Earring Size Chart: The Millimetres Against Your Own Jaw

Earring size for face shape: a chart indexed by a number you have

Two numbers are needed and both exist. The first is the jaw ratio: the run from one jaw corner to the other, divided by the widest horizontal measure across the cheekbones. This site's calculator prints it, and the tape method is jaw corner to corner over cheekbone point to point, which is a division and not a fitting room. The second is cheekbone width in millimetres, without which a ratio buys you nothing. The published adult band runs 128.2 to 148.8 mm and the working line used on two of this site's eyewear pages is 136 mm; the dataset behind the band and its date live on the page that publishes it, so the table below only uses the three numbers.

Find the row nearest your ratio and read across. The same ratio at the two ends of the width band is not the same jaw: at the bottom row of the ladder the two outer columns are 13.6 mm apart, and at the top row 23.5 mm apart — 0.92 in. That spread is why a single column of millimetres on anybody's sizing page cannot fit an outline it never asked about.

One warning about the ratio, because it is the easiest way to get a wrong number from a correct table. Use the value the calculator displays, which is stated on a soft-tape scale, and not a raw quotient of detected points. The published centre of the widest outline in the rule set is 0.95 on a tape; the detector's own quotient for the same row is 0.809, and at 136 mm those two are 19.3 mm of jaw width apart.

Every cell is one multiplication — ratio × cheekbone width — carried out exactly and rounded to one decimal, so any cell can be re-derived on the spot. The inch column is the product before rounding divided by 25.4 (0.66 × 136 = 89.76, and 89.76 ÷ 25.4 = 3.5339, so 3.53); carry the printed 89.8 through that division instead and it comes out 3.54. That is why the division is taken on the product. The published band for this ratio on a soft tape is 0.72 to 1.08, with a published standard deviation of 0.12; the ladder runs to 0.66 and 1.14, which is 0.06 — half a deviation — past each published end, and that overhang is this page's own choice rather than a measured range. Two rows sit exactly on the published ends, which is why the step is 0.06 and not 0.05. The seven published centres on this scale are 0.72, 0.72, 0.80, 0.85, 0.92, 0.95 and 1.08; they are deliberately not printed as a column, because a size question indexed by a category name is the question this page is not answering, and because two of those centres are shared and three are interpolated — see the limits section. The outer columns and the two end rows are combinations, not typical faces: the top row at the wide column implies a jaw of 169.6 mm, which is more than a tape finds on anyone. Read one row at your own width.

How big should hoop earrings be, once the answer is a ratio

A pair of earrings whose widest level sits on the line joining your two jaw corners adds twice its own diameter to that line, and nothing to the cheekbone line above it. So the ratio of the outline — the outline's, not your face's — rises by the pair's width divided by your cheekbone width. Four millimetres of diameter at 136 mm is 8 ÷ 136 = 0.059. Forty millimetres is 80 ÷ 136 = 0.588.

Two things fall out of that formula and both matter. The first is that the shift has no jaw term in it: two readers with the same cheekbone width get the same shift whatever their jaws are, so a size statement written as a ratio is not the same kind of claim as one written as a fit. The second is that it is not a small quantity in the units this rule set uses. The published standard deviation of the jaw ratio is 0.12 on a tape, which at the working width is a diameter of 8.16 mm. One small stud, worn level with the jaw corners, is a whole standard deviation of the measurement that is supposed to describe your outline. The entire distance between the narrowest and widest published outline centres is 0.36, three deviations — 24.5 mm of diameter. A pair of 40 mm hoops is 4.90 deviations — 1.63 times the full width of the scale this site uses to tell outlines apart at all.

That is the honest content of the word big, and it is the only content this page puts in it. The table reads a printed diameter three ways at once: the pair in millimetres, the ratio it adds at three cheekbone widths, and that addition as a count of published deviations at the working width.

Cells are 2 × diameter ÷ cheekbone width, exact and then rounded to three decimals; the last column is the same quotient before rounding, at 136 mm, divided by the published standard deviation of the ratio, 0.12 on a tape. That count is scale-invariant — the same quotient taken on the detector's own scale (0.026 for the same quantity) gives the identical figure, which is how this table was cross-checked — but the ratio itself is not, so the deviation column is the one number here that survives a change of measurement scale. Every row is the full value, which assumes the ornament's widest level sits on the line between the jaw corners; the offsets and what happens when it does not are in the limits section, where the 20 mm row is shown at 2.45 deviations and at 2.12 and where a hoop entirely above the line contributes nothing to it. The 8 mm row is 0.98 deviations, and one whole deviation at the working width is a diameter of 8.16 mm (0.32 in), 7.69 mm at the narrow end of the published cheekbone band and 8.93 mm at the wide end.

