Hats for Round Faces: How Much Crown Your Own Head Can Spend

How Much Room a Round Outline Has Above It

The range this site publishes for face length runs 1.05 to 1.65, meaning length divided by cheekbone width measured with a soft tape, and the stretch from the lowest figure it publishes to the highest is what this site calls a band. The round outline sits at 1.05 — the bottom edge itself, the same figure as the square outline and the lowest of the seven rows. A crown adds to the numerator of that division, the top number in it, and cannot subtract from it, because the excess of a crown above your head is either positive or zero; there is no such thing as a crown standing below the top of your skull. So the question the ten pages argue about — tall or low — has a shape nobody in those ten states: it is one ladder, running upward from the floor of a published band, and it has a ceiling.

The ceiling is the published top of that same range, 1.65, which is where the site's longest outline sits, and reaching it from the floor takes 0.60 of the division. Every cell below is a lift divided by a width, using the ratio column this site already published for the 136 mm working width, added to the round outline's starting figure. Nothing here is newly computed except the addition and one subtraction.

The practical reading: the worksheet most people stop at — a crown standing 45 mm high, a little under two inches — gets you to 1.381, and the top of the range is still 36.6 mm of crown away at that width. That gap is the page's answer to how tall is too tall: not a style limit, 36.6 mm of unused ladder.

Provenance, column by column. The millimetre and inch columns and the third column are the worksheet this site published on 6 October 2026, lifted from the table in `What a Crown Adds to the Length Ratio` on the face-shape hat guide; the inch figures are the millimetres divided by 25.4 and rounded separately, which is why dividing the millimetres beside an inch can differ from it by a hundredth. The fourth column is that published ratio column added to 1.05, and 1.05 is this site's published centre for the round outline — the single figure it gives for that outline — on the soft-tape scale, which is also the published minimum of the band; it is used here as a ruler, not re-derived. The 5 mm step is the edge of a worksheet, not a finding, and no band is published for a lift to sit inside. The final row is the full published span of the length range: 0.60 of a 136 mm width is 81.6 mm, 3.21 in, and that span is 3.29 of the published standard deviations of this ratio, a standard deviation being the spread the site publishes around its own middle figure. The spread published with it is 0.18, which is worth 23.4 mm of crown at the narrow published width, 24.8 mm at the working line — the width this site runs its arithmetic at — and 27.1 mm at the wide one. If your cheekbone measure is not 136 mm, redo the division with your own width; the three published widths of that worksheet are 128.2 mm, 136 mm and 148.8 mm, and the arithmetic is the same one step.

The Ten Do Not Agree on Which Way to Send the Crown

Read the ten as a tally and the count is smaller than it looks. Six of the ten have readable bodies; one is a marketplace category page that returned an error with four words of text; two are image-aggregation shells carrying 75 and 46 words on the load counted here — the same shell read three hours earlier gave 76, so a shell's total is a load rather than a fact; and one returns a status that says it exists but gives 15 words of body, so its advice is only visible as a search-result line. Of the six that can be read, two send the crown up and two send it down, one of those two also sends it up in the next sentence, and one names wide brims without any statement about height at all.

The page at the top of the list is the strangest row in that count. Its body carries no argument about a round face — 1,429 words with the two words for crown and brim present as menu labels — while the sentence usually quoted as its answer, about adding height and angles, is text a search engine wrote in its own summary of the page rather than text on it. A summary line was checked against the page it came from and is not on it.

Google's own generated answer for this question also sits above these ten and says add height, structure and angular lines. It is excluded from the count below because it is a composite written from four of the pages under it, and it carries no measurement either.

What the count is worth: a split of two against two, with one of the four holding both positions in one block of text, is not a disagreement you can settle by reading more pages. It is a sign that the quantity being argued about has no scale attached to it anywhere in the set — which is what the next two columns of the table below record.

