scans from 12 tools that store facial measurements contributed.

Table 1How the sample was built.
StepScans
Completed scans read3,250
With stored measurements3,016
Front-facing2,879
Removed: repeat scan from an account already counted1,450
Removed: measurements identical to a scan already counted25
Faces included1,404

Eight of the measurements were added to the scan record later than the first six. They are available for 1,363 of the 1,404 faces. Each table gives the count for every row.

The study uses no age, sex, ethnicity or country information. No figure here is split by any of them.

Section 3

Results

Figure 1Spread of 14 facial measurements across the sample. Each row has its own scale.
Canthal tilt
−1.5°+8.7°+4.2°
Eye aspect ratio
2.564.083.13
Eye spacing to eye width
1.081.431.23
Eye width to face width
0.1890.2280.208
Facial width-to-height ratio (fWHR)
1.441.881.63
Midface ratio
0.881.221.00
Face length to width
1.111.321.21
Nose width to eye spacing
0.981.281.13
Mouth width to nose width
1.141.471.28
Lower lip to upper lip
1.262.151.63
Chin to philtrum
1.362.832.01
Upper section
15.0%22.8%19.3%
Middle section
37.3%45.6%41.1%
Lower section
33.6%45.4%39.6%

Line: 5th to 95th percentile. Bar: 25th to 75th percentile. Tick and number below: median.

Eyes

Table 2Eye measurements.
Measurementn5th25thMedian75th95thMeanSD
Canthal tilt1,404−1.5°+2.4°+4.2°+6.2°+8.7°+4.1°3.0°
Eye aspect ratio1,4042.562.863.133.454.083.261.23
Eye spacing to eye width1,3631.081.161.231.311.431.240.11
Eye width to face width1,3630.1890.2000.2080.2160.2280.2080.012

The median canthal tilt is +4.2° and the middle half of faces fall between +2.4° and +6.2°. A positive tilt is the usual case: 91.2% of faces have one. The band of +2° to +8° that is often quoted for it holds 72.2% of all faces. It describes most people and not a rare trait. A tilt of zero or below is the less common case, at 8.8% of faces. The 5th percentile is −1.5°.

No single tilt is established as ideal. A photogrammetric study of composite faces found that the eye slant of attractive composites differed by group. It was less steep than average in the Asian composites and steeper than average in the Caucasian and African composites.1

The eye aspect ratio has a median of 3.13. Its mean of 3.26 sits above the median and its standard deviation is large beside the spread of the middle half. That pattern points to a small number of very high values, which is what a squint or a half-closed eye produces. The percentiles are the better guide for this measurement.

The two measurements of eye width tell a story together. One eye is close to a fifth of the width of the face: the median is 0.208 and 83.2% of faces fall between 0.18 and 0.22. The gap between the eyes is wider than one eye. Its median is 1.23 eye widths and 8.2% of faces come within 10% of 1.0. The rule of fifths holds for the eye and fails for the space between the eyes.

Face proportions

Table 3Proportions of the whole face.
Measurementn5th25thMedian75th95thMeanSD
Facial width-to-height ratio (fWHR)1,4041.441.541.631.731.881.640.15
Midface ratio1,4040.880.951.001.071.221.020.12
Face length to width1,3631.111.171.211.261.321.220.07

Face length is measured from the top of the tracked face area, which sits at or just below the hairline. See the note on vertical sections.

The median facial width-to-height ratio is 1.63 and the middle half of faces fall between 1.54 and 1.73. 11.8% fall inside the band of 1.8 to 2.1 that is often quoted for the ratio. This measurement depends on the camera. In one study the same people measured a smaller ratio in portraits taken with a shorter focal length.2 Camera distance was not controlled here, so our figures may read lower than figures taken from studio portraits. We have not tested that.

The median midface ratio is 1.00, which sits on the commonly quoted figure of about 1.0. 74.9% of faces are within 10% of it.

Face length to width has a median of 1.21. It should not be compared with the classical figure of 1.618, because the length used here stops short of the hairline.

