: Clearly distinguishable colours for scientific figures I am looking for pointers on how to pick a set of colours (or a colour gradient) that are clearly distinguishable, for use in scientific

@Kaufman565

Here's a handy splash of science.


Your eyes' retinas have receptors for red, green and blue light (cones) as well as receptors for general light levels (rods)
Between your eyes and visual cortex, there's a step (the LGN) where colour information is converted into opposing pairs: red:green, blue:yellow, and black:white. Yellow is created in special cases where red and green are in balance and blue is low, and most browns are effectively dark or low-saturation yellows (higher red and green compared to blue). You can see how these pairs are opposed to each other with after images - stare at a bright light of a colour, and look away, and the after image effect you will see will be made of its opposite colours.


So that gives you the four most clearly differentiable hues (plus no black or grey: no hue). Within each of these, you can use differences in lightness and saturation to create more colours. If you need still more colours, you can take mid points between these four hues, and then vary lightness and saturation again. (be careful with lowering saturation and raising lightness too high as it reduces the hue contrast).

Bare in mind that yellow is a special case - it is characteristically bright because it's being fueled by the activation of two types of receptors, not just one.

The question about how these convert to greyscale is important but I think probably best answered seperately. There are differences (e.g. reds tend to be darker) but it depends on other things like how the convesion is done, etc. The colour intensity chart from the above link about yellow might help as a starting point:



Two things to bare in mind:


Colour scales aren't perceived linearally. In colour heatmaps, like the second image you post, colour scales are good for helping specific values to be perceived, but this comes at a cost that the overall picture of what is high or low compared to what is harder to intuitively percieve. More info on this here. Use such 'rainbow scales' only when getting specific values from a visualisation is more important than seeing the bigger picture (usually, this means rainbow scales are okay for visualisations that are for analysis by experts, but bad for visualisations that are for general communications purposes)
A surprisingly high number of people (particularly men) are colour blind. In particular, something like 10% of men struggle to percieve differences between red and green - seeing those colours as the same hue. These people can usually make an educated guess based on lightness levels, but it's difficult and so it's best practice to make sure that where red and green communicates menaing, there are other ways to tell these colours apart than just hue. Use colour blindness simulation tools like vischeck to test designs, don't rely on just converting to greyscale as that's not an accurate simulation.

Vote Up Vote Down

@Kristi927

For your first question you might want to look into a color wheel, here's an example. Just divide the 360 degrees of the wheel into as many parts as you have required elements and you'll get colors that contrast from each other as much as possible.

For example, if you have 3 elements, your colors would be red, blue and lime.

As per figures usable in black and white, you'll often find that patterns might help identify different graphs, where such application is possible.
Where using patterns isn't possible your best bet is selective desaturation where different hues result in different shades, see an image example over here.

Selective desaturation applied to your original image can be seen here:

Vote Up Vote Down