Word and byte counts are of text after markup, scripts and comments were stripped, counted by whitespace tokens, so an aggregation shell reports a small number for being a shell and a catalogue-heavy page reports a large one for its menus. The two-word column counts the words for a crown and for a brim, singular and plural, with a separate note for how many of those occurrences sit in shop navigation rather than in body text. Crown and brim instruction columns are what each page's own heading block contains, taken by cutting the documents on their `<h1>`–`<h4>` headings and then reading sentences; cutting them by sentence adjacency gets two sentences wrong on one page, whose warnings about tall crowns belong to sections about other outlines. The two centimetre rows are stated as head-size selectors because that is what those numbers are: a ring that has to close around a skull, which does not follow a face shape. Six of the ten carry no readable publication date, two are shells whose only dates belong to a story list, one is an error page, and one shows a template stamp reused across several shops; those are recorded in the dated table below rather than quietly dropped.

One Page, Two Opposite Verbs

The single most useful thing the readable set contains is on one page, and it is not a style opinion. In the block that page writes for a round outline, a low crown with a wide brim is described as what will elongate the face and give it a more angular appearance. Two blocks away, the identical pair of instructions — low crown, wide brim — is what the same page says can help shorten a long face. Same two quantities, opposite verbs, one document.

That is not a typo and it is not fixed by choosing the other sentence. A low, wide shape lengthens a face that is read as short and rounds the reading of a face that is read as long, and which of those a hat does depends on the starting figure — the very ladder in the first table. A page that has no starting figure cannot say which verb it is using, so it prints both and calls it guidance. The next sentence in that same block recommends a tall top hat and a high crown without noticing it had just recommended a low crown.

The same page is also the only one in the ten that comes close to printing a measurement of a face. Where it tries, the visible markup reads `1. WIDE FOREHEAD: Widest section, broad top. Measures width X.` — the sentence is there, the unit is not, and the figure is a placeholder that shipped unreplaced. It was counted twice in the document because the guide text is embedded in the page in two forms, and both copies carry the same `X`.

A last reading from that page, because it changes how much weight its section names can bear. The page has six headings naming outlines, and the blocks they open are mostly empty: the one under the heading naming a long face holds 142 bytes and no sentences, while the long-face sentences sit inside the blocks whose headings name a round face and a heart. So the section a sentence is filed under is not reliable evidence of what the sentence is about on this page, and everything above is attributed by the subject a sentence names about itself.

For anyone deciding from this page alone: the disagreement you can act on is not which adjective a shop uses, but that no page in the set tells you how far the crown stands, which is the only part of the sentence that transfers to your head.

The Two Pages That Write to Men and to Women Separately

Two of the six readable pages are written to one reader rather than to every outline at once: the menswear guide, the longest readable page in the set, and the sun-hat guide, the shortest of those six. Their instructions differ in kind, not just in direction, which is worth noting because each is offered as the complete answer.

Ask the question as hats for round faces male and the menswear guide is the position in this set written for a man. It frames the goal as adding length and definition. It asks for height and angles or structure, credits a fedora's high crown and slanted brim with lengthening a soft outline, warns that a narrow brim reads back as roundness and asks for a brim between narrow and wide instead, and bans two things outright: a wide rounded crown of the bowler type, because a crown shaped like the face repeats it, and a bucket hat, because a brim sloping downward accentuates roundness.

Ask it as hats for round faces female and the sun-hat guide is the one written for a woman. It puts the whole answer on two quantities at once: a high crown with a shorter brim. It names the trilby and the soft bucket for that job, and it opens by telling readers to avoid circular, floppy shapes, because those mimic the outline they were meant to change.

Put those two paragraphs side by side and the set contains one hat recommended by each and banned by the other: the same soft bucket shape that the women's guide recommends for a round face, the men's guide bans for a round face. That reversal is not a matter of taste either — it is what happens when one page treats a brim as a length-changing object and the other treats it as a shape-changing one, and neither page has a number for the brim or the crown to argue about.

Those two are also the only positions in the readable set that name shapes to stay away from, each in its own round-outline section: one bans circular and floppy styles, the other a wide rounded crown and a downward-sloping brim. The other four readable pages stay at what to look for. Nobody in the set prints what either ban is worth in millimetres, which is the subject of the next three sections.

