Square vs Rectangle Face Shape: Half of the Usual Description Is Backwards

The Face Length Each Row Asks For, in Millimetres and Inches

One of the five checks this site publishes is a ratio: your face length divided by your cheekbone width, measured across the widest part of the face just under the eyes. Because it is one measurement divided by another, the units cancel — the same face gives the same number in centimetres or inches. The table below turns the two ends of those bands into the units a tape measure comes in.

The square row's band on that check runs 0.87 to 1.23. The row this site labels oblong — the column that carries the word rectangle in its header — runs 1.47 to 1.83. Between 1.23 and 1.47 there is nothing. This table turns those edges into lengths you can hold a tape against, at the three cheekbone widths this site works from: the narrowest and widest adult face in its published cheekbone band and its working line, all three named on the glasses-for-a-round-face page.

Every cell is one multiplication: a printed edge of a band times the width in the first column. Nothing in the table is a new measurement, a model, or a re-fit, and the middle row can be redone at your own cheekbone width in place of the printed one. The fourth column is the one that makes the rest of this page worth reading: the empty strip, in millimetres, for the same three faces.

What that means in plain terms: a face whose length lands inside the square row's numbers is not near the rectangle row by a small amount. At an average cheekbone width the two rows are roughly a hand's-breadth apart — the space in the fourth column is more than three centimetres.

Four things this table does not hide. One: the band edges multiplied here are the two-decimal numbers printed in this site's published table and on its about page, so a reader multiplying them at their own width gets these cells; the unrounded bands behind them move each edge by about 0.3 mm at the middle width and change nothing on this page. Two: the ± one tolerance that makes a band is a chosen band width, not a natural boundary — widen both rows to two tolerances and the square row runs to 1.41 and the rectangle row starts at 1.29, so the empty strip closes into an overlap of roughly 0.13. What survives either choice is that this check is where the two rows sit furthest apart. Three: on this check one tolerance is 23.1 mm at the narrow width, 24.5 mm at the middle and 26.8 mm at the wide (0.91 / 0.96 / 1.05 in) — the strip in the last column is wider than a tolerance at every width in the table. Four: the rectangle row is the same row this site labels oblong, and the oval column overlaps it from 1.47 to 1.63, which is 20.5 mm at the narrow width, 21.8 mm at the middle and 23.8 mm at the wide.

Where These Numbers Are Already Printed

None of the numbers above is new to this site, and the page that carries them is not this one. The centre for each row and the tolerance for each of the five checks are published on the about page, where the whole rule set is printed for anyone to check. The full table of seven rows and five checks is printed on the rectangle face shape guide, which is the one page on this site that carries all seven rows at once; the naming of that column as oblong is explained there too, in the section on where the name comes from.

The same rectangle guide already answers the question this page was written for, twice over. Its opening answer states that the square row sits at 0.87 to 1.23 and that the two bands do not touch; its first section adds that the oval band spans 1.27 to 1.63 and therefore shares everything between 1.47 and 1.63 with the rectangle family. That is the same fact this page turned into millimetres in the table above, and it is stated there in words. The site's square guide has carried a question on exactly this pair — square or oblong — for longer than this page has existed, and it answers with the widths and the corner rather than with numbers; its own answer points at the measurement table on that page for the ranges.

So this page is not the first place anybody has asked, and not the place where the bands were first printed. What it adds is only the column nobody has multiplied out: the size of the gap in millimetres, and the size of the overlap with the oval row, at the three face widths the site already works from.

Plain version: the bands and the reasoning live on the rectangle guide and the about page; what the table above adds is the size of that gap in a tape's units, at your own face width.

The Hybrid Sentence, Tested Against the Bands

Here is the sentence, quoted from the cosmetics guide that ranked fourth in a day's results for this question: `A rectangle face shape is a combination of an oval face shape and a square face shape.` A page ranked thirteenth that same day, outside the top ten, spells the same idea out with the two halves named separately: `the face length appears long like an oval, but the forehead, cheekbones, and jawline are similar in size like a square`. That second version is the more useful of the two, because it says which feature is supposed to come from which parent, and a claim that names its features can be tested.

