Skip to main content

Together we are beating cancer

Donate now
  • For Researchers

Research with integrity – why colour blindness is a research integrity issue

The Cancer Research UK logo
by Cancer Research UK | Analysis

28 February 2023

0 comments 0 comments

Of all the tricky-to-fix research integrity issues researchers contend with, making your research accessible to those with colour blindness is actually quite an easy fix. Not only that, says Dr Andrew Porter, it shows care and respect to all readers of your work, and that can only be a good thing…

This entry is part 4 of 14 in the series Research Integrity
Series Navigation<< Research with integrity – why building networks is vital for integrityResearch with Integrity – integrating training into research culture >>

It’s a classic kid’s question – does everyone see the same colours when they look at something?

Some readers may remember a certain amount of ambiguity around the colour of a particular dress which suggested that different people do indeed perceive colour differently.

For some people, this difference is even more extreme, with an estimated 300 million people worldwide having some kind of colour vision deficiency (commonly called colour blindness). Most common is red-green colour vision deficiency (deuteranopia), affecting around 1 in 12 men and 1 in 200 women in the UK.

True colours

I’m mindful of these numbers when reading research papers in our Institute’s pre-submission review process, as described by Catherine recently. One thing I look for are figure elements that may be difficult for colour blind people to distinguish. These typically fall into three categories – research images (such as microscopy images) where colour is part of the image itself; heatmaps and graphs where colour represents different data elements; and colour diagrams, models or schematics.

Helping researchers make these more accessible is part of our commitment to showing care and respect to “all participants in research, and for the subjects, users and beneficiaries of research”, one of our responsibilities as signatories of the Concordat to Support Research Integrity.

The complexity and depth of biological data is rising all the time. Heatmaps, tSNE and PCA plots, mutational signatures – all frequently use different colours to represent multidimensional data.  Look in any recent paper and you’ll see pie chart segments, bars, lines and data points in a whole spectrum of colours. Many papers include illustrations to helpfully demonstrate methods or summarise a model, and these often come in colour. But all these can be a barrier to accessibility; a recent survey found 48% of cell biology papers containing some figures which were inaccessible to people with deuteranopia (of 124 examined), with similar outcomes for physiology and plant research.

Back in black

Several factors are at work here, particularly trends in how data is published.

With the need to represent hugely complex datasets in an easy-to-understand fashion, colour is an obvious choice. Colour looks attractive and adds interest to a paper – especially in an Instagram-filled world. A decade or two ago, printing charges for colour pages in journals would have found many of these elements represented in greyscale to save money. Readers were more likely to print a paper than read it on screen, and colour printing has always been more expensive, so it was more common to design figures that could be read in black and white.

The continued rise of digital publishing, changes in payment models, and more people reading on screens now mean the full range of colours are employed, even for simple charts and graphs.

If I find figures that would lose meaning for those unable to see the colours, I flag these with the authors. I use the “simulate colour blindness” function in ImageJ/FIJI to create a new version of the image to demonstrate the issue to the authors. I feel this makes the point most clearly, because these issues are literally invisible to most people. I then provide simple suggestions for how the image could be made more accessible (see below for these takeaway tips).

a) A chart produced from sample data in GraphPad Prism, with three colours used qualitatively to distinguish between the data sets b) The same chart adjusted using the “Simulate Colour blindness” function in ImageJ to mimic the effects of deuteranopia, or red/green colour blindness, in which the two treatment groups are now virtually indistinguishable c) The same chart designed using a more colour blind friendly palette d) Even after simulating the effects of deuteranopia, the three data sets can be distinguished. The labels have also been moved closer to the data to further help identify the datasets.

Kind of blue

Another big challenge in terms of colour vision is the use of microscopy images where different colours represent the signal from different channels.

I’m sure you’re aware that putting a red and green image on top of each other is not a good idea, because these cannot be distinguished by people with common forms of colour-blindness. However, while it’s true this doesn’t occur often in papers I review, the briefest glance at the literature shows it does still happen.

As a very small example, I looked on Pubmed at papers citing a classic colocalization analysis paper from 2011 (which incidentally also shows red/green overlays) and 6 of the first 10 citing papers contained red/green overlays. So even this commonly understood issue persists in the literature.

Over the rainbow

I think there are two big reasons why we continue to see so many colour-based issues in publications. Firstly, it’s clear we still need greater awareness of the problem, and secondly, we need measures to address this that don’t add time-pressure for researchers.

Despite a wealth of great resources and advice (such as this piece from Nature in 2021), awareness remains low. While some journals have guidelines on colour use, these are rarely enforced. The International Committee of Medical Journal Editors guidelines (from which many journals draw their guidance) don’t mention colour blindness. We can all play a role by sharing good practice, highlighting potential issues in papers we review, and advocating for positive change. And this isn’t just about figures in papers; we should be mindful of our use of colour in presentations and posters, too, and many pieces of software have tools to help spot potential issues (e.g. the Accessibility Checker in Microsoft Office 365).

Researchers’ time is also an issue. It can be time-consuming to recolour figures, especially if they have been embedded into a PDF or separated from the raw data used to generate them, and researchers are under great time pressure already. This is why it’s best to adopt good practices in figure creation. Defining a colour-blind friendly colour scheme for the whole paper also helps ensure consistent use of colour (the same colour representing the same entity throughout the manuscript), increasing readability and comprehension.

Other approaches can help: using alternatives to colours like different shades of grey, hatching or shape markers; greater use of labels; splitting microscopy channels; sharing data in tables or even producing interactive figures that can be recoloured, such as the data in this article from eLife.

Goodbye yellow brick road

Unlike many other research integrity issues, ensuring that data is presented in an accessible way is a relatively simple problem to fix.

We know what the issue is, tools are available to solve it, and there is little downside to doing so. In fact, there can be further upsides, as using colours in an accessible way often leads to better design and communicability more broadly. And, speaking pragmatically for a moment, no researcher should want to get on the wrong side of a reviewer who happens to be colour blind.

Thinking about the accessibility of data is a practical way to show care and respect to all readers of your work, and that can only be a good thing.

Takeaway tips

  • Consider – do you need to use colour, or could you choose another way of distinguishing your data? There are some good examples here of using shapes, patterns, textures and labels.
  • Choose a colour-blind friendly palette when you start producing figures – this one, for example. Or generate and test your own – and introduce it as a lab standard.
  • If you’re coding figures in R language, add colour blind friendly options. More examples are here.
  • For immunofluorescence images, consider showing each channel using greyscale alone, and use a colour-blind friendly palette for your overlay images.
  • If you already have images, check them using a colour-blind simulator such as this one.

See some general resources and information on wider issues around colour blindness are available on the Colour Blind Awareness website.

Author:
Dr Andrew Porter is Research Integrity and Training Adviser at Cancer Research UK Manchester Institute.