A Cowboy Hat and a Beanie, Counted Rather Than Recommended

Ask for the best cowboy hat for round face and you are asking a question the ten have not answered, and the count is the evidence. Across the visible text of all ten positions, the word for a cowboy hat appears 15 times: five on the page at rank one and six on the retailer's guide, where ten of those eleven are shop navigation — a category label reading Western and Cowboy Hats — and one is a sentence whose own subject is shortening a long face. Four more appearances are on a sun-hat page, all four inside one product name in a list of styles for a round face. The longest readable page in the set does not contain the word at all. Nothing in the set gives a cowboy hat a brim angle, a crown depth or a height for a round face, because nothing in the set discusses one as a shape for a round face.

The knit cap fares differently and the difference is the point. Ask it as a beanie for round face and two pages do put that object and that outline in the same block of text, and what they give is not a size. The hatter's collection page tells you one can be worn pulled low. The retailer's guide prints two verdicts in the round-outline block: a cap pulled low over the face will elongate it, and — in a sentence whose stated subject is a long face — avoid these for that reason. The menswear guide uses the word exactly once in 3,680 words, and it is in the section telling a long face what to avoid, where a slouchy one is said to accentuate elongation.

A fold on a knit cap, and the depth a crown leaves of a forehead once the cap is on, are quantities the first table takes directly: the excess above the head divided by your own cheekbone width. No position in the set has written that division for either object. What they hand over in its place is shop navigation, a product name, and a sentence about how to wear the thing rather than how large it is.

Both names carry a height claim in ordinary speech, which is why the count belongs to a decision and not to a catalogue. Thirty uses of the two names across the ten, fifteen of each, and the arithmetic content is seven sentences: one about a cowboy hat, attached to a long face, and six about a knit cap, three of them inside text written for a round outline and none of them giving a figure. What those two names can therefore settle for you is not which one flatters, but whether the object you buy has a crown depth and a brim width written anywhere on it — and in this set, neither object does.

These are counts of word forms, not of advice. A category label is counted where the page prints it as navigation text, which is why the totals are larger than the number of claims anyone could quote. Attribution to an outline follows the block a page's own headings open, with the one exception recorded in the section above about displaced headings — where the sentence naming its subject wins over the heading it sits under. The page at rank six embeds its guide text twice, once as the text a person reads on the page and once inside a machine-readable copy of it — the version written for a search engine rather than for eyes — and the counts printed here are from the readable side only, so they are lower than the same strings counted across the whole document.

The Dated Record, Sorted by the Date Each Source Carries

The table below puts what can be dated in date order and leaves what cannot at the bottom with its word count, so the hole in the record stays somewhere a reader can see it rather than being smoothed out of the page. Four of the ten positions carry a date that belongs to the article they are standing on; six carry none. The top row is not one of the ten. It is a printed millinery handbook from 1847, read in transcription, and it is the oldest instruction found anywhere that addresses this outline. Its date is an imprint year with no day beside it, which is why the span it opens is stated as a range.

The range first. From 1847 to 20 January 2026 is 65,034 to 65,398 days depending on which day of that year the book left the press — 178.05 to 179.05 years on the same divisor this site used for its 1915 rows. The four dated positions span 917 days among themselves, 2.51 years, from 18 July 2023 to the January 2026 row; inside that, 47 days or 0.13 years separate the 29 January 2025 row from the 17 March one, 309 days or 0.85 years separate that March row from the newest dated row, and 356 days or 0.97 years separate 29 January 2025 from 20 January 2026. Across the whole 178 years the count of millimetres attached to a crown or a brim for a round outline is zero. The ten use the two words 123 times, and the only centimetre figures anywhere in the set are two head-size pickers, 54 to 65 cm and 56 to 61 cm, which price a ring around a skull rather than a shape above a face.

