In the past years, there is a surge in the amount of colour that appears in scientific papers. There are several reasons for it. The ease of creating coloured figures increases with every new release of a software version. Journals are accepting (and encouraging) colour more than ever. Additionally, in the world of online publishing, there are no added costs for colour.
In the past, people had only black and white (and grey scale if lucky) to use. This forced them to think about line types (dashed vs dotted), symbols (circles and squares), etc. to make the graph easy to understand. Today, many people submit their pretty colour figures without putting too much thought into the act.
Let’s take a look at an example:
Seemingly, there’s nothing wrong with that graph. There are two data series, both of which are coloured (red and green) and appear in the legend. However, this figure fails in two critical aspects:
Black and white photocopying and printing
Even though it’s possible to view articles online nowadays, many people still print out papers. Just take a look at any desk here at RSES. While you may think that the green looks brighter than the red so when printing one will be dark grey and the other will be light grey, it is not the case. The two colours have the exact same luminosity and thus will appear identical in print. It is then left to the reader to try and figure out which are better, kittens or spiders. And it is open to subjective interpretation.
There is a simple solution to that: use different symbols in addition to colour:
This way, people with no access to colour printing can still understand your figure, while people with colour printing or screen viewing can benefit from your use of colour. If you prefer to use only lines and no symbols, there is a solution for you as well – line types:
Accessibility and colour blindness
Colour blindness affects mostly males (in some countries close to 10% of the population) but also females. This means that you can be confident that some of your readers are certain to be colour blind. Most commonly, people cannot distinguish red from green. This makes our figure a great example of how not to do it. While people can use the symbols to distinguish the two data series, the entire point of the colour is reduce the time the reader needs to understand the figure. Which colours should we use then? This is a hard question because of the multitude of colour blindness forms. Generally, blue and orange are a good combination:
However, with increasing data series this quickly becomes a hard task. Fortunately, there are online tools to assist in colour choice. One of my favourites is ColorBrewer: http://colorbrewer2.org/ but there are others (just search online for ‘colorblind safe colors’).
Remember – colour in scientific figures is a tool to assist understand the figure better and faster. It is not to make it pretty, although that is usually a welcome side effect.
There are two ways to test your figures after you’ve created them. First of all, print it black and white. Can you still understand the different data series? Second, you should run your figure through a simulator (or ask a colour blind friend). My favourite is the one at http://colortest.it/, but it requires that your figure already exists somewhere online. Finding other simulators should be easy using a search engine. Try to run the red-green and the blue-orange versions through the simulator!
Final note about legends
Legends are terrible in my opinion: your eyes keep moving around from the actual data to the legend. I always try to avoid using them. A better alternative is quite obvious in our case:
That’s it for now. Stay tuned for part 2 where I will discuss colour gradients. There plots were produced using R. If anyone is interested in seeing the code I used, feel free to ask me.