Test the length half first. If a rectangle really is a mixture of a square face and an oval face, the mixture should sit somewhere between them on the check the sentence names: length divided by cheekbone width. On this site's published centres the square row is 1.05 and the oval row is 1.45, so the midpoint between the two named parents is 1.05 and 1.45 added together, then halved — 1.25. That is the whole calculation, one step, done on numbers printed on the about page.

Now look at where 1.25 actually falls. The square row's band on that check stops at 1.23, so the midpoint is outside the first parent's own range. The oval row's band starts at 1.27, so the midpoint is outside the second parent's range too. Meanwhile the rectangle row this sentence is defining starts at 1.47, which leaves the 0.24 of unused ground this page has been counting since the first table. A face reading 1.25 on length is not inside any of the three rows' bands on that check; read on that check alone it is inside the bands of three other rows — heart, diamond and triangle — which is a reminder that one reading on one check is not a classification, and never has been on this site.

So the length half of the hybrid sentence does not describe the row it is attached to. The rectangle row's real neighbour on length is the oval row, and the two share ground from 1.47 to 1.63 — the last column of the first table. The sentence names the oval connection as a parent and the square connection as the shape itself, while the published numbers put the length overlap on the oval side and the sameness with square on the widths. That is the same pair of facts in the opposite order, and the order is what the sentence gets wrong.

The widths half is a different story, and it survives. Forehead, cheekbones and jaw close together, and a corner under the ear rather than a curve, is exactly the arrangement the square row is named for; the clinic page ranked eleventh that day, outside the top ten, puts it as well as any styling guide does: `These two are confused more often than any other pair, because horizontally they are identical.` Horizontally identical is a claim about the widths, and on this site's table the two rows are indeed the same arrangement across the width checks. It is only when you add the vertical that they stop being relatives.

What a reader should take from this section: use the hybrid sentence for the horizontal half — near-equal widths, a cornered jaw, and styling that follows from that — and drop it entirely when it tries to tell you where your face length sits, because on that check it points at the wrong neighbour.

Which Checks the Two Rows Actually Share

A tempting way to write this pair is to count how many of the five checks give the same answer for both rows, and that count is already on this site's record. The diamond-versus-heart page prints a table that works out, for several pairs including this one, how many cells two rows give identical readings for, and it names which check is the identical one. This page does not recount that number or restate it: go and read that row.

What matters for anyone holding a tape is a shorter version of the same fact. Three of the five checks are width checks and one is an angle, and on none of those four do the two rows land as far apart as they do on the length check — on those checks the two rows are either close or the same, and the corner under the ear, which is the feature that makes both shapes look alike to a mirror, is the single reading this site's own form lets you leave blank. That last point belongs to the round-versus-square page, which is where this site has already worked out what happens to a result when that field is empty; this page only points there.

The one check where the two rows genuinely part company is the length ratio, and it is worth saying what that does and does not buy you. It buys you a decision rule: above 1.47 you are not reading a square row's length, below 1.23 you are not reading a rectangle's, and in between neither row claims the reading. It does not buy you a verdict on which label to wear, because the site's own classifier does not run one check — it scores all five readings together and returns a margin, the gap between the row it puts first and the row immediately behind it, and its documentation says what to do when two rows are close.

The short form for anyone holding a tape: the length number tells you which of these two pages to read, and nothing else. Which word ends up on your result is decided by all five readings together, on the tool, not by this one ratio.

The Answer Is Filed Under the Other Word

Before any of this can be searched for, there is a naming problem, and it is not a minor detail: no page in that day's results takes this pair on under the word rectangle. The highest-ranked page in that day's results renames it outright — `There are two subtypes of the oblong face shape; the round oblong and the square oblong` — which dissolves the rectangle into a variety of another shape. The only page in the top ten that is about this pair and gave up its text carries the title with the other word and explains itself: `Square and oblong (also called rectangular) face shapes are frequently confused.` The glasses guide ranked ninth lists both words in its navigation as if they were separate shapes and then defines the rectangle as `its elongated appearance`.