A date is used only where the source labels it as the publication date of the article it is standing on, in machine-readable form or printed in the text. A copyright year, a date belonging to a story list, a template stamp reused across several shops, and a bare date string with no field name beside it are recorded inside the row rather than used as the sort key — which is why five of the six undated rows do have a date string somewhere on the page and are still listed as undated. The word counts are of text after markup, scripts and comments were stripped, so a shell reports a small number for being a shell and a catalogue reports a large one for its menus; the two shells are given separate rows because their counts came from separate loads. The 1847 row is described in this page's words rather than quoted, because it was read through a transcription and a transcription is not clean, and because no licence for the text is recorded with it; it is dated by imprint year alone, which is why the span is a range. One row of the ten is absent from the instruction column entirely: the page that returned an error.

What Moved in Those 178 Years, and What Did Not

The oldest row of that table and the newest row of it are not the same kind of document, and the difference is worth naming exactly because it is not progress. The 1847 handbook writes about the head-dress: it asks for height without breadth, and for a reader of short stature it raises that height again and moves the whole thing to one side. The January 2026 position writes about the reader: it classifies an outline and answers with a style name, in two tables that sort one column of faces against one column of hats. What changed is the subject of the sentence. What did not change is that neither sentence contains a figure.

The number-free passage in 1847 is not a pre-modern oversight either. That book is not illiterate about measurement — it writes inches 31 times when it instructs somebody to cut wire and cloth, and nothing at all when it instructs somebody to choose. The ten carry the same division of labour 178 years later: 123 uses of the two words, zero millimetres, and twelve centimetre figures apiece on the two positions that print numbers at all, every one of them a head size chosen from a menu rather than a shape measured against a face.

One thing did get dropped on the way. The 1847 instruction places the object — higher for a short reader, and set to one side rather than square on the head. In the modern set the only placement verb given to this outline belongs to a knit cap pulled low, and every other verb is about the object itself: taller, lower, wider, narrower. Those are the quantities a catalogue prints. Where a thing sits on a head is not a column anybody can filter, which may be the plainest available reason the same hole survived 178 years of advice.

The Only Way Back Down Runs Over the Forehead

A crown can only add, so if the reading is to go down, something has to lose height, and the only place a hat can take it off is the forehead. One position in the set does write the step — in the section it addresses to a long outline, where it asks for a brim cut across the forehead so the face reads shorter. That is the same numerator with the sign reversed: the height of forehead a brim covers, divided by your own cheekbone width. It is the only move in this subject that goes backwards, and no page in the set states it as an instruction for a round outline, where it is the only way down.

The brim has to be bought as a width, and a width has a floor that does not belong to the face. On this site's ruler a forehead is a ratio times your cheekbone width, and the published forehead ratios for the seven outlines run from 0.80 at the narrow end to 1.05 at the wide, which are published centres rather than ends invented for a table. The spread published with that ratio is 0.08 — at the 136 mm working line, 5.5 mm per side — and the whole 0.25 span of the band is 34.0 mm across the face, 17.0 mm or 0.67 in per side. A brim wide enough to clear a forehead at the row where a forehead equals the cheekbone width, by 10 mm on each side, comes to 156.0 mm across. The 10 mm is the site's own allowance, defined on the published brim table rather than measured, and its sensitivity is published beside it: 5 mm per side gives 146.0 mm, 15 mm gives 166.0 mm.

Then pay the head first. In the 1915 worked example the head's own oval measures 177.8 mm across at the level a brim is attached, and 156.0 mm is 21.8 mm narrower than that. For a head at the example size, the covering brim is simply not available: the arithmetic is describing where an edge would fall if you could buy it. That is the sentence the two guides written for one reader each are missing when one asks for a shorter brim and the other for a wider one. A brim verdict issued without your forehead width and your head width is not a verdict; it is a preference stated as an instruction.

Three of the Five Numbers Sit Below Anything a Hat Reaches

This site publishes five readings for a face: face length divided by cheekbone width, forehead width divided by the same, jaw width divided by the same, the angle at the jaw corner, and how much narrower the chin is than the jaw. A hat enters two of them. The crown stands above the hairline and changes where the top of the outline reads as being; the brim hangs over the forehead and can hide part of it. The other three are taken on the lower face, below the level anything in this subject reaches, and no crown height or brim width on the ladder above moves one of them by a hundredth.

