Evidence review

The golden ratio and facial attractiveness: what the evidence shows

Studies that tested 1.618 against rated attractiveness, the proportions they found instead and how often real faces land on the classical targets.

Published
Last reviewed
Version
1.1
Reading time
10 min
Author
LooksMeter Research

Key findings

  1. 01Two studies measured real faces against rated attractiveness. Closeness to 1.618 explained little or none of the difference between faces.1, 2
  2. 02In three-dimensional scans of young adults, 7 of 10 facial ratios were statistically different from 1.618.2
  3. 03Where an experiment did find preferred proportions, they matched an average face and were not derived from 1.618.3
  4. 04Classical proportion rules fitted a minority of the people measured in every group studied.4, 5, 6
  5. 05Across 1,363 faces measured by LooksMeter, 5.9% had a mouth to nose width ratio within 10% of 1.618.
  6. 06The better supported cues to facial attractiveness are averageness, symmetry and sexual dimorphism. None of them is a single number.7, 8

Section 1

The claim

The golden ratio is the number 1.618, often written as phi. The claim made for it is simple. A face is said to be more attractive the closer its proportions sit to 1.618.

The claim reaches people in two forms. The first is a list of ratios between facial distances, each compared with 1.618. One orthodontic study tested 19 such proportions taken from a published scheme.1 The second is a single template drawn from the ratio and laid over a face. The best known is the Phi mask.9

Both forms make a prediction that can be checked. Measure a set of faces. Have people rate them. See whether distance from 1.618 tracks the ratings. This review covers the studies that did that, the proportions that did predict preference and how often real faces reach the classical targets at all.

Table 1The claims made for the golden ratio and what the studies report.
ClaimWhat the studies reportEvidence
Faces closer to 1.618 are rated as more attractiveNot in the two direct tests. One found 4 of 19 proportions weakly related to ratings. The other found no relation in 9 of 10 ratios.1, 2Not supported
The proportions of real faces equal 1.618Mostly not. 7 of 10 measured ratios were statistically different from 1.618.2Not supported
Classical proportion rules describe a typical faceNo. In each group measured, the best fitting rule held for between 26.7% and 40% of people. Most rules held for fewer.4, 5, 6Not supported
One ideal template fits every faceA published analysis found the best known template a poor fit for non-European faces. Survey data show that facial measurements differ between groups.9, 10Not supported
Some facial proportions do affect preferenceYes. In one set of experiments preference peaked when the eyes to mouth distance was about 36% of face length and the distance between the eyes about 46% of face width.3Limited
Average proportions are attractiveYes. Composite faces were judged more attractive than almost all of the faces that went into them. A later review with meta-analyses agrees.7, 11Consistent

Consistent: several independent studies point the same way. Mixed: studies point in different directions. Limited: one or two studies, not yet repeated. Not supported: tested, with no effect found or one too small to matter.

Section 2

Direct tests against rated attractiveness

Nineteen proportions in adolescents

Kiekens and colleagues showed photographs of 64 white adolescents to 76 adult laypeople, who rated each face on a scale from 0 to 100.1 Three observers placed 13 landmarks on every face. From those the team calculated 19 proportions that had been proposed as golden and how far each one sat from 1.618.

Four of the 19 proportions were related to the ratings in the predicted direction. Faces rated higher sat a little closer to 1.618 on those four. The other 15 were not. Taken together the four explained 16% of the variance in ratings. The authors concluded that this is too small to be of clinical importance.

Ten ratios in three-dimensional scans

Rossetti and colleagues used three-dimensional stereophotogrammetry, which the authors describe as the reference method in facial anthropometry.2 A jury scored 400 healthy young adults for attractiveness on a scale from 0 to 40. From those who scored above 28 and those who scored below 12 the team drew 15 men and 15 women per group at random, 60 people in all. Ten ratios were computed for each person, compared between the groups and tested against 1.618.

Nine of the ten ratios showed no significant effect of attractiveness. The one exception was the eyes to mouth distance over the height of the lower jaw, which differed by both sex and attractiveness. Seven of the ten ratios were statistically different from 1.618. The authors concluded that the ratios were not related to attractiveness and that most of them were different from the golden ratio.

How far these two studies go

These are small studies. One rated 64 adolescents and the other compared 60 young adults. Each covered one age group. They are still the direct test of the claim and they point the same way. A relation between 1.618 and rated attractiveness was either absent or too weak to be useful.

