How facial thirds and the facial width-to-height ratio (fWHR) are measured, what the research actually says about fWHR, why the golden ratio doesn’t hold up, and how our app uses each.
Published Updated
Short answer
Facial thirds compare the heights of the upper, middle and lower face; fWHR divides the face’s width at the cheekbones by the height from the brows to the upper lip. Our app scores the middle-to-lower thirds balance (typical range 0.89 to 1.21) and shows fWHR for interest only (typical range 1.61 to 1.99, our own estimate). Neither has a single "ideal", and the golden ratio doesn’t describe real attractive faces well.
Facial thirds
The classic rule splits the face into three heights: hairline to brows (upper), brows to the base of the nose (middle), and nose to chin (lower). Real faces vary around equal thirds, and the hairline moves with age and hair loss, so the upper third is the least reliable. The app compares only the lower third with the middle, measured on the mesh.
A longer lower third often reads as a strong chin; a shorter one as a softer lower face. Neither is wrong, and a typical range wide enough to cover many populations is part of the scoring rule, not an afterthought.
fWHR: the facial width-to-height ratio
fWHR is the width across the cheekbones (bizygomatic width) divided by the height from the brows to the upper lip. It became famous through studies linking wider faces to how dominant or aggressive men are judged to be.
A 2015 meta-analysis pooled those studies (Geniole and colleagues). It found fWHR only slightly larger in men than in women, linked to how masculine a man’s face is judged, related to threat behaviour in men, and also related to body mass index, so weight moves it. That is about perceived threat and dominance, not attractiveness. So the app shows your fWHR for interest and never scores it.
What about the golden ratio?
The "phi mask" claimed that beautiful faces follow the golden ratio (about 1.618). An analysis of it found that it best describes a narrow look seen in fashion models, fits non-European faces poorly, and doesn’t describe an ideal face even for the group it was drawn from (Holland, 2008). We don’t score anything against phi.
A note on symmetry
Every face is a little asymmetric: a study of 321 young, healthy adults measured asymmetry between the two halves of the face across the group (Ercan and colleagues, 2008). The app reflects one side of your mesh onto the other and measures the gap; a turned head makes a symmetric face look uneven, so it checks the pose first.
How the app handles this
Harmony is scored on facial thirds (lower face against mid-face, since a photo can’t find your hairline) and lower-face balance (chin height against upper-lip height). fWHR, the mid-face ratio and the forehead ratio are shown beside their typical ranges and never scored. Ranges are widened on purpose so proportions common in some families or backgrounds are not marked down.
Questions people ask
What is a good fWHR?+
There isn’t a "good" one. Research ties fWHR to how threatening a face looks, not to attractiveness. The app shows yours for interest and doesn’t score it.
Why doesn’t the app measure the upper third?+
The face mesh stops on the upper forehead and can’t find the hairline, which also moves with age and hair loss. So the thirds score compares the lower face with the mid-face.
Is the golden ratio real?+
Phi is real maths, but the claim that attractive faces follow it doesn’t hold up across people. We don’t use it.