Corners and Radius
A radius scale is a list of numbers, and a list of numbers never says which corner belongs on a card. Three sentences do: what is structure, what is an object on the page, and what can be pressed.
A list of numbers is not a set of decisions
Your brand book may well have four corner values in it, something like 2, 4, 8 and 16, set out in a row with a name under each. That is a good thing to have and it settles nothing. It tells you which corners are allowed. It does not tell you which one a card takes, and it never will, because that answer is not in the list.
The list is your vocabulary. What is missing is the sentences, and there are three of them.
Three rather than one, because a corner is not decoration. It is one of the few things on a screen that tells a reader what kind of thing they are looking at before they have read a word of it. A hard corner reads as structure: this is part of the page rather than something placed on it. A softened corner reads as an object, distinct and self-contained and set down on top of the layout. A fully rounded one reads as pressable, and it reads that way so reliably that a rounded label with nothing behind it is a small lie a reader falls for every time.
So the question this row asks is not what your corner values are. You have those. It is what each of them means, three times over.
The three, and what each is answering
| The corner | What it covers | What it says |
|---|---|---|
| Structure | Page sections, dividers, the blocks a page is built out of | This is the page itself, not something sitting on it |
| Containers | Cards and panels: things that are distinct and self-contained | This is one object, and it ends here |
| What can be pressed | Buttons, chips, the things a reader acts on | You can do something to this |
They are asked together, on one row, with an example of each drawn in your own colours, because none of the three means very much on its own. A container corner worth having is one that is visibly not the section behind it. A pressable corner worth having is one nobody mistakes for a card. Read down the three rather than at any one of them, and the answers arrive faster than they would in isolation.
Picking the values
If your book states them, transcribe and move on. Most books do not state them. What most books have instead is the scale plus a set of finished screens, and that is enough: look at what your own product or your own templates already do and read the three answers off them. You are recovering a decision that has already been made rather than making one.
The arrangement almost every brand lands on runs in one direction: the sharpest step for structure, something in the middle for containers, the largest or fully round for what can be pressed. Treat that as where brands tend to end up rather than as a rule, because the memorable ones are often the ones that break it. Sharp buttons on a soft page is a deliberate and legible thing to be. What is not legible is arriving there by accident.
If you are deciding rather than transcribing, two tests will get you there faster than any reasoning about the numbers. Put a card and a button side by side at the values you are considering, take the words out, and ask whether you can still tell which one is pressable. Then set the card on a section and ask whether it still reads as a separate object. If both hold, you have your three, and the particular numbers matter much less than those two answers.
When two of them come out the same
That is a real answer, not a mistake, and nothing on the card treats it as one.
Plenty of brands have one corner and mean it. An editorial brand may be square on everything it draws, a consumer brand round on everything, and in both cases the three sentences come out identical and are still three sentences. All three count as ruled, because they are.
What is worth being clear about is what you have said when you do it. Ruling two corners to the same step is not a repetition, it is a claim: in this brand the corner is not what carries that distinction, and something else does the work instead. Your pressable things are telling a reader they are pressable by their colour, their weight or their position, and you have decided that. It is a decision, and it is a far better thing to have made than to have drifted into.
The check is one question. If a card came back with the button's corner on it, would you send it back? If you would not, they genuinely are the same and you can rule them the same. If you would, they are not the same and you have not finished, whatever the two numbers currently say.
The scale, and this row
They live on separate cards and neither does anything useful without the other.
The scale sits above and holds the numbers. This row holds the meanings, and each of the three is stored as a pointer at one of the steps on that scale rather than as a value of its own. There is no box on this row to type a length into, and that is deliberate: a number typed here would be a second copy of your scale, and it would stop agreeing with the original the first time somebody edited one of them.
The good consequence is that one edit moves everything. Change what a step is worth and every corner standing on it moves with it, and all three sentences stay true without anybody rewriting them. Nobody has to go back and restate what a card's corner means because the number behind it changed.
It also means the scale has to exist first. A brand with no length values anywhere has nothing for these three to point at, and rather than offering a box to type one into, the form says so and sends you up to where a scale gets added.
The other direction is the state most brands arrive in: a complete set of corner values and not one of the three ruled. That is reported as three findings rather than one, because they are three separate questions, and it is the exact case this row exists for. Numbers, with nothing saying what they mean.
One more thing worth knowing: the three meanings are fixed, and their names are not. If your brand has its own word for the corner a card takes, you can rule it under that word and the mapping is kept for you, so nothing downstream ever has to learn our vocabulary in order to read yours.
Fully round
Designers reach for an enormous number here, and they are usually right to, but rarely for a reason they could state.
The habit comes from drawing tools. Most of them have no setting that means "however round this needs to be": there is a slider that stops when the corners meet, or a field where you type something absurd and the tool quietly holds it down to whatever fits. So a very large number became the way to say fully round, and the habit gets carried somewhere else without anyone being certain what it will do there.
Here is what it does here. The number is kept exactly as you gave it and passed on unchanged. Nothing converts it, caps it or rewrites it on the way.
What gets drawn is a corner that rounds until the two ends meet and then stops. On a wide button that is a capsule with semicircular ends. On a square chip the same value is a circle. Nothing overflows and nothing distorts, and any value at or past half the shorter side draws the identical picture, which is precisely why a very large one is the safe way to say it: it stays round enough at every size the thing turns out to be, including sizes nobody has drawn yet. If you are building a scale here from scratch, the last step it offers means fully round and is stored as a very large length, around nine thousand pixels, for that reason.
The one thing that is not the same is a percentage. Fifty per cent sounds as though it ought to mean fully round and it does not, because a percentage measures each corner against the box it is drawn on rather than against a fixed length. On a square that gives you a circle, which is fine. On a wide button it gives you an ellipse: the sides bow out and you get a lens rather than a capsule. If a pill is what you are after, use a large length.
What an unruled corner costs
Nothing breaks, which is the difficulty with all of these.
All three are raised on the card at the middle of the three priorities, as something this brand cannot make yet, below anything that would make a piece of work confidently wrong. Everything you ask for still gets made. Your corner values still travel. The numbers are all there.
What is not there is the assignment, and the drawings on this page are honest about that in a way worth looking at once. Where a corner is unruled the preview does not draw a square one and present it as your answer. It declines to state a corner at all and marks the spot instead, because a square corner and an unstated corner are the same picture the moment they reach a screen, and only one of them is a decision.
Out in the work it lands the way an unstated design intent lands. Whatever is building picks, the pick is always plausible, and nobody queries a plausible corner.
This one has its own particular way of going wrong, though, and it is worth naming. The corner is a signal about what kind of thing something is, and when nobody has assigned it the same corner tends to turn up on a section, on a card and on a button. Each screen looks fine. What has quietly gone is the cue telling a reader which of the three they could press, and readers never report that. They just get slower.
The other rules
The corner is the fork a design intent is usually explained with, so that is the one to read next: two corners are both legal, every rule you have stated is silent, and one sentence picks. Grid and margins sets out the idea underneath the three rules below this row, that what gets stored is a relationship and never a measurement. Brand components is the card further down, where a card or a button gets its own corner field, pointing at a step of the same scale these three point at.