Guide

Best Colour-Blind-Safe Palettes for Scientific Figures

Colour-blind-safe palettes for scientific figures with hex codes: Okabe-Ito, Paul Tol, viridis, cividis and ColorBrewer, plus how to test them.

Sciicons · Published October 11, 2026 · Last updated October 11, 2026

Last updated: October 11, 2026 (added the interactive palette explorer, colour-blind simulator and more palettes). Hex codes and statistics are taken from the sources listed at the end.

Quick answer: The best colour-blind-safe palettes for scientific figures are Okabe–Ito (8 colours) or Paul Tol's Bright scheme for categories, viridis or cividis for continuous data, and ColorBrewer's diverging schemes (such as RdBu) for data around a midpoint. Pair colour with labels or shapes and test in a colour-blindness simulator.

Why do colour-blind-safe palettes matter?

Colour vision deficiency (CVD) is common: the US National Eye Institute estimates that about 1 in 12 men and 1 in 200 women have red-green colour blindness. In a lecture of 100 researchers, several people may not be able to tell your red and green lines apart. Journals have responded: Nature's research figure guide asks authors to use an accessible palette and cites Bang Wong's Nature Methods column on colour blindness.

What are the three types of palettes?

• Qualitative (categorical): distinct hues for unordered groups, e.g. treatments or cell types.
• Sequential: light-to-dark ramps for ordered values, e.g. expression level.
• Diverging: two hues meeting at a neutral midpoint, e.g. fold change around zero or correlations.

Choosing the wrong type is a more common mistake than choosing the wrong colours.

Which colour-blind-safe palettes should you use? Comparison table

Interactive palette explorer: how do these palettes look to colour-blind readers?

Use the explorer below to compare every palette in this guide. Pick a palette to see it applied to a sample bar chart, line chart and heatmap, and to a Sciicons illustration recoloured on the fly. Switch the simulator between protanopia, deuteranopia, tritanopia and greyscale to see what readers with colour vision deficiency, or a black-and-white printout, will see. Click any swatch to copy its hex code, or export the whole palette as a hex list, CSS variables, a matplotlib snippet or a ggplot2 snippet.

The simulator uses the model of Machado, Oliveira and Fernandes (2009) at full severity, which is also used by tools such as Colorspace in R. Real colour vision varies between people, and anomalous trichromacy (a weaker form) is more common than complete dichromacy, so treat the simulation as a worst-case check rather than an exact picture.

What are the Okabe–Ito hex codes?

Yellow #F0E442 is pale on white backgrounds; use it for fills rather than thin lines or text.

What are Paul Tol's palettes?

Paul Tol (SRON Netherlands Institute for Space Research) publishes colour schemes designed to be distinct for colour-blind readers and in print. His Bright scheme is #4477AA, #EE6677, #228833, #CCBB44, #66CCEE, #AA3377 and #BBBBBB. His High-contrast scheme (#004488, #DDAA33, #BB5566) also works in greyscale, which makes it useful for three-group plots.

Which colormap should you use for heatmaps and images?

Use a perceptually uniform sequential colormap. Viridis (Matplotlib's default since version 2.0) increases evenly in lightness, prints well in greyscale and is readable with common forms of CVD. Cividis was designed specifically so that people with and without CVD see nearly the same colours. For values above and below a meaningful midpoint, use a diverging map such as ColorBrewer RdBu or Matplotlib's coolwarm, and centre it on zero.

Avoid jet/rainbow colormaps: their uneven brightness invents boundaries that are not in the data.

When should you use magma or inferno instead of viridis?

Magma and inferno are part of the same perceptually uniform family as viridis, created by Stéfan van der Walt and Nathaniel Smith for Matplotlib. They run from near-black to pale yellow, so they have a wider lightness range than viridis. That makes them useful for dense images such as microscopy intensity maps, astronomy images or spectrograms, especially on dark backgrounds. Viridis remains the safer default for heatmaps printed on white paper because its darkest colour is a dark purple rather than black, which keeps cell borders and labels readable.

Which colours work for fluorescence microscopy overlays?

