Inverted Triangle Face Shape: The Measurements Behind the Label

Not an eighth shape: what this page is for

The outline this label describes is wide across the forehead and the temples, narrows through the cheekbones, and ends at a chin that comes to a point rather than a curve. Every clause of that is also a true description of a heart outline. That is the whole difficulty, and it is why this page exists as a distinction instead of a profile: no measurement separating the two has ever been published, so there is nothing to classify against.

This site lists seven shapes. An eighth would need a threshold — a forehead ratio above which the label applies and below which it does not — and no source offers one that could be checked against a tape. Printing a number we could not defend would be worse than saying plainly that the category is absorbed into a neighbour, so that is what we say, here and on the about page where the entire rule set is published.

The practical consequence for you is small and worth stating before anything else: if a guide has told you that you are an inverted triangle, the numbers behind that label are the heart column of the table on this page. Nothing about your face changes because two naming traditions disagree about it, and every recommendation this site makes for a heart outline already applies to you unchanged.

The second consequence is not small. Because the label sits inside heart rather than beside it, the shape worth distinguishing from yours is not heart at all — it is triangle, where the widest point is the jaw and the narrowest is the forehead. The two are the same triangle pointing in opposite directions, and three of the five checks in the table below separate them with no band in common. That separation, printed as numbers, is what this page adds.

Two criteria for one label, and which of them can be measured

The guides that use this label divide it from heart in one of two ways, and the two do not agree with each other. The first reads the hairline: a centre point in it means heart, a straight or rounded hairline means inverted triangle. The second reads the width gradient: how the forehead compares with the cheekbones, how the cheekbones compare with the jaw, and how sharply the chin narrows beneath them.

Neither is a mistake. They are different criteria, and a face can satisfy one and fail the other — a straight hairline over a full upward gradient is common, and so is a hairline with a centre point over an outline whose widest point is the cheekbones. Two guides using different criteria will both report confidently and will disagree, which is exactly what happens across the pages answering this query.

This site reads the second criterion, for a reason that is structural rather than a preference. A hairline is where hair grows: a fringe moves it, recession moves it, and holding the hair back moves what a tape can reach. The five features the classifier reads are all bone-and-tape quantities — face length against cheekbone width, forehead against cheekbone width, jaw against cheekbone width, the angle at the jaw corner, and chin width against jaw width. A criterion resting on a hairline cannot be taken twice with the same result, so it cannot become a threshold, and a category with no threshold cannot be classified against.

If the guide you read used the hairline criterion and landed on inverted triangle, there is no conflict for you to resolve. The label tells you which tradition measured you; the table below tells you what your face measures. Read the heart column either way.

The table: five checks, two columns

Rows are the five checks this site classifies on. Columns are the two shapes this label sits between: heart, which is where this site puts an inverted triangle, and triangle, the outline running the other way. The other five shapes are not printed here — the full seven-column version lives on this site's rectangle face shape page, and this page carries two columns because only these two are in dispute over the label.

The four notes under this table are part of the reading, not a disclaimer attached to it afterwards.

• Bands overlap, and on this pair two of the five do. The length bands share most of their span and the angle bands share almost all of theirs, so neither row settles anything on its own. The other three rows share no span at all — that is the finding the next-but-one section works through.
• Neither column is measured. Heart had one photograph in the calibration set and triangle had none, so both centres are interpolated along the published ordering of the seven shapes rather than fitted to real faces. The about page prints the whole method and both of its limits.
• The jaw angle is the weakest of the five checks in our own data. It could not separate round from square — two groups that measured within a degree of each other, though the published literature places the two shapes far apart — so treat the angle row as a tie-breaker and the ratios as the evidence.
• The detector's raw readings sit on a different scale from a tape measure and are converted before they are shown anywhere on this site. If you measured yourself with a tape, no conversion applies to you: you are already on the scale this table is printed in.

Taking four of the five numbers at home

Four lengths, taken with a soft tape over bare skin, hair held back off the forehead, head level. No tape? The string-and-ruler method described on this site gets the same quotients out of four marks. Face length runs from the centre of the hairline straight down the midline to the tip of the chin. Cheekbone width runs apex to apex across the highest points of the cheekbones, the tape passing just under the eyes. Forehead width runs across the widest part of the forehead. Jaw width runs from one jaw corner to the other along the outside of the jaw, straight across and not wrapping round under the chin.

