Short answer
A rainbow (jet or HSV-style) scale runs through hues whose lightness goes up and down rather than in one direction, so it has no perceived order, creates sharp bands where the data are smooth, and hides differences inside its wide green–cyan–yellow stretch. It also depends on red–green discrimination and falls apart in greyscale. Perceptually uniform maps such as viridis, cividis or the scientific colour maps avoid all of these.
The table below samples a fully saturated hue sweep from blue to red — the shape of the classic 'jet' and HSV maps — every 30°. Its CIELAB lightness starts dark at blue, climbs steeply to cyan, stays almost flat through cyan, green and yellow-green, peaks at yellow, then falls again to orange and red. The data run in one direction; the lightness does not. A reader's eye treats lightness as magnitude, so the brightest part of the scale, yellow, draws attention as though it were the extreme, although it sits about three quarters of the way along. Crameri and colleagues make the same observation: yellow, the brightest colour, attracts the eye yet is neither the middle nor the end of the map.
Look at the neighbour-to-neighbour column. Some 30° steps change colour enormously (blue to azure, yellow to orange) and others barely at all (the greens). On a continuous field that means a smooth gradient in the data is drawn with apparent boundaries at the big steps and apparent uniformity across the small ones. Borland and Taylor summarised this as the rainbow's two failures: it imposes structure that is not in the data and hides structure that is. Crameri and colleagues estimate that non-uniform maps can introduce visual error of more than 7% of the displayed data range. The practical consequence is worst in exactly the fields that use rainbow most — medical imaging, geophysics, fluid dynamics — where a false edge can be read as a real boundary.
The rainbow's order depends on telling red from green and on hue in general. For deuteranopes and protanopes, the red end and the green middle can look similar, so the scale folds back on itself. Crameri and colleagues cite a worldwide prevalence of colour-vision deficiency of about 8% of men and 0.5% of women, and higher in some populations. In greyscale, the problem is worse still: the flat green–yellow plateau becomes a single grey and the two ends, blue and red, become similar mid-darks. Nuñez and colleagues designed cividis specifically so that a map looks nearly the same to readers with and without red–green deficiency — the opposite design goal from a rainbow.
Rainbow maps stay common because they were the default in widely used software for decades and because they look vivid. More hue does help readers name locations on a map ('the yellow region'), and that is a legitimate need. The answer is not to ban hue but to make lightness carry the order and let hue change along a path that keeps lightness monotonic, which is what viridis and the scientific colour maps do. Cyclic data — wind direction, phase, time of day — are the one case where a hue circle genuinely fits the data, and even then a cyclic map with even lightness steps is preferable to a raw hue wheel.
Switching away from rainbow is usually a one-line change: in Matplotlib, since version 2.0, viridis is already the default; in other tools, choose viridis, cividis or a scientific colour map explicitly.
| Hue | Colour | CIELAB L* | Change in L* from previous | ΔE00 from previous |
|---|---|---|---|---|
| 240° | #0000FF | 32.3 | — | — |
| 210° | #0080FF | 54.7 | +22.4 | 28.7 |
| 180° | #00FFFF | 91.1 | +36.4 | 43.9 |
| 150° | #00FF80 | 88.5 | -2.6 | 27.3 |
| 120° | #00FF00 | 87.7 | -0.7 | 9.0 |
| 90° | #80FF00 | 89.9 | +2.2 | 5.2 |
| 60° | #FFFF00 | 97.1 | +7.2 | 18.5 |
| 30° | #FF8000 | 67.1 | -30.1 | 42.4 |
| 0° | #FF0000 | 53.2 | -13.8 | 21.2 |
Why: The rainbow map's large perceptual jumps between some neighbouring hues.
Fix: Replot with viridis, cividis or a scientific colour map and compare.
Why: Yellow is the lightest point of the rainbow and reads as extreme.
Fix: Use a map whose lightness increases monotonically with the data.
Why: The map's order depends on red–green discrimination.
Fix: Use a map designed for colour-vision deficiency and add contour labels where values matter.
Each statement is labelled by kind — established fact, a standard’s requirement, observed market data, a convention, or Colourwise’s own interpretation or analysis — with the strength of the evidence behind it.
FactStrong evidence
Crameri, Shephard and Heron argue that rainbow colour maps distort data because their lightness is uneven — yellow is brightest but neither central nor at the end, and greens form a wide band of low perceived contrast — and estimate visual errors of more than 7% of the displayed data variation for non-uniform maps.
Caveat: The 7% figure is the authors' estimate for the maps they analysed, not a universal constant.
FactStrong evidence
Borland and Taylor's 2007 article summarised evidence that the rainbow colour map lacks perceptual ordering and introduces artefacts, and recommended against it for scientific visualisation.
Colourwise analysisStrong evidence
In a fully saturated blue-to-red hue sweep, CIELAB L* rises from about 32 at blue to about 97 at yellow and falls again to about 53 at red, with a near-flat stretch across cyan, green and yellow-green.
Based on: Computed by Colourwise from CSS HSL at 100% saturation and 50% lightness; see the table.
Source: CSS Color Module Level 4; CIE (International Commission on Illumination) publications
FactStrong evidence
Cividis was designed so that a colour map appears nearly identical to viewers with and without red–green colour-vision deficiency.
Reviewed 29 September 2026. Colourwise summarises its sources in its own words and does not reproduce standards text or proprietary colour data. Spotted an error? Tell us.