Nose, mouth and chin

Table 4Measurements of the nose, mouth and chin.
Measurementn5th25thMedian75th95thMeanSD
Nose width to eye spacing1,3630.981.071.131.191.281.130.09
Mouth width to nose width1,3631.141.221.281.351.471.290.10
Lower lip to upper lip1,3631.261.461.631.812.151.650.28
Chin to philtrum1,4041.361.732.012.272.832.040.48

The nose is a little wider than the gap between the eyes. The median ratio is 1.13 and 37.6% of faces come within 10% of the classical figure of 1.0. A published anthropometric study found that this rule held in 40% of young adults.3 The two figures use different tolerances and methods, so the agreement should not be over-read.

Mouth width to nose width has a median of 1.28, well below the golden ratio target of 1.618. 5.9% of faces come within 10% of that target.

The lower lip to upper lip ratio has its median at 1.63, close to 1.618. It also varies widely. The 5th and 95th percentiles are 1.26 and 2.15. 24.5% of faces fall within 5% of 1.618 and 47.5% within 10%. Lip height changes with expression, so part of that spread is likely to come from the photo.

The chin to philtrum ratio has a median of 2.01 and the middle half of faces fall between 1.73 and 2.27.

Vertical sections

Table 5Vertical sections as a share of face height.
Measurementn5th25thMedian75th95thMeanSD
Upper section1,36315.0%17.8%19.3%20.9%22.8%19.2%2.4%
Middle section1,36337.3%39.6%41.1%42.8%45.6%41.3%2.5%
Lower section1,40433.6%37.2%39.6%41.8%45.4%39.6%3.7%

These are sections of the tracked face height. They are not the classical facial thirds. See the text below.

These three figures divide the height of the tracked face area. That area ends at or just below the hairline and not at the hairline itself. The upper section therefore reads short, with a median of 19.3%. The other two read long, with medians of 41.1% and 39.6%.

For that reason they must not be compared with the classical rule of equal thirds. No face in the sample has all three sections between 30% and 36%. That says where the measurement starts. It is not a finding about faces. Published anthropometry on the classical sections also finds that the three sections of the face are not equal.3, 4

Section 4

How many faces meet the commonly quoted targets

Guides and forums quote a target for most facial measurements. The table below sets each one against the sample. The tolerance is stated in every row because a share means nothing without one.

Table 6Share of faces inside each commonly quoted target.
TargetnShare of faces
Canthal tilt above 0°1,40491.2%
Canthal tilt from +2° to +8°1,40472.2%
fWHR from 1.8 to 2.11,40411.8%
Midface ratio within 10% of 1.01,40474.9%
Eye spacing within 10% of one eye width1,3638.2%
Eye width from 0.18 to 0.22 of face width1,36383.2%
Nose width within 10% of eye spacing1,36337.6%
Mouth to nose width within 5% of 1.6181,3632.2%
Mouth to nose width within 10% of 1.6181,3635.9%
Lip ratio within 5% of 1.6181,36324.5%
Lip ratio within 10% of 1.6181,36347.5%
Face length to width within 10% of 1.618 *1,3630%
Lower section from 30% to 36% *1,40416.2%
All three sections from 30% to 36% *1,3630%

* These rows depend on where face height starts. They are shown for completeness and are not a fair test of the classical figure.

Two patterns are worth drawing out. A target that most faces already meet, such as a positive canthal tilt, does not single anyone out. A target that almost no face meets, such as a mouth to nose width ratio of 1.618, is a poor description of faces. In both cases the share tells a reader how much weight the target deserves.

Our review of the golden ratio covers the studies that tested these targets against rated attractiveness.

Section 5

Method

Measurement

Each scan locates a set of landmarks on a single photo and computes angles and ratios between them. The computation runs in the browser at the time of the scan. Only angles, ratios and percentages are recorded. A single photo has no scale, so no distance in millimetres is taken. Our measurement standards define each measurement.

Inclusion

  • Completed scans up to the end of 17 November 2026 (UTC) on tools that store facial measurements.
  • Scans that carried a stored set of measurements.
  • Scans marked as front-facing at the time of the scan. A face that is clearly turned distorts every ratio.