Three divisions turn those four lengths into three of the checks: length over cheekbones, forehead over cheekbones, jaw over cheekbones. Units cancel, so centimetres and inches give identical quotients and the table above carries no centimetres at all.

The fourth check, the jaw angle, is not something a tape will hand you. On the calculator path this site estimates it from your three ratios rather than asking you to guess at a protractor, and the page says that it is an estimate. On the photo path it comes from landmarks, in your browser, with nothing uploaded.

The fifth check, chin taper, is the one this site deliberately does not ask you to measure. A tape cannot take a chin width reliably — the placement is ambiguous by a few millimetres and a few millimetres is most of the band — so the calculator has no field for it and the classifier gives it zero weight on that path rather than feeding a made-up number into the result. Only the photo path produces it. Keep that in mind for the next section: of the three checks that separate the two shapes, a tape gives you two.

One measurement is worth getting right twice, because it decides this page. Take the forehead width across the hairline at its widest, with the hair held back. A fringe lying over the forehead lowers what the tape reads without changing your face, and on this outline the forehead is the widest point of the face — under-read it and the gradient you are trying to measure quietly disappears.

Which checks separate the two, and by how much

Read the table by row instead of by column and three of the five checks settle the label outright, because the two bands do not touch. Each figure below is a single subtraction of two endpoints already printed above: the nearer edge of one band taken away from the nearer edge of the other. What is left is ground that belongs to neither column, so a face reading there cannot be both shapes at once.

Forehead against cheekbones leaves 0.09 of the ratio scale belonging to neither. On this outline that single number is the difference between the widest point of the face sitting above the cheekbones and sitting below them — which is why the section above asks you to take that one measurement twice.

Jaw against cheekbones leaves 0.12, the widest of the three gaps, and it carries a reading of its own: the heart column never reaches a jaw as wide as the cheekbones, while the triangle column runs past that point. This is the check with the most daylight in it and, unlike the chin, the one a tape will give you.

Chin taper leaves 0.03. It is real, it is the narrowest of the three by a clear margin, and it is the reading most sensitive to where the tape crosses the chin — which is the same reason the calculator does not ask for it. Treat it as confirmation of the other two rather than as the check that decides.

The remaining two rows separate nothing. A face can sit inside both columns on length, and inside both on angle, which is not a defect in the table but the reason the classifier reports a margin against every shape instead of a single label, and the reason a mixed-shape result exists at all.

One limit travels with all of this, and it is the second note under the table rather than a new caveat. Neither column is measured, so what these three gaps prove is that this site's published rule set draws a definite line between heart and triangle — not that the line has been validated against a sample of faces. We print the gaps because a line you can inspect beats a category nobody can test, and we print the limit beside them for the same reason.

Why the numbers are hard to find

Across the pages that answer this query the pattern is consistent enough to describe without naming anyone. The ones that offer a method offer a formula and then stop short of putting numbers into it. The ones that do print numbers print them for a neighbouring shape rather than this one, and source the count to their own tool instead of to a stated sample. What almost none of them carry is a band you could hold a tape against.

That is not a mystery about the subject. It follows directly from the first section: a category with no agreed threshold has no number to publish, which leaves two honest options — print a threshold you cannot defend, or say that the category is absorbed into a neighbour. We took the second, and then published the neighbour's bands, which is the part that makes this page something other than an opinion.

Two disagreements run through those pages and are worth flagging, because you will meet them if you keep reading elsewhere. Whether a frame carrying its mass along the top rim helps or harms this outline is asserted both ways, sometimes within a single article. Whether this label and heart name one category or two is asserted both ways as well. Neither disagreement can be settled from the pages themselves: the first is a styling question with no measurement attached to it, and the second is the naming question this page has already answered.

Weight and lines against this outline

This section is a styling judgement and not a measurement, and it is marked as one: nothing in it comes from the rule set, and none of it is a correction. The full treatment — hair, frames, facial hair, hats, earrings, necklines and the makeup placements — is on this site's heart face shape guide, because it is the same outline and the same reasoning, and duplicating it here would only give you two places to disagree with. What follows is the part specific to a taper.