That is where the roundness question actually lives, and it is why the ten point a reader toward a place a hat cannot reach. A crown is scored against the length division; the reading that carries a jaw's narrowing and the one that carries the angle of it are not in the same arithmetic, and never will be while the hat stays on the head. Two of the ten make the lower face their promise anyway — one says a crown will give a round outline a more angular appearance, the other that structure adds sharpness to softer features.

This page prints no figure for those three readings and points instead. The two outlines that share the bottom of the published length band are told apart elsewhere on this site, in a guide written for exactly that comparison, and the pillar page for this outline carries the five readings with their bands. The angle at the jaw corner does not separate those two outlines on this site's data, and the published jaw figures are taken with a detector's landmarks — the fixed points the site's photo tool puts on a face — rather than a soft tape, so they belong on a different ruler from every cell above and are not mixed with them here. For anyone holding a hat in a shop the consequence is small and firm: a hat can move two of the five readings, and the roundness everybody argues about lives in the three it cannot reach.

The Order to Run It In

Both operands of the division on this page are yours rather than the site's: a cheekbone width taken with a soft tape, and a crown's lift read off a ladder with a ruler. The published figures here are the band, the spread and the three working widths — not your head. Take the width every cell on this page divides by ↓

Everything on this page is two readings and one division, and none of it needs a shop's help. The order below is the order that produces a number you can check, rather than a number you have to trust.

• Take the width across the cheekbones in millimetres, the same way this site's measuring guide takes it, and write it down — every cell above is divided by that figure, and it is the only one of the two readings your face has to supply.
• Put the hat on and stand it the way you would wear it. Measure how far the outside top of the crown stands above the top of your head, in millimetres. That lift is the only new quantity in this subject, and it is positive or zero; there is no crown standing below the top of a skull.
• Divide the lift by your width. At the 136 mm working line, 25 mm of lift is 0.184 and 45 mm is 0.331; at your own width, do the division instead of reading the row.
• Add the result to the figure this site publishes for a round outline, 1.05 on the tape scale, and compare the sum with 1.65 — the published top of the range, which is where the site's longest outline sits. That comparison is the entire decision about height, and it is a distance rather than a verdict.
• If you want the reading to go down instead of up, do the brim arithmetic before you buy anything, and pay the head first: a covering width has to exceed the width of the skull it sits on, or the edge you are describing does not exist to be bought.

What This Page Does Not Prove

Ten positions read on one day, from one network exit, in one country setting. Rankings move, and one of the two aggregation shells here changed its own word count between two loads three hours apart — that is stated because it is the only measurement on this page which disagreed with itself, and a reader should know which numbers here were stable. What is claimed from the set is narrow: no position in it printed a millimetre for a crown or a brim, and four of the ten carry a date belonging to the article they stand on.

The 1847 row is a transcription, and no sentence of it is reproduced here. It is evidence that an answer in terms of shape and placement existed, not evidence that it worked, and its author's name is withheld because the transcription of the title page is damaged in the way that would make a printed name unreliable.

The ladder ranks nothing. 1.05 is the floor of a published classification band, used here as a ruler because the ladder has to start somewhere; it is not a diagnosis, not a target, and not a claim that a longer outline is preferable to a rounder one. This site publishes no measure of preference of any kind. Its five readings are taken on a bare face, and three of those five are below the reach of anything with a crown.

There is no fit claim here either. The ring a hat closes around is a different quantity from both divisions on this page, and no head-size range is printed here — the site publishes its own on the tool named above, and a standard oval drawn by a 1915 course manual is a different object from that band, so the two are kept apart rather than averaged. And nothing above was measured on a person: a crown standing off a head is a quantity this rule set — the site's own published list of checks, rows and figures — has never taken on anybody, which is why every grid here is arithmetic with published operands and your own two numbers in it, built to be redone rather than believed.

Questions this page answers