Sequential palettes must increase monotonically in perception

What it is

A sequential palette maps "larger value" to "looks more/intense," and the mapping holds only if the perceptual quantity is monotone: walking the ramp from low to high, perceived intensity (lightness-led overall strength) must climb without ever turning back. A turning-back segment is lethal: two far-apart values (on either side of the reversal) look equally strong, and readers take them for peers — the ramp pulls the distant together and pushes the near apart, and ordinal encoding fails wholesale. The test is not a glance but a plot: convert the ramp to perceptual lightness (L*) and look at the curve — a qualifying sequential palette is an ascending line that never doubles back.

Why it happens

Ordinal reading rests on the stable conversion "brighter/stronger-looking = numerically larger," and monotonicity is the precondition for that lookup table. A non-monotone ramp errs not as uniform noise but as directional deception: values flanking the reversal pair up (equal visual strength), differences at the gradient reversal flatten, and order across it inverts. Perceptual uniformity is the second layer: equal value increments should occupy equal perceptual distance, or the ramp's front crams together while its back leaps, and the "density structure" read off the chart is the ramp's artifact, not the data's — precisely the rainbow's indictment (next leaf).

Where it stops holding

Monotonicity binds sequential encoding (continuous value mapped); categorical palettes are exempt (categories unordered, needing mutual parity instead); diverging palettes are "two sequential segments joined at a midpoint," each monotone on its side (diverging group). The perceptual quantity is lightness first, saturation second (third leaf), and "perceptually monotone" is not "HSL-lightness monotone" — HSL's L is not perceived lightness (yellow and blue at equal L look far apart), so validation must use a CIELAB/OKLab-class space.

Applying it

• Run the monotonicity check before a palette enters the library: take L* (or OKLCH L) per colour, plot index-versus-lightness, and reject on any descending segment.
• Choose established perceptually uniform palettes (viridis family, ColorBrewer sequential); custom ramps only via interpolation in uniform colour spaces, re-checked for monotonicity.
• Verification: take the ramp's two ends and midpoint, ask readers to order them by strength; failure to order or mis-ordering means monotonicity broken.

Related

• Same group: U4.02.2 Rainbow palettes create false boundaries · U4.02.3 Luminance carries the sense of order
• Nearby: U4.03.1 Diverging palettes suit data with a meaningful midpoint · U2.10.3 Colour scale choice directly shapes the conclusion
• Search terms: sequential palette · perceptual monotonicity · viridis · OKLab