The classic red/green merge is the single most common colour-blindness problem in biology figures, because yellow overlap is invisible to many protanopes and deuteranopes. Use magenta and green for two-channel merges (overlap appears white), or show each channel in greyscale side by side with a coloured merge. For three channels, magenta, green and blue (or cyan, magenta and yellow) remain distinguishable. Always add a channel legend in text, not just colour.

How many colours can a palette safely have?

Most people can reliably match about six to eight distinct hues across a chart and its legend. Okabe–Ito offers eight colours including black, and Paul Tol's Muted scheme offers nine, but ColorBrewer rates many of its qualitative sets as colour-blind safe only up to three classes. If you need more categories, combine colour with marker shape, line style or direct labels, group minor categories as "Other" in grey, or split the data into small multiples.

How does ColorBrewer mark colour-blind-safe schemes?

ColorBrewer 2.0, designed by Cynthia Brewer for maps, lets you filter schemes by "colorblind safe", "print friendly" and "photocopy safe". The rating depends on the number of classes you choose: diverging schemes such as RdBu, PuOr and BrBG stay colour-blind safe at most sizes, while qualitative sets such as Set2 and Dark2 are only rated safe for small numbers of classes. Check the flags on the site after choosing your number of classes.

How do you apply a palette to a figure step by step?

1. Decide the data type (categories, ordered values or diverging values) and choose the palette type.
2. Limit the number of colours. Beyond 6–8 categories, group or facet instead of adding hues.
3. Assign colours by meaning. Keep the same colour for the same condition across every panel and figure.
4. Use neutral grey for controls or context and save saturated colours for what matters.
5. Add redundant cues: direct labels, line styles or marker shapes.
6. Test with a simulator such as Color Oracle or Coblis, and in greyscale.
7. Match illustrations. Recolour schematic icons to the same palette; Sciicons SVGs can be recoloured in the icon detail view or in any vector editor so diagrams match your plots.

How do you set the palette in R and Python?

• R (ggplot2): scale_colour_manual(values = palette.colors(palette = "Okabe-Ito")); for continuous data scale_fill_viridis_c().
• Python (Matplotlib): cmap="viridis" or cmap="cividis"; for categories pass the Okabe–Ito hex list to color= or set plt.rcParams["axes.prop_cycle"].
• Seaborn: sns.color_palette("colorblind") gives a CVD-friendly categorical palette.

Common colour mistakes in scientific figures

• Red vs green as the only contrast in plots, gels or microscopy overlays (use magenta/green instead for fluorescence merges).
• Rainbow heatmaps.
• Different colours for the same condition in different panels.
• Too many saturated colours, leaving nothing to stand out.
• Coloured text that becomes unreadable when printed.

Related guides

• How to choose colours for scientific figures
• Journal figure requirements: DPI, formats and sizes
• How to make a scientific poster

Sources

• National Eye Institute, colour blindness: https://www.nei.nih.gov/eye-health-information/eye-conditions-and-diseases/color-blindness
• MedlinePlus Genetics, colour vision deficiency: https://medlineplus.gov/genetics/condition/color-vision-deficiency/
• Okabe & Ito, Color Universal Design: https://jfly.uni-koeln.de/color/
• Wong B., Points of view: Color blindness, Nature Methods 8, 441 (2011): https://www.nature.com/articles/nmeth.1618
• Nature research figure guide: https://research-figure-guide.nature.com/figures/preparing-figures-our-specifications/
• Paul Tol, colour schemes: https://personal.sron.nl/~pault/
• Matplotlib colormaps: https://matplotlib.org/stable/users/explain/colors/colormaps.html
• Nuñez, Anderton & Renslow, cividis, PLOS ONE 13, e0199239 (2018): https://doi.org/10.1371/journal.pone.0199239
• ColorBrewer: https://colorbrewer2.org/
• Machado, Oliveira & Fernandes, A physiologically-based model for simulation of color vision deficiency, IEEE TVCG 15(6), 1291–1298 (2009): https://doi.org/10.1109/TVCG.2009.113
• Smith & van der Walt, A better default colormap for Matplotlib (SciPy 2015): https://bids.github.io/colormap/

More guides on the Sciicons blog · Browse scientific illustrations · Free figure tools