Eye Shapes: Five Readings Before You Pick a Label

Five readings, and the mean span they fall in

The opening between the lids is the palpebral fissure. Its width runs from the inner corner, the medial canthus, to the outer corner, the lateral canthus. Its height is the widest vertical gap between the upper and lower lid margins, measured over the pupil. Canthal tilt is the angle of the line from the inner corner to the outer corner, read against a horizontal line. The inner-corner gap is the intercanthal distance; the pupil gap is the interpupillary distance, the number an optician writes on a glasses prescription.

These five readings were measured on 200 South Indian adults aged 18 to 26, 100 men and 100 women, in a 2013 study that used standardised frontal photographs and two independent examiners. The table prints the male and female means as one span. That span compares two group averages; it is not a normal range. The study also reports a standard deviation for every mean, and individual readings sit well beyond the two averages.

The 2013 study reports a standard deviation for each mean: opening width 1.79 mm in men and 2.18 mm in women, height 1.66 and 1.65, tilt 2.47° and 2.85°, inner-corner gap 3.57 and 3.09, pupil gap 3.84 and 5.24. A standard deviation is the spread around a mean, so the two means above are not the edges of the sample. Two examiners measuring the same photographs differed by 0.4 mm on opening width, 0.14 mm on height and 0.23° on tilt; the study calls that the inter-observer technical error, meaning the spread between repeated measurements of the same image. A clinical reference puts adult pupillary distance at about 63 mm with a working range of 50 to 70 mm, wider than the two means here; that is a cross-check on the pupil row, not a validation of the other four.

What the five readings decide, and what they leave open

Three ratios fall out of the means. Opening width divided by height is 2.58 in the female mean and 2.75 in the male mean. The inner-corner gap is 1.10 to 1.12 eye widths. The pupils sit 2.09 to 2.15 eye widths apart. Ratios are the part that travels between face sizes, because the absolute width cancels; they are also the part that neither dated guide in the next table prints.

What the readings cannot do is produce one label. Crease visibility separates a hooded lid from a monolid, and it is a yes-or-no look rather than a millimetre. Depth separates a prominent eye from a deep-set one, and depth is a profile question: a frontal photograph can suggest it but not measure it. The outline words sit on top of both. A sentence such as hooded, almond-shaped, slightly upturned and average-spaced is coherent because each word names a different axis.

There is also no published cutoff between almond and round. The 2013 study prints means, not categories. The guides print categories, not numbers. A width-to-height ratio of 2.6 can be called almond by one writer and round by another, and both descriptions can be honest. What the ratio does is make the description checkable, which a name alone is not.

What this means for you: take the five readings first, then treat the label as shorthand for the axes it names. If a label does not say which axis it is using, it is not yet a description you can check.

How to take the five readings with a ruler and a mirror

Work in even light, facing a mirror, with the eyes open and looking straight ahead. For opening width, hold a millimetre ruler level and place the zero at the inner corner; read the outer corner without pressing the lid. For height, measure the widest vertical gap over the pupil, from the lower lid margin to the upper. For canthal tilt, draw a horizontal line through the inner corner and read the angle up to the outer corner; a positive number means the outer corner sits higher. For the inner-corner gap, measure across the bridge from one medial canthus to the other. For the pupil gap, look at a distant point and have someone else read centre to centre, or use the PD on a prescription.

The study's own error figures are a useful scale. Its two examiners, measuring the same photographs, differed by 0.4 mm on opening width, 0.14 mm on height and 0.23° on tilt. Those are small numbers, but they were produced by trained examiners with a scale bar in the frame. A hand-held ruler in a bathroom mirror will do worse. Take each reading twice, on two days if possible, and keep the average.

One millimetre on a 30 mm opening is a thirtieth of the reading. That is small enough that a single reading can look precise while being wrong, and large enough to move a width-to-height ratio by about 0.09. The fix is not a better ruler; it is a second reading and a note of where the ruler sat.

What this means for you: write down the method beside the number. A reading of 30 mm taken from the inner corner is comparable with the table; a reading taken from the tear duct is not.

The four families behind the labels

The 2026 guide read for this page organizes its advice around four questions: 'Can you see your crease?', 'Do you see white above or below your iris?', 'Do your outer corners tilt up or down?' and 'How prominent are your eyes?' Those are four different axes — crease visibility, iris coverage, tilt and depth — and none of them is the same question as whether the opening is round or almond.

The 2024 guide lists six labels: almond, hooded, deep-set, monolid, downturned and prominent. The 2026 list adds round and upturned. Round is an outline word; upturned is a tilt word. Neither list includes a spacing label, even though the inner-corner gap is the reading a ruler settles most easily. That omission is why the labels can feel incomplete: they cover the lid and the tilt, then stop before the distance between the eyes.

Two of the families can be read from the five measurements, one is partly visual and one needs a profile. The table below sorts them that way rather than pretending all labels are the same kind of thing.

The labels are listed as the two dated guides use them, not as a closed taxonomy. The 2024 guide does not use round or upturned; the 2026 guide does not use close-set or wide-set. The one-eye-width rule for spacing is a rule of thumb, not a threshold published in the 2013 study: that study reports a mean ratio of 1.10 to 1.12 and no label. A ratio of exactly one eye width is therefore a convention, not a measured boundary.