The outline runs wide at the top and narrow at the bottom, so the styling question is never which line to introduce but where weight sits. Weight carried at or below cheekbone height travels with the taper; weight carried at the temples or the crown adds to the part of the outline already at its widest. Both produce a look. The second is a recognisable one with a name attached to it, and it is not a mistake.

The same reasoning runs through the small categories, and it is the direction of the taper doing the work in each. A frame carrying its mass on the lower rim puts weight in the half of the outline that has least of it. A drop earring widest at the bottom mirrors the taper instead of repeating it. A neckline holding a horizontal under the jaw stops the point; a narrow deep one continues it downward, which is the same effect a pointed goatee has, seen from further away. Facial hair kept fuller along the jaw than at the chin fills the narrow end without extending the point.

There is no forbidden list on this page and nothing above is something to fix. What is described is where weight lands and what that does to a reading; a preference running the other way is a styling choice with its own recognisable result, not an error to be corrected out of your face.

Frequently asked questions

If an inverted triangle is a heart, why do guides list it separately? Because the two traditions are describing one outline at different resolutions. One names everything that widens upward; the other splits that family by how extreme the taper is and by what the hairline does. Neither split has a published threshold, so the extra category can be preferred but not measured. This site keeps the shorter list and says so rather than absorbing the label quietly.

Does the hairline settle which label applies? Not here, and not by measurement anywhere. A hairline moves with a fringe, with recession, and with how the hair is held back on the day, while the five checks are bone-and-tape quantities that do not. A guide using the hairline is applying a different criterion, not a sharper one — the second section sets the two side by side.

Can a face be an inverted triangle and a triangle at the same time? Under no criterion. The two are the same triangle pointing in opposite directions: this label is widest at the forehead and narrowest at the chin, a triangle outline is widest at the jaw and narrowest at the forehead. Three of the five checks leave no band in common between them, so a face reading clearly on any one of those three cannot read as both.

My length reading sits where the two bands overlap. Now what? Use the other rows. Length and angle share most of their span on this pair, so neither is the check that decides; the forehead, jaw and chin rows are. If those disagree with each other as well, the result page reports both shapes with the margin against each rather than picking one for you.

Is this shape rare? We do not know, and we will not guess. The prevalence figures for face shapes in circulation carry no stated sample and no stated method, so there is nothing to check them against — and this site publishes none, for any of the seven shapes.

What will the detector report if my face is an inverted triangle? Heart. Those seven categories are the only ones it can return, and the result page prints the margin against every one of them, so you can see for yourself how close the reading sat to diamond or to oval instead of taking a single label on trust.

Diamond ends in a point too. Where does that leave this page? The chin is the least useful feature on either outline, since both narrow to a point, so the pair is decided one storey higher. Read the forehead row: on this outline the forehead is the widest thing on the face, and on a diamond the cheekbones are. The diamond column is not printed here because it is not in dispute over this label; the diamond guide on this site carries its own bands.

I measured with a tape and the detector shows different numbers. Which is right? Both, each on its own scale. The detector reads a different set of points from a tape, and its raw length and jaw quotients come out lower; every reading shown anywhere on this site is converted to the tape scale first, which is the scale the table above is printed in. A tape measurement needs no conversion, and the calculator will not ask you for a chin width at all.

Where the numbers come from

Every band in the table is this site's published rule set: the seven prototypes, their standard deviations, the affine mapping between the detector's landmark scale and the tape scale, and the weights the classifier runs on. The whole of it is printed on the about page, together with the two limits that matter most here — that the heart and triangle centres are interpolated rather than measured, and that the jaw angle could not separate round from square in the calibration data.

The three gaps quoted in the fifth section are single subtractions of endpoints printed in the table, and the arithmetic is stated there so you can redo it. No other figure on this page comes from anywhere else and nothing on it is estimated: where a number could not be traced back to the rule set it has been left out, which is why there are no prevalence figures here, no attractiveness ranking and no centimetre targets.

The styling reasoning in the seventh section is ours and is labelled ours. It follows from how the rule set describes the outline rather than from any measured outcome, and it has not been reviewed by a licensed cosmetologist — the about page carries that gap openly instead of inventing a credential to fill it.

What this page does not do: no eighth prototype, no threshold we cannot defend, no rating of your face, and no claim about how this outline photographs. On every path through this site your photo stays on your device.