Section 3

What predicted preference instead

If 1.618 does not predict preference, the next question is whether any proportion does. One set of experiments says yes.

Pallett and colleagues ran four experiments in which people compared pairs of faces.3 The features in each pair were identical. Only the distance between the eyes and the mouth or the distance between the two eyes was changed. A face was judged most attractive when the eyes to mouth distance was about 36% of the length of the face and the distance between the eyes about 46% of its width.

The authors called these new golden ratios. They were found by experiment and not derived from 1.618. The authors also report that they match the proportions of an average face.

That last point connects to a long-standing finding. Langlois and Roggman digitised photographs of faces and averaged them by computer.11 The composite faces were judged more attractive than almost all of the individual faces they were made from. Composites became more attractive as more faces were added.

A later review with meta-analyses reached the same conclusion for three cues: averageness, symmetry and sexual dimorphism. All three were attractive in male and female faces and across cultures.7 A review written for surgeons names averageness, sexual dimorphism, youthfulness and symmetry as the four most important cues.8

None of these is a number. A face whose layout is near the average of its population tends to be preferred. That is a statement about a distribution of real faces. It is not a statement about a constant from geometry.

Section 4

How often real faces reach the classical targets

Golden ratio calculators usually mix two things. One is 1.618 itself. The other is a set of older rules known as the neoclassical canons, such as equal facial thirds and eyes set one eye width apart. Both can be checked against measured faces.

Published anthropometry

Farkas and colleagues tested nine neoclassical canons on 153 young adult North American Caucasians.4 The two rules that held most often were both horizontal. The distance between the eyes equalled nose width in 40% of people. Nose width equalled a quarter of face width in 37%. The vertical rules fitted worst. The authors wrote that the canons do not represent average facial proportions.

The same test has been run in other groups. Among Chinese, Vietnamese and Thai adults the three sections of the facial profile were not equal. The same was true of the North American Caucasian comparison group.5 The five other canons tested held for 16.7% to 36.7% of the Caucasian group and for 1.7% to 26.7% of the Asian groups.

In 108 African American women the best fitting canon included 30.6% of the women. The next two included 13.0% and 9.3%.6 The authors concluded that the average African American woman does not fit the neoclassical standard.

An international survey of 1,470 healthy adults found that facial measurements differ between groups, most of all around the eyes and the nose.10 No single set of proportions describes all of the groups it measured.

LooksMeter measurements

We ran the same kind of check on our own data. The figures below come from the 1,404 faces in our facial measurement norms study. They show how many faces sit near a classical target. They cannot show whether sitting near it matters, because the dataset holds no attractiveness ratings.

Table 2Share of faces near each classical target. LooksMeter norms dataset, version 1.0.
MeasurementClassical targetMedian measuredFaces within 10% of target
Mouth width to nose width1.6181.285.9%
Lower lip to upper lip1.6181.6347.5%
Face length to width *1.6181.210%
Eye spacing to eye width1.01.238.2%
Nose width to eye spacing1.01.1337.6%
Eye width to face width0.200.20883.2%

n = 1,363 faces. * Face length is measured from the top of the tracked face area, which sits at or just below the hairline. The ratio therefore reads lower than a hairline to chin measurement would. This row is not a fair test of the classical figure and is shown for completeness.

Two results stand out. The lip ratio has its median at 1.63, almost exactly on 1.618. Yet only 24.5% of faces fall within 5% of the target, because the ratio varies so widely from person to person. The mouth to nose width ratio sits far from 1.618. Its median is 1.28 and 2.2% of faces come within 5%.

One median near 1.618 is not evidence for the claim. With many ratios measured, some will land near any given number. The three-dimensional study above saw the same pattern. Three of its ten ratios were similar to 1.618 and seven were not.2

The eye spacing rule shows how far a classical target can sit from the typical face. The rule of fifths puts the eyes one eye width apart, a ratio of 1.0. The median face here measures 1.23 and 8.2% of faces come within 10% of 1.0.

A related rule sets eye spacing equal to nose width. The 1985 anthropometric study found that it held in 40% of its sample.4 Our share within 10% for that rule is 37.6%. The two figures use different tolerances and different methods, so the agreement should not be over-read.