This site does the same thing. Its column for this family is labelled oblong, its square and oblong guides say the words describe one outline, and its blog index quietly files the two rectangle pages under the oblong column so that a reader filtering by that word finds them. How the two words came to name a single column, and where the softer and sharper ends of that family should be drawn, is a different question from which band your face sits in, and this page does not settle it.

What it means for you: if you keep landing on pages that talk about oblong faces when you searched with the word rectangle, nothing has gone wrong, on their side or on this site's.

Every Dated Answer Found for This Question

The table below is ordered by the date printed on each page, oldest first. Rows are described by what kind of page they are, not by who runs them; the dates are the publication date on the page with the revision date where the page carries one. It is included because a reader comparing this pair deserves to know how long the current advice has been the current advice, and whether anyone ever put a number on the border. What you can do with it is smaller: before trusting a rule of thumb about this pair, check the year printed on the page that handed it to you.

Identities are given by type, not by name or link, for every row that is not this site. Dates are the publication date printed on the page, with the revision date where the page carries one; the self-test row carries a publication date and no revision date on the page. Nine answers that day are not in this table, and each is missing for a reason about its date rather than about its content: four of the ten best-ranked pages returned an error, an interstitial or a sign-in wall instead of an article, and one question page ranked outside the top ten did the same; two readable pages carry no date printed anywhere on them, and those two are the pages ranked first and ninth, whose exact words section 5 quotes when it describes the renaming — no date is claimed for either; two readable pages ranked outside the top ten had their addresses recorded in a form this page could not re-check, and neither is used for anything. The generated answer block at the head of the results for this question was not counted as a source.

What Moved, and What Did Not

Down the middle column of that table, over seven years and four months, the format has not moved. Seven answers written outside this site, and in all seven the border is drawn in words: an outline read off a shape traced on paper, a comparison between length and width, a list of six named shapes, one shape described as another one stretched. Three of the seven build the shape out of something else — stretched, or two of them combined — the earliest printed in 2020, and all three carry a revision date in 2025 or 2026. Not one of them hands a reader a figure in millimetres for the border between these two shapes, and not one says what to do when a face fits the description of both.

What did move is the honesty about the arithmetic, and it moved outside the top ten. The page ranked eleventh that day is the only row in this table that says a number is not enough by itself: `the ratio alone will not settle it`. Even its length range is written in words — one and a half to two times — which spans 1.5 to 2.0: it starts just above this site's published 1.47, runs past the published top edge at 1.83, touches neither exactly, and covers the upper part of the oval row's band from 1.5 to 1.63 on the way. Two of the ten answered the question by renaming the shape into another one, which is a way of not having to draw a line at all.

So the gap this page sits in is not a missing opinion. Eleven dated answers are in that table; the seven written outside this site agree about what these two faces look like, and one of this site's own rows already prints both length bands and says they do not touch. What none of them prints is those bands as numbers you can hold against a tape. The strip between them, in millimetres and inches at three published widths, is the part this page adds, and the table above is what it is trying to be different from.

One sentence to carry away: across those seven answers the only movement is an admission, printed by the newest of them, that a number cannot settle this on its own — and no page in the table, older or newer, has put the two length bands side by side in millimetres.

If Your Numbers Land Near the Border

The length check needs two measurements, and the honest difficulty in this pair is not deciding which number you have, it is deciding whether you measured the length the way the table means. Face length on this site runs from the hairline down the middle to the tip of the chin, so anything covering the hairline takes length out of the reading. The site's square guide puts the consequence for this specific pair directly: `a fringe covering the hairline shortens the measured length and pulls an oblong reading down toward square`. Take the hair back off the forehead before trusting a reading that sits low.