Why an 8mm earring size is a standard deviation and 40 mm is a third of a jaw

The table above fixes an ornament in ratio units. This one asks the other question, the one a shopper actually has: how large is the thing compared with the line of the face it hangs beside. At the working width the three rows below are jaws of 97.9, 122.4 and 146.9 mm — 3.86, 4.82 and 5.78 in across — the published lower end of the band, the middle row of the ladder, and its published upper end. All three are rows of the first table, so any cell here can be rebuilt from a cell there.

Read a hoop size off the trade's own ladder and the numbers are not marginal. A 12 to 14 mm hoop is 9.8 to 11.4 per cent of the middle row, so its pair covers a fifth to nearly a quarter of that jaw line. A 30 to 40 mm hoop is 24.5 to 32.7 per cent, and its pair runs from 49.0 to 65.4 per cent of it. A pair of 50 mm hoops spans 81.7 per cent, within nineteen points of covering the face from corner to corner.

Going the other way, a pair exactly equal to the middle-row jaw line needs a diameter of 61.2 mm — 2.41 in — and that is not an absurd size on a listing. It is inside the top band of a hoop ladder printed by the same February 2026 guide that supplied the face sentence above, a ladder that runs from huggies to “60mm+” for its largest hoops, and it is the size a December 2025 jeweller's table reaches in its last row, “30mm - 50mm+”, whose description column for that row reads “Jawline length”.

Two of those are the trade naming the jaw line — worth saying plainly, because it is the closest any of the fifteen pages gets to this page's own landmark. Neither asks you for a number on it. The description column names a level and the ladder stops there; the 60 mm band is a size category with an adjective attached. The arithmetic below needs the line measured, and the measurement is a division you can do with a tape.

These are the numbers a claim about proportion has to be made of. Whether a reader's face looks smaller next to any of them is not in the table, and the arithmetic does not lean that way: the added width is added to the face's own silhouette, which is the opposite direction. The perceptual half of that sentence is dealt with below rather than borrowed from this one.

Jaws are 0.72 × 136 = 97.9 mm, 0.90 × 136 = 122.4 mm and 1.08 × 136 = 146.9 mm, the published lower end, the middle row and the published upper end of the first table's ladder, and each cell is diameter ÷ jaw × 100, or twice the diameter for the last two columns, exact and then rounded to one decimal. Those three are rows of a ladder rather than three people: a jaw of 146.9 mm beside a cheekbone width of 136 mm is a jaw wider than the face above it, which is what the published top of this band describes. Read against the 128.2 mm and 148.8 mm columns of the first table instead and every cell in this one scales by 136 ÷ 128.2 = 1.061 or by 136 ÷ 148.8 = 0.914 — 6.1 per cent of itself upward, 8.6 per cent downward — which is the arithmetic consequence of a cheekbone width you have measured being different from the working line used here. The columns are shares of a line at one level, and an ornament is widest at one level chosen for it; where the two do not coincide is the offset in the limits section. Nothing here is a perceptual quantity: it ends where the ornament's share of a measured line ends.

Earrings for square face shapes, and the column that cannot divide them

Three outline names recur in the pages this page was written against — square, round and oval, each of them on two of the fifteen, while two further names appear once each and nobody mentions the rest — so it is worth being exact about what separates the first two here, because the honest answer is an uncomfortable one. The rule set carries two columns for the lower face: a width ratio and an angle at the jaw corner. The angle is defined by three points that are all on the face — each jaw corner, measured against its own cheekbone point and the chin tip, the two sides averaged — and nothing worn on the ear moves any of them. An earring cannot change an angle whose vertices are all facial landmarks.

