Masculine and feminine face features
What actually differs
Most lists of masculine and feminine features come from impression: somebody looked at a lot of faces and wrote down what they noticed. This is the list Canthion measures against instead. Every entry on it has a direction, a size, and a source for the number.
Also searched as facial masculinity, facial femininity, masculine face features, feminine face features and how masculine is my face.
The scale
Every trait sits on the same line. Fifty is the midpoint of the reference population you pick, and fifty points is one pooled standard deviation in the masculine direction. So a face a full standard deviation toward the female mean sits at 0, and a full standard deviation the other way sits at 100.
The bands: under 15 very feminine, under 40 feminine, 40 to 60 balanced, 60 to 85 masculine, above 85 very masculine. A typical man lands in the masculine band and a typical woman in the feminine one. That is the check that the scale is calibrated rather than flattering.
The traits, and how far apart the sexes sit on each
Fourteen traits, ordered by effect size. Cohen's d is the distance between the male and female means, counted in pooled standard deviations. It doubles as a measure of overlap: the smaller the d, the more the two distributions sit on top of each other.
| Trait | Direction | What it looks like | Cohen's d | Source |
|---|---|---|---|---|
| Brow height above the eye | Higher in women | A low-set brow close to the eye | 0.9 large | Whitehouse 2015; Burton 1993 — brow height is the strongest single photographic sex cue |
| Jaw width | Higher in men | A jaw carried wide relative to the cheekbones | 0.8 large | Farkas 1994; Perrett 1998 |
| Nose width | Higher in men | A wider nose base | 0.7 medium | Farkas 1994 (nose width d≈1.0 absolute) |
| Lip fullness | Higher in women | Thinner lips for the face | 0.7 medium | Farkas 1994; Bisson & Grobbelaar 2004 |
| Nasal index | Higher in men | A broader nose | 0.6 medium | Farkas 1994 |
| Chin share of the lower face | Higher in men | A tall chin for the lower face | 0.6 medium | Farkas 1994; Perrett 1998 |
| Face width to height ratio | Higher in men | A wider face for its height | 0.5 medium | Weston 2007; Lefevre 2013 (fWHR, contested) |
| Eye openness | Higher in women | A narrower eye opening | 0.5 medium | Burton 1993; Farkas 1994 |
| Philtrum length | Higher in women | A bigger chin for the philtrum | 0.5 medium | Perrett 1998; Farkas 1994 |
| Lip ratio | Higher in women | A thinner upper lip relative to the lower | 0.4 small | Farkas 1994; Popenko 2017 |
| Canthal tilt | Higher in women | A flatter outer-eye line | 0.3 small | craft/morphometric; small |
| Brow arch position | Higher in women | A flatter, less arched brow | 0.3 small | craft; small |
| Mouth width | Higher in women | A narrower mouth for the nose | 0.3 small | Farkas 1994 (nose grows more than mouth in men) |
| Facial index | Higher in men | A longer face for its width | 0.3 small | Farkas 1994; small |
4 of the fourteen reach 0.7 or above: brow height above the eye, jaw width, nose width, lip fullness. The rest sit between 0.3 and 0.6. At that size the male and female distributions overlap so heavily that any one trait, taken on its own, separates almost nothing. So the position is read across all fourteen at once, rather than by picking a feature and calling it masculine.
Three of the fourteen are measured straight from the mesh rather than from a published ratio table: brow height above the eye, lip fullness relative to the face, and the chin's share of the lower face. They are among the strongest photographic sex cues, and no standard ratio table covers them.
What the preference evidence says
A masculine reading is not a score to raise. Neither is a feminine one. What the rating literature actually reports is narrower than either: masculine jaw and brow rate well in men, feminine eyes, lips and skin rate well in both sexes, and the preferred face is not at the extreme of either end.
Then there is how much of this is personal. Roughly 40% of how any one viewer rates a face is idiosyncratic to that viewer (Leder et al. 2016; Germine et al. 2015). That is a hard ceiling on what a measurement of this kind can predict about one particular person's reaction.
This is not a measure of hormone level, of bone thickness under soft tissue, or of gender. It is one thing only: where a photograph of your face sits against published means for the population you selected, carrying the interval that landmark noise leaves behind. Nothing here is medical. Nothing here is a recommendation to change anything.
Where each of these is measured
Eleven of the fourteen have their own measurement page, with the formula, the reference tables and the error bar:
The masculinity and femininity axis is part of the complete report, built from these fourteen traits on your own photograph. The free overview covers the proportions themselves. Your photo is measured in the browser and never uploaded.
Common questions
What actually makes a face look masculine or feminine?
Four traits do most of the work: brow height above the eye, jaw width, nose width, lip fullness. Those are the biggest separations among the traits with published sex differences. Brow height is the largest single one. A brow sitting low and close to the eye reads male, a brow set high above it reads female, and that difference holds up in a photograph as well as it does in person.
Can a photograph tell me how masculine my face is?
It can place your proportions on a scale built from the sex differences the literature reports. That is not the same thing as measuring your bone or your hormones. Landmark noise puts an error bar on every value, and the position inherits it. What comes back is a description of a photograph, not a verdict on a person.
Is a more masculine face more attractive in men?
Up to a point, and not at the extreme. The rating literature finds masculine jaw and brow rating well in men, and feminine eyes, lips and skin rating well in both sexes. Toward the ends of the scale, preference falls away rather than rising (Perrett 1998; Rhodes 2006; Said and Todorov 2011). And roughly 40% of how any one person rates a face is idiosyncratic to that person, so no measurement predicts a particular reaction.
What is Cohen's d and why is it in the table?
It is how far apart two group means sit, counted in pooled standard deviations. Around 0.2 is a small difference, 0.5 a medium one and 0.8 a large one. It is in the table because size is the part feature lists leave out. A trait with a d of 0.3 overlaps so heavily between men and women that on its own it tells you close to nothing.
Which of these can be changed without surgery?
The soft-tissue reads move. Brow shape and height respond to grooming, and they carry the largest effect in the table. Facial hair changes the outline of the jaw and the lower face in a photograph. Body composition changes the fullness of the cheek and the definition along the jaw. The underlying bone does not move, and nothing on this site suggests it does.
Where does the scale come from?
Fifty is the midpoint of the reference population you choose. Fifty points is one pooled standard deviation toward the male mean, so a full standard deviation feminine is 0 and a full standard deviation masculine is 100. Where the reference tables genuinely separate the sexes, the published means are marked on the bar. Where they barely separate them, the marks come from the effect size for that trait.
References
- Farkas 1994 (Anthropometry of the Head and Face) — sex differences in nose, jaw, lip and brow dimensions
- Whitehouse et al. 2015; Burton, Bruce & Dench 1993 — brow height/eye size are the strongest photographic sex cues
- Perrett et al. 1998 Nature; Rhodes 2006; Said & Todorov 2011 — masculine jaw/brow rate well in men, the extreme does not
Related: face shape · the golden ratio face test · looksmaxxing, measured · attractiveness test · methodology · glossary · all 35 insight pages