Then compare your own quotient with both edges before comparing it with a word. Divide face length by cheekbone width; if the answer is above 1.47 you are not reading a square row's length, and if it is below 1.23 you are not reading a rectangle's. If it lands between those two numbers, neither row claims it, and that is a property of the published bands rather than of your hands — the fourth column of the first table is how wide that ground is at a face like yours.

If your reading is inside one band and the mirror still says the other word, stop trying to average the two sets of advice. The horizontal half of this pair is genuinely shared, which is why the two pages read alike; the difference between them is only how much length to allow, and that is the thing the number settles.

If you use the calculator rather than a mirror, the same ground comes back in the open: it scores all five readings at once and shows a margin against the other rows rather than a single word, which is what you want here, because the honest answer for a face near the edges of a band is two names and a gap rather than one name.

• Push the hairline into view before measuring length, or the reading moves toward the square row whether your face did or not.
• Divide face length by cheekbone width and compare it with both edges — 1.23 and 1.47 — before comparing it with any word.
• If the answer lands between the two, treat it as the bands being quiet, not as a bad measurement, and re-check the hairline first.
• If the reading is inside one band but the mirror disagrees, style for the horizontal half the two shapes share rather than merging two sets of instructions.
• Where the number is close to an edge, let the tool report a margin instead of picking one word for you.

The tape positions, the order to take them in, and what to do when a number is refused: How to measure your face, step by step ↓

What This Page Cannot Prove

The bands are not measurements of two kinds of face. They are a centre and a chosen tolerance per row, fitted on a set of fifty photographs labelled by one person, with no second rater and no published agreement figure — this site says so on its about page rather than hiding it, and the same note says that three of the seven rows, including heart, diamond and triangle named in the third section above, were placed by interpolation between measured neighbours rather than by measurement. Six of those fifty carry the square label, twelve the oblong label and eighteen the oval one, so thirty-six labelled faces is the whole evidence base behind the three bands this page multiplied out. Those three counts are printed on the round-face neckline page; the about page publishes the weaker form of the same fact, that four rows had five or more labelled examples each while heart, diamond and triangle had one, three and zero, and the single spread every band on the table uses comes from those four fitted groups pooled together.

The strip in the first table is arithmetic on chosen band widths, and this page says it twice rather than once. One tolerance is 0.18 on that check; widen the bands to two tolerances and the empty strip becomes an overlap, printed in the footnote of the first table, and the claim that the two rows never touch goes with it. What survives either band width is the order: this pair is further apart on length than the rectangle row is from the oval row, and the hybrid sentence points at the wrong neighbour either way.

The 1.25 in the third section is a position on a ruler, not a claim about anybody's face. It says that the midpoint of two published centres falls outside the two bands it was derived from; it does not say that faces sit there, that the rows were fitted to place them anywhere, or that the mixture idea is meant to be read as arithmetic at all. The styling writers who use that sentence are describing an impression, and an impression is not a band.

No perception has been measured. This page says where a tape reading falls against a published table. It does not say that a stranger can see the difference between the word square and the word rectangle, or that a gap of 0.24 on that check is visible rather than administrative, and it says nothing about which of the two reads better on a person. There is no study, here or anywhere in the table above, mapping these ratios onto what other people actually judge, so any sentence on this page about what you look like would be a guess wearing a number.

And the comparison of published answers has a hard edge that the table's own footnote names but the prose cannot undo. Several of the pages that answer this question would not give up their text at all: five returned an error, an interstitial or a sign-in wall instead of an article — four of them inside that day's top ten, the fifth a question page ranked outside it. Those pages are not evidence that nobody has written a number, only that nobody's number could be read here; they are not rows in the table above, and nothing on this page rests on what they might have said. The claim this page makes about the absence of arithmetic is therefore limited to the dated pages whose text could be read. For a reader the rule that survives that limit is a small one: use these numbers as a check you can redo against the about page, and treat anything they do not settle as still open.

Questions this page answers