Section 5

The problem with one template

A template makes a stronger claim than a list of ratios. It says there is one ideal layout for every face.

Holland examined the best known template, a mask derived from the golden ratio.9 The paper reports three problems. The method that had been used to judge how well a face fits the mask was faulty. The mask suits non-European populations poorly, in particular sub-Saharan African and East Asian faces. Its proportions come closest to those of masculinised white women as seen in fashion models.

The paper concludes that the mask does not describe an ideal face shape even for white women, because its proportions are at odds with what most people prefer.

This is one analysis by one author and it has the limits of any single paper. It is consistent with the survey data above. Faces differ between populations in measurable ways, so a single template will fit some groups better than others.10

Section 6

What this means for a golden ratio score

LooksMeter offers a Golden Ratio Face Test. It measures the proportions discussed here and shows how each compares with its classical target. This review sets the limits on what that result can mean.

  • It is a description of proportion. It shows where a face sits relative to a set of classical reference figures.
  • It is not a measure of attractiveness. The studies above do not support reading distance from 1.618 as a beauty score.
  • A low match is the normal result. Most measured faces sit away from most classical targets.
  • The measured spread of real faces is a better yardstick than a constant. Our norms study gives that spread for 14 measurements.

Researchers have built an attractiveness index that combines neoclassical canons, symmetry and golden ratios.12 Any score built this way is a summary of proportion. It inherits the limits of the evidence behind each of its parts. That applies to our score as well.

Section 7

Limits of this review

  • This is a narrative review of selected peer-reviewed studies. It is not a systematic review and it may have missed relevant work.
  • The two direct tests are small and each covers one age group.
  • Attractiveness ratings depend on who is rating. The studies used different panels and different scales.
  • The LooksMeter figures come from self-selected users and single photos. They carry no ratings and no demographic split. The norms study sets out those limits in full.
  • Each study is summarised from its published abstract. Where a paper reports more detail than its abstract, we have not relied on it here.

References

  1. 1Kiekens RM, Kuijpers-Jagtman AM, van 't Hof MA, van 't Hof BE, Maltha JC. Putative golden proportions as predictors of facial esthetics in adolescents. American Journal of Orthodontics and Dentofacial Orthopedics. 2008;134(4):480-483. doi:10.1016/j.ajodo.2006.10.041 · PubMed
  2. 2Rossetti A, De Menezes M, Rosati R, Ferrario VF, Sforza C. The role of the golden proportion in the evaluation of facial esthetics. The Angle Orthodontist. 2013;83(5):801-808. doi:10.2319/111812-883.1 · PubMed
  3. 3Pallett PM, Link S, Lee K. New "golden" ratios for facial beauty. Vision Research. 2010;50(2):149-154. doi:10.1016/j.visres.2009.11.003 · PubMed
  4. 4Farkas 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
  5. 5Le 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
  6. 6Porter JP, Olson KL. Anthropometric facial analysis of the African American woman. Archives of Facial Plastic Surgery. 2001;3(3):191-197. doi:10.1001/archfaci.3.3.191 · PubMed
  7. 7Rhodes G. The evolutionary psychology of facial beauty. Annual Review of Psychology. 2006;57(1):199-226. doi:10.1146/annurev.psych.57.102904.190208 · PubMed
  8. 8Bashour M. History and current concepts in the analysis of facial attractiveness. Plastic and Reconstructive Surgery. 2006;118(3):741-756. doi:10.1097/01.prs.0000233051.61512.65 · PubMed
  9. 9Holland E. Marquardt's Phi mask: pitfalls of relying on fashion models and the golden ratio to describe a beautiful face. Aesthetic Plastic Surgery. 2008;32(2):200-208. doi:10.1007/s00266-007-9080-z · PubMed
  10. 10Farkas 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
  11. 11Langlois JH, Roggman LA. Attractive faces are only average. Psychological Science. 1990;1(2):115-121. doi:10.1111/j.1467-9280.1990.tb00079.x
  12. 12Schmid K, Marx D, Samal A. Computation of a face attractiveness index based on neoclassical canons, symmetry, and golden ratios. Pattern Recognition. 2008;41(8):2710-2717. doi:10.1016/j.patcog.2007.11.022

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. The golden ratio and facial attractiveness: what the evidence shows. Version 1.1. 18 November 2026. https://www.looksmeter.com/research/golden-ratio-evidence

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