The site's own calibration is the reason this page does not lean on that column. On fifty photographs the measured jaw angle for the square and round groups came out at about 133 degrees for both, while the printed rule set places their centres 14.1 degrees apart on the detector's scale and 24 degrees apart on a tape. That printed gap is ordered from the literature the rule set was fitted against; it is not a difference this site's measurements settled. Which is also why the size advice in the comparison keeps drifting into that column's vocabulary — corners, softened, angular — and prints no measurement beside it, as the 19 March 2026 row below shows.

The width column does carry a difference between those two published centres, and it is small in exactly the units a size claim needs. They sit 0.1016 apart on a tape, which is 0.85 of one published deviation, and at the working width it is 13.8 mm of jaw. A pair of 6.9 mm earrings, if their widest level were on that line, would move the outline by precisely that distance. So a sentence of the form this outline needs that many millimetres is comparing an ornament with the difference between two rows of a table whose angle column its own source cannot separate. That is the whole reason the first table above is indexed by a ratio and not by a name: the ratio is the one quantity here that a reader can measure, and the measurement is worth the same whatever it is called.

Do big earrings make your face look smaller?

There is an answer to this that has arithmetic in it and an answer that has a mirror in it, and this page only has the first. The arithmetic: your jaw ratio is computed from three widths of your face, and an earring is none of them, so the measurement does not move. The outline moves, by the pair's width divided by your cheekbone width, and the direction is wider, not narrower. At the working width that is 0.176 of a ratio for a 12 mm pair and 0.588 for a 40 mm pair, against a published deviation of 0.12 and a published band width of 0.36. Whatever a reader sees in a mirror, the quantity this table changes has the opposite sign to the phrase.

So the defensible version of the claim is a comparison of sizes rather than an effect: an ornament whose diameter exceeds 24.5 mm — a pair wider than 49 mm — is adding more width at the jaw line than the entire distance between the narrowest and widest outline this rule set recognises. That is a statement about the ornament's scale against a published spread, and it is testable from the two tables. Anything stronger than that — that a wide hoop reads as a smaller face — would need a comparison of faces judging faces, which is a study, and this site has not run one. The fifteen pages read for this page do not contain one either; they contain assertions, and the only quantity any of them asks a reader to measure is a distance on an ear.

If you want the version of this question that the tables can actually settle, it is a different sentence: does this ornament cross the line of my jaw, and by how much of it. Read the diameter against your own row in the first table, take the pair share from the third, and you have a number that will still be true when the listing changes its adjectives.

What has been printed about this, in date order

The rows below are the readings behind the first paragraph of this page: fifteen positions retrieved on 29 September 2026 and read in the same sitting, thirteen of them in a text layer, and one book. Positions are described by what they are and none of them is linked. Dates are given as each position carries them, including the two kinds a reader cannot see on the page itself. One qualification about the dates, because it is the trap in this set: several of these pages contain strings that look like publication dates and are not — a theme name, a content-delivery path, an app's creation stamp, a feature flag. Those were opened and identified, and only the rows below carry a date belonging to the page itself.

Two things moved and one did not. Units arrived early and thickly: six of the six dated web rows print millimetres — 29, 50, 59, 15, 104 and 12 of them — and eleven of the fifteen readable positions carry tables, twenty-one between them. The face arrived too, in one sentence that puts a millimetre band and a word for a face in the same breath, and in three more where a figure sits somewhere in the same paragraph.

What never arrived is the third thing, and that one needs its count stated carefully. Stripped of navigation and of stylesheet noise, cheekbone, chin, forehead, face length, face width and measure your face return zero occurrences across the readable positions, and jaw on its own returns zero. Jawline returns three times on three different pages: once as an adjective about square faces with no figure beside it, once as the description of the largest hoop row in the set's most carefully graded table, and once to say where the lowest point of a drop falls. Three uses of one word and not one measurement asked of the line it names.

What the set does measure is the ear: lobe and lobes occur 54 times across the seven pages that use them, and one page tells you to put a ruler from your piercing hole to the bottom edge of your lobe and add a millimetre for a snug fit. That is a genuine instruction, with a genuine number taken off a genuine body, and it is about an ear. Nothing in the set asks you to bring a number taken off your face and divide it into a size.

