The number of distinguishable categories has a ceiling

What it is

Every swatch in a categorical palette must satisfy two conditions at once: distinguishable from every other swatch in the palette, and still distinguishable on small-area marks (a 3-pixel dot is far harder than a large chip). Stacked together, these constraints give the usable colour count a hard ceiling — the empirical band is 6 to 12 colours: a carefully crafted palette under ten (uniform hue sampling plus alternating lightness) works; past twelve, colours begin to "borrow" from each other (crowding on the hue wheel) and confusion rates climb steeply. This is not a palette-design skill issue but the colour system's categorical capacity.

Why it happens

Colour discrimination lives in a perceptual space of three dimensions (hue, lightness, saturation), and categorical encoding demands pairwise separation — n categories need all n(n-1)/2 pairs spaced apart, so the required "perceptual distance" grows quadratically while the space stays constant. Only a few hues on the wheel are genuinely far from one another (human languages' basic colour terms also stop near a dozen), and filling the wheel necessarily creates near-neighbour pairs. Small marks cut once more: cone sampling is thin in peripheral vision and on small colour patches, hue differences fade, and only lightness still carries — which is why "separable on the swatch card, indistinguishable on the chart" is the norm, not an accident.

Where it stops holding

The ceiling counts categories coexisting in one chart: multi-page layouts (≤ 8 colours per chart, one palette reused across pages) are not bound by it; interactive hover hints and selection highlighting relieve some of the "must identify by colour alone" pressure, though static charts lack that buffer. Maps are the disaster-zone exception — administrative divisions run to dozens of colours, but they lean on boundary lines plus labels, not pure colour encoding. This leaf sets the ceiling; the exits beyond it (grouping, aggregation, redundant encoding) occupy the next two leaves.

Applying it

• Hard-code the ceiling in palette specs: ≤ 8–10 categorical colours per chart; exceeding triggers the next leaf's grouping/aggregation flow — "squeeze in two more" is not an option.
• Selection strategy: uniform hue sampling plus alternating high/low lightness (adjacent hues must differ in lightness), preferring established palettes over ad-hoc mixing.
• Verification: run pairwise identification at the actual mark size (real dots and strokes, not swatch cards); a confusion rate above one in ten means over the ceiling or poorly mixed.

Related

• Same group: U4.01.2 Group or aggregate when categories multiply · U4.01.3 Redundant shape or pattern encoding alongside colour
• Nearby: U1.02.2 Hue is the preferred channel for categories but has limited capacity · U2.13.3 Dense networks collapse into hairballs
• Search terms: categorical palette · colour discriminability · palette size · colour naming