A mark being visible against the background and being told apart from its neighbor are separate demands

What it is

A readable multi-category chart must satisfy two independent contrast requirements simultaneously: each mark's contrast against the background ("can be seen") and contrast between adjacent category marks ("can be told apart"). Palettes checked only against the background often fail the second test—five mid-lightness colors that all pass vs the background may differ so little from each other that no one can tell which bar belongs to which category. The two requirements map to two reader tasks: locating (where is the data) and attribution (which category is this), and both are necessary.

Why it happens

The requirements' independence comes from measuring different color relationships: mark-vs-background is the lightness difference of "mark color vs a single background," while adjacent contrast is the difference "mark A vs mark B"—the former has one reference, the latter N−1 pairs, and the direction that maximizes background contrast (high lightness separation) often conflicts with maximizing adjacent separation (an evenly stepped lightness ladder). The practical approach for multi-category palettes is a qualitative palette: a set of colors distributed evenly in lightness or hue, designed specifically to maximize adjacent separability; ad-hoc "looks different" colors guarantee nothing about adjacent pairs. Color-vision deficiency sharpens the adjacent-contrast problem: red-green pairs converge toward the same perceived color for deuteranope readers while looking distinct to typical vision—accessible category palettes must be verified under color-blindness simulation. Shape, texture, and direct labeling (series names on the marks) are color-independent alternative or reinforcing channels: when adjacent contrast cannot be widened further (too many categories), adding attribution through labels and shapes is the solution, not infinitely stretching hue distances.

Where it stops holding

The demand for adjacent contrast runs into the ceiling of category count: beyond roughly 7-9 categories, no palette guarantees adjacent separability (adjacent knowledge: too many categories are unreadable in themselves), and the right move is merging categories, faceting, or interactive highlighting rather than straining hue differences. "Adjacent" needs definition: not every category pair—check pairs spatially adjacent on the chart or sequential in the legend, and in stacked charts, adjacent in stacking order. In multi-line time series, "adjacent" means vertically neighboring lines at the same x—the spacing between lines itself separates them, so color-contrast demands can relax accordingly; the two requirements' weights differ by chart type.

Applying it

• After choosing colors, run both checks: each color vs background ≥3:1; every adjacent pair's difference (ΔE or contrast difference) at a discernible standard.
• Re-verify adjacent separability under color-vision deficiency simulation (protanopia, deuteranopia, tritanopia filters).
• Beyond 8 categories, switch to faceting, merging, or interactive highlighting; never rely on color alone.
• Verification: grayscale the chart and check whether adjacent categories remain distinguishable (is the lightness ladder stretched?); adjacent categories that fuse in grayscale are a palette failure instance.

Related

• Same group: U11.04.1 Chart marks need discernible contrast against the background · U11.04.3 Too-thin lines vanish on low-contrast screens and projectors · U11.04.4 In-chart text minimum size is judged by final rendered size, not the design file · U11.04.5 Interactive marks are also bounded by touch-target requirements
• Nearby: U11.04.1 Chart marks need discernible contrast against the background · U1.05.4 Adjacent colors mutually influence hue judgments
• Search terms: qualitative palette · color blindness · adjacent contrast