The outer corner sits clearly above the inner one. The eye reads as upturned or "almond", and photographs tend to pick it up as alertness. Liner pulled further up adds little; a softer, rounder shape balances it.
Level · 0° to 2°
Corners at roughly the same height, which is where most measurements land. The eye shape is neither lifted nor dropped, so it takes almost any liner, brow or frame direction.
Negative · below 0°
The outer corner sits below the inner one. Usually read as soft and approachable. If you want more lift, liner flicked up at the outer corner and a brow tail kept high do more than anything else.
What actually moves the number
Three things change the reading before anatomy does. Head pitch: chin down adds apparent positive tilt, chin up removes it, and half a degree of head movement is easily two degrees on this measurement. Camera height: shooting from above tilts the whole face plane; a phone held at chest height and angled up will flatten your result. Lens distance: a phone held closer than about 40 cm distorts the mid-face, and the eye corners move with it.
If you want to compare yourself across two photos, match all three first. Otherwise you are measuring the photograph, not the face.
Measuring it yourself
Camera at eye level. Not above, not below. If you can see the top of your head, or up your nose, the reading is already wrong.
Chin neutral, and look into the lens — not at the screen. On most phones those are a few centimetres apart, and that gap is worth about a degree.
Arm's length or further, roughly 40–60 cm. Closer than that and the lens stretches the mid-face, taking the eye corners with it.
Even light on both sides. A window on one side deepens one socket and moves where you think the corner is.
Then the angle itself: find the inner corner (medial canthus) and the outer corner (lateral canthus) on each eye, draw the line between them, and measure it against the horizontal. Average the two eyes.
Done carefully by hand you will land within a degree or two of what the test above reports, and that is close enough — the four points above matter far more than the method. What the scan adds is that it places both corners from 478 landmarks instead of your judgement of where a corner is, and it does both eyes the same way every time.
What counts as typical
Most published series on adults sit a few degrees positive. The figure usually quoted from Western sources is around 4°, but it is not universal: a 2020 series of 76 Korean patients measured a mean of 8.45° on the right and 8.42° on the left, with not one negative tilt in the group. That series was patients presenting with facial asymmetry rather than a general population sample, so it is not a norm either — but it sits far enough from 4° to make the point that "normal" depends entirely on who was measured.
Which is why this page gives you the angle in degrees and shows you where the band boundaries are, rather than telling you whether you passed. If your reading lands near a boundary, head angle alone will move it across.
Two questions this measurement gets asked
Does it change with age? Slightly, and downward. The tendon that anchors the outer corner loosens over decades and the lower lid descends with it, so a tilt measured at fifty tends to read a degree or two lower than the same face at twenty. It is a slow drift, not something you will see between two photos a year apart — if your number moved that fast, look at head angle first.
Is there a tilt that is most attractive? There is no figure the evidence supports, and the sites that quote one are quoting each other. What the research does support is broader and duller: averageness, symmetry and sexual dimorphism all rate as attractive across cultures. Canthal tilt is a small component of the first two and is not studied on its own as a preference. Anyone selling you a target angle is ahead of the literature.
Where it sits among everything else
Canthal tilt is one angle. It is not a score, and on its own it says very little: two faces with identical tilt can read completely differently once brow position, eye aperture and mid-face length come in. The full scan puts the tilt next to ten other proportions and a percentile, which is a more honest way to look at it than one number in isolation.
References
The measurements and judgements on this page rest on the work below, including findings that cut against what we do. Reference values are ranges the literature reports, not targets to reach.
Kim et al. (2020). Diagnostic analysis of vertical orbital dystopia and canthal tilt for surgical correctionJournal of the Korean Association of Oral and Maxillofacial Surgeons 46(6):379–86 · PMC7783177Measured 76 Korean patients: mean canthal tilt 8.45° ± 3.53° on the right and 8.42° ± 3.81° on the left, with not one negative tilt in the series. A caution about reference values — these were patients presenting with facial asymmetry, not a general population sample, but the figure sits well above the 4° often quoted from Western sources.
Rhodes (2006). The evolutionary psychology of facial beautyAnnual Review of Psychology 57:199–226 · doi:10.1146/annurev.psych.57.102904.190208A review and meta-analysis finding that averageness, symmetry and sexual dimorphism are all rated attractive, in both sexes and across cultures. This is the evidence behind measuring symmetry at all.
FAQ
Questions people ask
What is canthal tilt?
It is the angle between the horizontal and the line running from your inner eye corner (medial canthus) to your outer eye corner (lateral canthus). If the outer corner sits higher, the tilt is positive and the eye reads as upturned. If it sits lower, the tilt is negative and the eye reads as downturned.
What is a normal canthal tilt?
Published measurements on adults cluster around 4 to 6 degrees positive, with a wide spread on either side, and the figure varies by ethnicity and by how the measurement is taken. This page reads the sign the way the definition above does: 2 degrees and up is positive (upturned), 0 to 2 degrees is level, and anything below 0 is negative (downturned). Because published adult values cluster around 4 to 6 degrees, most people measure positive.
How is it measured here?
The 478-point face model finds both eye corners, and the tilt is the angle of the inner-to-outer line against the horizontal, taken separately for each eye and then averaged. Averaging matters, because one eye is almost always a degree or two off the other.
Why does my result change between photos?
Head pitch is the biggest factor. Tipping your chin down or shooting from above adds apparent positive tilt; lifting your chin or shooting from below takes it away. For a stable number, put the lens level with your eyes, keep your chin neutral and look straight into the camera.
Can canthal tilt be changed?
Makeup, brow shape and glasses change how the tilt reads without changing the anatomy, and that is what this site covers. Anything surgical is a medical question for a qualified clinician, and nothing here is medical advice.