Three dated sources, sorted by publication day

The table below records what each source actually publishes, in the order the pages were published. The point is not that one source is better than another. It is that the measurements and the labels live in separate documents, and a reader is left to join them.

The 2013 article was received on 29 September 2012 and accepted on 8 December 2012; its issue date is January 2013. The 2024 and 2026 dates are the pages' own publication dates. The 2026 guide's four questions are quoted in the section above. Neither guide prints a number, and the study prints no label; the dated table records that split rather than ranking the sources.

What changed, what did not

What changed, what did not, between 2013 and 2026 is the vocabulary, not the measurement. The 2013 study measures five readings and never names a shape. The 2024 guide names six shapes and prints no number. The 2026 guide names eight and turns the description into four questions, which is the closest any of the three comes to a method — but the questions still return a word, not a reading.

What did not change is the separation between the two literatures. No guide in the set prints a millimetre, inch or degree; no study in the set prints a shape label. A reader who wants both has to join them, which is what the first table and the family table above do.

The count change is informative too. Six to eight is not two more shapes measured; it is two axes added. Round is an outline word and upturned is a tilt word. The faces did not change between 2024 and 2026; the list did.

What this means for you: use the dated sources for their vocabulary and their questions, and use the 2013 means for the numbers. A later date does not make a guide more precise.

What eye shape does not tell you

A shape label describes the opening and the lid. It is not a test of visual acuity, and it says nothing about refractive error, colour vision, glaucoma risk or dry eye. It is not an ancestry test either: the 2013 sample is South Indian, and its means do not transfer to another population as a norm. Even within that sample, the male and female distributions overlap. For four of the five readings, the difference between the male and female means is smaller than either group's standard deviation; pupil spacing is the exception, at about one standard deviation.

Nor does a label rank a face. The site's attractiveness page treats proportions as one input and reports a tolerance around its result; a single reading is not a verdict, and this page does not convert any of the five into one. A canthal tilt of 5° is a description of where the outer corner sits, not a measure of how a person looks to anyone else.

What this means for you: read the label as a sentence about the eye opening, not as a claim about the person. The number and the label answer different questions, and only one of them is in the table.

What the readings change for liner, lashes, frames and brows

Liner is the place where the canthal line becomes visible. A wing drawn along the line the eye already has extends its tilt; a wing drawn above that line changes the apparent angle. On a hooded lid, a line close to the lashes stays visible when the lid is open, while a thick band across the moving lid disappears into the fold. A rounder opening reads longer when the liner runs past the outer corner; an almond opening already has that point.

Lashes work on the same visible gap. Curl lifts the lashes away from the opening; length adds a line above it. On a deep-set eye the brow bone can hide the root of the lash, so the visible effect is smaller than the length suggests. On a prominent eye the same length sits further forward. None of this needs a measurement, but the five readings explain why the same technique looks different on two faces.

Frames are a fit question first. Frame width follows face width and the bridge, not the eye opening; eye shape changes the lens shape, not the hinge position. The glasses guides linked at the foot of this page do the fit arithmetic. Brows are the one lever that moves the apparent height of the opening without touching the eye: a brow placed higher opens the space above the lid, and a low brow closes it.

What this means for you: change one thing at a time. Liner, lashes, a frame and a brow all alter the apparent opening, and changing all four at once makes it impossible to tell which one did the work.

Mirror versus photograph

A photograph measures a projection, not the eye itself. A phone held below the chin lengthens the lower lid and shortens the visible opening; a phone held above the brow does the reverse. A wide-angle lens close to the face adds the same distortion at the edges. The 2013 study used photographs, but standardised ones: a fixed head position, a camera aligned to the eye, and a scale bar in the frame. A casual selfie has none of the three.

Expression matters as much as angle. A raised brow, a squint and a smile all change the apparent height of the opening. So does fluid: a poor night's sleep can make a hooded lid look heavier and a prominent eye look puffier. Makeup changes the visible crease and the apparent tilt. Measure in a neutral state, at the same time of day, before deciding that a reading has changed.

If the tape and the photograph disagree, trust the tape for the five readings and use the photograph for the two visual checks. The tape can measure width, height and spacing; it cannot see a crease or a profile. The photograph is the second opinion, not the first measurement.

What this means for you: take the photo after the readings, at eye level, and keep it as a record. A photograph is good at catching a crease you have stopped noticing; it is bad at settling a millimetre.

Method and sources

The five readings come from a 2013 anthropometric study of 200 South Indian adults aged 18 to 26, 100 men and 100 women, all medical students in Manipal. The study measured standardised frontal photographs with a scale bar, took the right and left eye together, and had two examiners measure each image independently. Its table reports means and standard deviations; its second table reports the technical error of repeated measurements, which is where the 0.4 mm and 0.23° figures in this page come from. A student sample is not a population census, and the paper's own conclusion is that orbital norms should be ethnically specific.

The 2024 guide was read for its six labels, the 2026 guide for its eight labels and four questions. Neither prints a measurement, so the only numbers in the dated table are the study's. The clinical pupil-distance reference was read for a working adult range and is used once, as a cross-check. No source read for this page defines a cutoff between close-set and wide-set, or between almond and round, so this page prints none.

The page does not claim that a shape predicts eyesight, health, ancestry, character or how a person is perceived. The labels describe a visible opening at one moment. The five readings describe it more precisely. Neither is a diagnosis, and neither is a ranking.

Questions this page answers