Statistical Atlas of the United States, Based Upon the Results of the Eleventh Census (1890) - plate 27: Proportion of the Colored to the aggregate population: 1890
The pattern
This still is filed under Data as proof.
Define
A chart, map, or count that is the argument, not an illustration next to the argument.
Mechanism
Sighted evidence. The reader is asked to believe their eyes: a bar, a share, a spread on a map.
Verdict
The claim is comparative or quantitative and the figure can be read without a caption doing the sell.
Why it works
The classification, not the colour, is what makes the pattern legible. The breaks are deliberately uneven, one, seven, seventeen, thirty-five, sixty, which spreads the Southern counties across the top three tints instead of saturating them all the same red and flattening the structure. Switching from count to proportion changes the question the map answers, so the reader sees composition rather than mass. Withholding colour from sparsely populated counties is a quiet integrity device: percentages built on almost no people are not drawn at all.
Steal this when
Use unequal class intervals whenever your values are heavily skewed and equal bins would put most of the data in one bucket, and say the breaks out loud in the key. Use the suppression rule whenever you are mapping rates or percentages and some units have too small a base to be trusted.