Sheet 05Revised 20 Aug 2026

Anatomy of a typeface: the parts, and what each one decides

The parts of a letter have names because the names are load-bearing. A face is chosen, rejected, paired or set at a particular size on the strength of four or five measurements, and none of the arguments can be had without the vocabulary. Every figure below is ruled from the font binary this page serves.

A clear French curve template and a fine technical drafting pen on pale grid paper, with a violet pencil.

No. 408A curve template does one job: it makes a shape repeatable. So does a name for a part of a letter.

The five lines every letter is built on

A roman letter is drawn against five horizontal references, and almost everything a reader perceives as the character of a face comes from where those five sit relative to one another.

Baseline: The line the letters sit on. Round letters overshoot it very slightly, because a curve that stops exactly on the line looks as though it stops above it.
x-height: The height of the lowercase x, and therefore of most of the lowercase. This is the single number that determines how large a face looks at a given point size.
Cap height: The height of the capitals. Usually a little below the ascender line, which is why a capital next to a lowercase l rarely reaches the same height.
Ascender: How far b, d, h, k and l rise above the x-height.
Descender: How far g, j, p, q and y drop below the baseline. Short descenders allow tight leading; long ones do not, and a face with long descenders set tight will collide with itself.

Fig. 3 · Ruled from the font binary this page serves, at a 216 px setting: ascender ink 158, cap height 151, x-height 114, descender ink 43, and an advance of 263 for the pair. The x-height is 75.1 % of the cap height, against 66.7 % for the text face set below it and 76.9 % for the monospaced face used in these captions, which is why a display face this open survives being scattered and still reads.

Measured, not estimated

The figures in the drawing above are not approximations taken from a screenshot. They are read out of righteous-400-latin.woff2, the file this page serves, and scaled to a 216 pixel setting: unitsPerEm 2048, cap height 1434 units, x-height 1077 units, the ascender taken as the ink extent of b, d, h, k and l, and the descender as the ink extent of g, j, p, q and y.

The same measurement run over the three faces this site sets its text in gives the comparison that matters more than any single number:

Face	Role	x-height, % of em	x-height, % of cap
Righteous	Display	52.6	75.1
Kumbh Sans	Text	48.8	66.7
Syne Mono	Figures and captions	50.0	76.9

The three faces differ by under four percentage points of the em and by more than eight points of the cap height. That second figure is the one that decides how a heading sits over a paragraph, and it is the reason every heading size on this site was set against the text face rather than chosen.

The parts of the stroke

Stem: The main vertical stroke. In most seriffed faces the stem is the thickest part of the letter and everything else is measured against it.
Bowl: The closed curve of b, d, o, p. Its width against the stem weight is most of what makes a face feel narrow or open.
Counter: The enclosed space inside a bowl. Counters close up first when a face is set below its intended size, which is precisely how a display face fails in body copy.
Aperture: The opening of a partially closed letter such as c, e, s or a. Open apertures are the strongest single predictor of legibility at small size and low resolution.
Terminal: How a stroke ends where there is no serif: cut flat, angled, or finished with a small ball. A great deal of a face's personality lives here and nowhere else.
Shoulder: The curved stroke leaving a stem, as in h, m and n.
Spur, ear, tail, leg, link: The small identifying features: the spur on a capital G, the ear on a lowercase g, the tail on a Q, the leg on an R or K, the link joining the two bowls of a double-storey g. These are what distinguish revivals of the same historical face from one another.

Axis and contrast, the two that name the family

Draw a line through the thinnest points of a lowercase o. That is the axis, and it records the angle of the tool the letter descends from. Angled axis means a pen; vertical axis means an engraver or a compass.

Contrast is the ratio between the thinnest and the thickest part of the same letter. Together the two are almost enough to place a face in its family, which is what the classification page is built on.

The space, not the shape

Three terms cover the space, and the space is what actually gets adjusted in practice.

Sidebearing: The space designed into the left and right of each glyph. It belongs to the font, not to the setting, and it is why some faces look loose at every tracking value.
Kerning: A pair-specific adjustment (AV, To, Ye) stored in the font and applied automatically. Turning it off is visible immediately in capitals and almost invisible in lowercase.
Tracking: A uniform adjustment applied to a whole run by whoever is setting it. Large sizes generally want negative tracking; small sizes and all-capital settings generally want positive.

Leading, the distance from one baseline to the next, is the fourth, and it belongs to the line length and the x-height together rather than to the face. The pairing bench exposes all four so the relationship can be seen rather than described.

Why the names earn their keep

Every one of these terms exists because a decision hangs on it. "This face is too small" is not actionable. "This face has a low x-height, so it needs a point size two steps up to match the one beside it" is. "That heading looks wrong" is not actionable. "The counters are closing at this size because it is a display face set below its range" is.

The vocabulary is small, around twenty words, and it converts an argument about taste into an argument about measurement, which is an argument that can be finished.

Questions about the letterform

Queries

What is the difference between x-height and cap height?

The x-height is the height of the lowercase x and therefore of most lowercase letters; the cap height is the height of the capitals. The ratio between them varies widely: across the three faces measured on this page it runs from 66.7 % to 76.9 %, and that ratio, not the point size, is what makes one face look larger than another set at the same size.

Why do round letters extend past the baseline?

Because of how the eye reads a curve. A circle whose lowest point stops exactly on the baseline appears to float above it, so type designers push round letters very slightly below and above the lines. The correction is called overshoot and it is deliberate in essentially every face.

What is an aperture and why does it matter more than it sounds?

It is the opening of a letter that is nearly closed, such as c, e, s or a. Open apertures keep those letters distinguishable when the image is small, blurred or low contrast, which describes most real reading conditions. It is the feature most closely linked to legibility research.

Where do these measurements come from?

From the font binaries this site serves, read with a font-tooling script and scaled to the stated setting. The script is kept with the build so the figures can be regenerated if a face is ever changed; a figure whose numbers no longer match its own font is worse than no figure.