Colourblinded by the Light

Graphs are necessary in most academic papers. They are one of the best ways to show the data collected and illustrate relationships (or lack thereof) and support to the conclusions of the study. They also wreak havoc on a lot of authors and students -- what is the best way to show my modelled data? Bar graph? Scatterplot? Do I attempt a donut graph just to see if I can get away it it (hint: you can't).

    Published online at PLoS Computational Biology just two weeks ago, Rougier et al's "Ten Simple Rules for Better Figures" has gained a lot of press and popularity, and for good reason[1]. The format of an online-only open-access journal like those in the PLoS collection benefit this article for two reason: First, the authors were able to display remarkable graphs in full colour alongside an easy-to-understand narrative. Second, they are not limited by distribution rights and and their audience is not hindered lack of access.

    I won't break down all ten rules, but one in particular got me thinking about the accessibility of my own figures. Rule 6 is to Use Color Effectively, and ends with, "Lastly, avoid using too many similar colo[u]rs since color blindness may make it difficult to discern some colo[u]r differences." This includes a reference to Masataka Okabe and Kei Ito's page on the University of Tokyo's Drosophila research group: "Color Universal Design (CUD) - How to make figures and presentations that are friendly to colo[u]rblind people." [2]

   While the prevalence of colourblindness varies among ethnic groups and between sexes, the figures worldwide indicate that around 7.0% of the males and 0.4% of females are red-green colourblind, and a very small percentage are blue-yellow colourblind[3]. Those with red-green colourblindness are said to have either protanope vision (reduced sensitivity to red) or deuteranope vision (reduced sensitivity to green), depending on the cone cells affected. Blue-yellow colourblindness is a reduced sensitivity to blue light and the individual is said to have tritanope vision[4]. Within these catergories are variation based on the amount of working cones individuals possess. Okabe and Ito go into great detail about the different ways that colourblind individuals may be left out of scientific discussion when it comes to the colours of graphs, presentations, and stains. I feel that the staining is the most important - if you're working with fluorescent double-staining, the most commonly used stains are green and red, a sharp contrast for non-colourblind individuals, but end up blending together for those with red-green colourblindness. You can read about this and more, including the 3(+1) principles of universal colour design, on the CUD page.

    This is all interesting to me, but you may wonder what it has to do with my research. Well, I'm studying the effects of different colours of artificial lights at night affect bat behaviour - white (nearly-full spectrum), green, and red. And when it comes time to make my graphs, I use colour in addition to labels to indicate the different treatments. I never really considered the properties of my graph colours - I just want them to look nice. But looking at the examples presented by Okabe and Ito, I became intrigued. What if my graphs are not legible or aesthetic to someone despite my best intentions? Luckily, there's a way to see through the eyes of someone with colourblindness, or at least to view images with an appropriate filter.

    I use ImageJ for my research, and there is a plug-in called VisCheck that is easy to install and use[5]. Five minutes, and I was staring at a glaring issue. I originally went for an orangey-red and a yellowish-green, a palette I love. However, under the filters for each condition, they looked very similar. Too similar. While my groups have written labels as well, it's good that I checked this now before going on with my more complicated graphs. I've made a few example graphs that look like the data I'm working with and ran them through VisCheck (my actual graphs were a bit worse when filtered).

Counterclockwise from top left: Original, Protanope simulation, Tritanope simulation, and finally Deuteranope simulation.

    There may be those who would argue that I likely won't have someone who is colourblind at my defense. I likely won't publish my work in a journal that supports full colour graphs. That being said, while I'm not colourblind, that shouldn't stop me from making sure that the way I communicate is inclusive. Knowledge needs to be accessible to those that wish to gain from it. Textbook publishers have addressed this, and so should scientists, journal publishers, and educators. The tools are out there, they just need to be advocated for and used.

    Back to my research, I'm left with on question that I have yet to find an answer to: if red or green light is recommended for large-scale use outdoors, how does that affect those with red-green colourblindness?


[1] Rougier, N.P., M. Droettboom, and P.E. Bourne. 2014. "Ten Simple Rules for Better Figures"
[2] Masataka Okabe and Kei Ito's CUD page
[5] VisCheck.


Popular Posts