Before a reader processes the letters of a word, they register its outline. Reading research calls this outline the bouma — the silhouette formed by which letters rise, which drop, and which sit in the middle band.
That middle band is the x-height: the height of a lowercase letter with no part rising or falling, like "a", "e", "o", "x". Letters that rise above it have an ascender — b, d, f, h, k, l, t. Letters that drop below the baseline have a descender — g, j, p, q, y. The pattern of peaks and valleys is what makes one word's shape distinct from another's at a glance.
A word with both ascenders and descenders has a jagged, recognisable profile. A word built entirely from x-height letters is a flat rectangular block — clean, but visually anonymous. "Spotify", "Google", and "Pfizer" each carry at least one peak and one valley. "Aeon", "noun", and "casa" do not. Read in running text, the flat ones are slower to find and easier to confuse.
No industry takes this more seriously than pharmaceuticals, because there the cost of a misread name is measured in patient harm. The risk has a regulated name: look-alike sound-alike confusion. The FDA runs a tool called POCA — Phonetic and Orthographic Computer Analysis — specifically to screen proposed drug names for visual and phonetic collision with existing ones, and its Name Differentiation Project tracked real confusion incidents across documented drug pairs.
This is why drug names are visually engineered, not just sounded out. Brand Institute's case work on the name "Tepezza" is explicit about the construction: three syllables, double consonants for stability, a crossed letter, and a descender on the "p". Every one of those choices adds a distinguishing feature to the written silhouette so the name survives being scrawled on a prescription pad and read under pressure.
The lesson generalises. Most names will never be handwritten by a hurried clinician — but every name is read quickly, in a list, next to competitors, at small sizes. The pharmacy is just the place where the consequences are visible.
Consider "aeon". Phonetically it is pleasant. Semantically it is evocative. But written in lowercase it is four x-height letters in a row — a small, featureless bar. Set it in a sentence next to other words and the eye slides past it. Set it as an app label among a grid of icons and it has no distinguishing outline to lock onto.
This is not an argument against the word. It is an argument for *knowing* the property before you commit, because there are cheap fixes. Capitalising the first letter adds a peak — "Aeon" already reads better than "aeon". Length helps too: a longer word made of x-height letters at least has horizontal mass to anchor on, where a short one has almost nothing. The all-flat penalty is real but it is sharpest for short, lowercase, peak-free words.
Orthography — how a word looks — is a measurable property of a name, separate from how it sounds. A name can score well on phonetics and still be visually weak. The two are independent, and a serious naming process checks both.
Outside regulated industries, silhouette still pays. App icons and favicons render the name at sizes where the bouma is doing most of the recognition work. Editorial and book-title typography depends on word shape for fast scanning. A wordmark with balanced vertical mass — some rise, some drop — simply holds up better across the contexts a brand actually lives in.
There is also a class of letter combinations that look fine in isolation but blur together at speed or at small sizes — "rn" reading as "m", "cl" as "d", "vv" as "w". These are not deal-breakers for most brands, but they are worth knowing about, especially anywhere the name will be set small or written by hand. processing fluency research is consistent on the underlying point: the easier a name is to take in, the more familiar and trustworthy it feels, even on first sight. Visual ease is part of that fluency, not separate from it.
Visual legibility enters the score as a component of the Language pillar — the pillar that already measures the *written* and structural properties of a name, as distinct from how it sounds. That placement is deliberate: letter shapes are a property of the grapheme, not the phoneme, so they belong with the language structure analysis, not the phonetic one.
The component is intentionally small. The Language pillar fine-tunes the score rather than dominating it, and visual legibility sits in a tight range — a real but modest lift for a name with a strong, balanced silhouette, a small penalty for a short all-flat one, and a flag either way so the user sees the reasoning. Most names land near neutral, which is correct: silhouette is a tiebreaker, not the headline.
It is also font-agnostic and deterministic. The analysis works from the letters themselves, not from any particular typeface, so the same name always produces the same legibility profile, with the ascenders, descenders, and flagged combinations listed explicitly. As with every PhonoPair factor, you can see exactly what moved the number and why.
1. Write the name down in lowercase and squint. If it is a flat bar with no peaks or valleys, you have a silhouette problem — regardless of how it sounds.
2. Capitalisation is a free fix. A leading capital adds a peak to almost any word. If the name will always be displayed capitalised, score it that way.
3. Short and flat is the worst case. A long word made of x-height letters at least has mass. A short one has nothing for the eye to grab.
4. Check it at the size it will live at. Favicon, app label, footer, a list of competitors. The bouma matters most exactly where the type is smallest.
5. Phonetics and legibility are independent. A name can sound excellent and look weak. Verify both before you commit — they are different tests, and a strong name passes each on its own terms.
Check how your name candidates sound and read