Sorted by publication date, with the nine positions carrying no date of their own placed last and in rank order; the two structured-data-only rows are stated rather than smoothed over, because they are the difference between a page that was dated and a page that was stamped. The span runs from the 1901 imprint to 1 September 2026 — one hundred and twenty-five years — while the six dated web rows cover 254 days, and inside the ten highest-ranked positions the dated rows cover 169 days from 21 December 2025 to 8 June 2026. Three addresses were removed from the list before it was read: two image results and one page about ring sizes, which is why fifteen content pages are described as ten ranked positions. Two positions were read by a browser where a text fetch was refused, and their figures come from that reading rather than a second one; one of the fifteen, the twelfth row, could not be re-resolved to a stable address on a second pass, so its reading is from the first pass of the same day and was not reproduced. The word counts in the paragraph above are from a fresh pass over the same saved pages, with scripts, styles and tags stripped and each term matched at word boundaries. No row here has been replaced with a substitute, and none of the readings is quoted at length beyond the clause the position is being judged on.

What these tables are not

The position is a model, and it is the soft part. The ratio addition assumes the widest level of the ornament sits on the line between your jaw corners. Treat the hoop as a circle of the printed diameter and that line misses the widest level by d millimetres, and the pair's contribution there falls to four times the square root of the radius squared minus d squared, or 0 if the line misses the circle entirely. On that reading a 20 mm hoop offset by 5 mm contributes 34.6 mm rather than 40.0 — 2.12 deviations instead of 2.45 — and the same hoop offset by its own radius contributes nothing at that level.

A stud whose widest level sits at the ear rather than the jaw is off that line by more than its own radius for most diameters sold, so at jaw level it adds nothing and the second table reads 0 for it. What survives is the third table, where a 4 mm stud is 3.3 per cent of a middle-row line — a comparison, not an addition.

The offsets are stated at 0, 5 and 10 millimetres rather than folded into one tidy figure because no position in the comparison measures the level at which an earring is widest, and this site does not either.

Read the same rule from the other line and the sign turns over: an ornament whose widest level sits at the cheekbones is added to the denominator, and there a 12 mm pair moves the widest outline centre in the rule set that rests on photographs — 0.9508 — down to 0.8082, which is 1.19 deviations in the direction of narrower. Both of those numbers describe an outline, not a face; the face's own ratio is fixed, and no earring appears in its definition.

The deviations used as a ruler here are estimated quantities. They were pooled from the four groups in the calibration set with five or more photographs each — eighteen, twelve, ten and six, forty-six faces and forty-two degrees of freedom — and the published standard deviation of the ratio is 0.12 on a tape scale. Two of the seven outline centres are shared rather than distinguished, and three rest on one, three and zero photographs respectively and are ordered along the literature rather than measured. That is why this page spends its deviations on a diameter and not on a diagnosis: an 8.16 mm shift is the same claim for anybody at 136 mm, whereas any statement about which centre you are nearer belongs to the rule set and to the page that publishes it, with the same limits written there.

The cheekbone band is a band. 128.2 to 148.8 mm is one standard deviation either side of the measured sex means this site publishes, and 136 mm is a working line rather than a person. If your own width is outside that band the tables are still usable, because the arithmetic is one multiplication and one division and every cell is derived from the row it prints. And nothing here is about the ear itself: not healing, not irritation, not what a heavy stone does to a piercing over time. One of the fifteen pages builds its whole sizing section on that ground and this page does not follow it there.

How to use the three tables

Take two numbers, then do four steps. The numbers are your jaw ratio and your cheekbone width: from the calculator, or off your own face with a tape and a piece of string, jaw corner to corner divided by cheekbone point to point, and the width in millimetres rather than in a unit you have to trust. Then: read your jaw width off the first table at your width column; take the diameter off the listing and double it; divide that by your cheekbone width to get the ratio the ornament adds, and by your jaw width to get the share of the line it covers; and decide which of the two questions you were actually being asked when you started, because the listing is answering neither.

Write down the two quotients, since they are the only things here that stay true. A 20 mm hoop at a 136 mm cheekbone width adds 0.294 of a ratio as a pair and covers 16.3 per cent of a middle-row jaw line on its own; a 40 mm hoop adds 0.588 and covers 32.7 per cent. If the second number feels too large for your face, the check is not a mirror but the offset in the limits section: move the level at which the pair is counted and the whole row shrinks. If the first number feels too small, the check is the published scale — three deviations cover the distance between the narrowest and widest outline this rule set recognises, and 24.5 mm of diameter is all three of them at once.