Repeat scans

One scan was kept for each signed-in account. Scans made without an account cannot be linked to one another. They were removed only when their measurements were identical to those of a scan already counted. A person who scanned different photos without an account can therefore appear more than once.

Statistics

Percentiles were computed by linear interpolation between ranked values. The standard deviation is the sample standard deviation. Shares are the percentage of valid readings inside the stated range, with both ends of the range included.

Privacy

The figures were computed in one read-only pass over stored measurements. No photo was read. Only the aggregates in this document were saved. No account detail and no single person's measurement was kept or published.

Section 6

Limitations

  • The sample is self-selected. It is made up of people who chose to use an appearance analysis tool. It is not a sample of any population.
  • There is no demographic split. Published anthropometry shows that facial measurements differ between groups.5 These pooled figures are not norms for any one group.
  • Each face is one two-dimensional photo. Camera distance, lens, pose and expression were not controlled. Published work shows that camera distance changes facial proportions in a photo.2, 6
  • Repeat scans were removed by account only. People who scanned without an account may be counted more than once.
  • Face height starts at the top of the tracked face area and not at the hairline. The three vertical sections and face length to width are affected.
  • Landmarks are placed automatically and carry their own error. We have not yet published how much a reading changes between two photos of the same person.
  • This document has not been peer reviewed. Its figures are not clinical norms and must not be used for diagnosis or to plan treatment.

Section 7

Data and versions

The percentile table is available as a CSV file: LooksMeter facial measurement norms, version 1.2. It is free to cite with a link to this document.

Version 1.2 is fixed. A later sample will be published as a new version with its own date and sample size. The figures above will not be edited in place.

Table 7Version history.
VersionDateFacesChange
1.212 September 20261,404First publication

References

  1. 1Rhee SC, Woo KS, Kwon B. Biometric study of eyelid shape and dimensions of different races with references to beauty. Aesthetic Plastic Surgery. 2012;36(5):1236-1245. doi:10.1007/s00266-012-9937-7 · PubMed
  2. 2Třebický V, Fialová J, Kleisner K, Havlíček J. Focal length affects depicted shape and perception of facial images. PLOS ONE. 2016;11(2):e0149313. doi:10.1371/journal.pone.0149313 · PubMed
  3. 3Farkas LG, Hreczko TA, Kolar JC, Munro IR. Vertical and horizontal proportions of the face in young adult North American Caucasians: revision of neoclassical canons. Plastic and Reconstructive Surgery. 1985;75(3):328-337. doi:10.1097/00006534-198503000-00005 · PubMed
  4. 4Le TT, Farkas LG, Ngim RC, Levin LS, Forrest CR. Proportionality in Asian and North American Caucasian faces using neoclassical facial canons as criteria. Aesthetic Plastic Surgery. 2002;26(1):64-69. doi:10.1007/s00266-001-0033-7 · PubMed
  5. 5Farkas LG, Katic MJ, Forrest CR. International anthropometric study of facial morphology in various ethnic groups/races. Journal of Craniofacial Surgery. 2005;16(4):615-646. doi:10.1097/01.scs.0000171847.58031.9e · PubMed
  6. 6Pressler MP, Kislevitz ML, Davis JJ, Amirlak B. Size and perception of facial features with selfie photographs, and their implication in rhinoplasty and facial plastic surgery. Plastic and Reconstructive Surgery. 2022;149(4):859-867. doi:10.1097/prs.0000000000008961 · PubMed

Every reference was checked against Crossref before publication. Those indexed in PubMed were checked there as well. Findings are summarised in our own words. Read the original before relying on a number.

How to cite

LooksMeter Research. Facial measurement norms: percentiles for 14 facial ratios across 1,404 faces. Version 1.2. 18 November 2026. https://www.looksmeter.com/research/facial-measurement-norms

About this document

LooksMeter is a consumer appearance analysis product. This document is written by the LooksMeter team for general information. It has not been peer reviewed or reviewed by a medical professional. The measurements it describes are not clinically validated and are not a diagnosis or a medical device. See a qualified clinician about any health or surgical decision.

Corrections are welcome at [email protected]. Material changes are recorded by raising the version number above.

Continue