Facial width-to-height ratio: what the dominance research actually found

Facial width-to-height ratio, usually shortened to fWHR, shows up in the same online spaces as canthal tilt: forums and short-form videos treating it as a settled predictor of dominance, testosterone, and attractiveness. The underlying research is real and peer-reviewed. It is also narrower and more contested than the online version suggests.

Definition

fWHR is the ratio of bizygomatic width (the distance between the two most prominent points of the cheekbones) to upper-face height (the distance from the highest point of the upper lip to the highest point of the eyelids, roughly brow level). A wider, shorter upper face produces a higher ratio. A narrower, taller upper face produces a lower one.

The measurement and the term were introduced together by Justin Carré and Cheryl McCormick in a 2008 study published in Proceedings of the Royal Society B, which measured fWHR from photographs of university and professional hockey players and compared it against penalty minutes as a proxy for aggression.

What the original study found

Carré and McCormick reported that fWHR predicted reactive aggression in men, accounting for roughly 15 percent of the variance in a laboratory aggression task, and correlated with penalty minutes in both varsity and professional hockey players. The effect held for men. It did not hold for women in the same study.

That single study is the origin of most of the claims now circulating: that a wider face signals dominance, that fWHR tracks testosterone exposure, and that it predicts how threatening a face reads to a stranger.

What happened when other labs checked

Independent replication has been mixed, and the mixed result is the part that gets left out of most summaries.

• A 2012 re-analysis of professional hockey data by Deaner and colleagues found that body weight, not fWHR, was the better predictor of penalty minutes once weight was included in the model. Heavier players draw more penalties; heavier players also tend to have a higher fWHR, which had been confounding the original correlation.
• A 2015 meta-analysis by Geniole, Denson, Dixson, Carré, and McCormick, published in PLOS ONE, pooled the accumulated studies and found the underlying effect was real but small: an average correlation of roughly 0.16 between fWHR and threat behaviour in men, and 0.12 for dominance behaviour across both sexes. The same meta-analysis flagged a low number of stimuli involving female and older-adult faces, and noted signs of publication bias in the threat-behaviour literature specifically.
• Perceived threat, how dominant a face looks to a rater rather than how the person actually behaves, showed a stronger correlation in the same meta-analysis. That distinction matters: fWHR predicts a snap judgment about a face more reliably than it predicts the underlying behaviour the judgment assumes.

What fWHR does not establish

It does not establish a testosterone link. Several studies have looked for a direct correlation between circulating testosterone and fWHR and failed to find a consistent one; the ratio is set largely by skeletal growth patterns, not by a single hormone reading. It does not establish attractiveness as a clean outcome either. Male dominance cues linked to a higher fWHR read as more attractive in some short-term-relationship contexts and less trustworthy in others, which is a trade-off, not a straightforward "higher is better" relationship. And it does not transfer cleanly to women: most of the underlying research was conducted on male faces, and the meta-analysis explicitly names the shortage of female stimuli as a limitation of what it can claim.

Why the online version overshoots the paper

The gap follows a familiar pattern: an r-value of 0.12 to 0.16 is a real, publishable, small effect in behavioural science. Compressed into a video or a forum post, a small effect becomes "wide face equals dominant, narrow face equals weak," a claim the original data was never powered to support. Carré and McCormick's own study excluded women from the aggression finding, which the popular retellings routinely drop.

How Facet measures it

Facet computes fWHR from the MediaPipe landmark set using the same definition as the original literature: bizygomatic width divided by upper-face height, measured from cheekbone landmarks and brow-to-lip landmarks respectively. Facet reports the ratio as a descriptive structural measurement, alongside the module's other parameters, rather than as a dominance or attractiveness score, because the research behind it does not support treating it as either.

A full example of how Facet's report presents structural parameters like this one, alongside all ten modules, is posted at [/sample-report](/sample-report). The citation list behind every threshold, including this one, is at [/methodology](/methodology).