Figures are the part of your scientific story that your reader is likely to see first. If you want to publish a killer paper, you should, therefore, not only tell a concise story in your text but also in your figures. Don't forget to download the free checklist!
Figures are the central part of almost every scientific paper. They both capture your main findings and are often the first thing your reader will notice. This is because our brains can register visual images faster than text. Your figures will not only complement your story but be a part of it. Based on the points that typically come up when I work with scientists on their papers, I developed these 10 steps that will help you make brilliant figures to fit right into your scientific story.
1. Define the message of each figure
When you decide which and how many figures you want to show in your paper, decide on one message that you want to communicate with each figure. It should ideally represent one of your conclusions or parts of it. When you follow through the next steps, think about this message in each desicion you make.
2. Pick the most suitable plot type
Choose a plot type that conveys your message in the simplest and most accurate manner. This could be a bar chart, a scatter or line plot, a heat map or a histogram. If you are unsure, show your figure to a colleague and check whether their interpretation of your data matches your intended message.
3. Be selective
When you select the data that you want to present in your figure, be minimalist. Put as little data in your plots and have as few figures as needed to accurately represent your findings. You recorded several sets of reproducible data? That’s great to mention in your main text or something to show in the supplemental information but there is no need to put all sets of data in your main figures. Including error bars can sometimes be a way to reduce the number of data series (and is always encouraged anyway). If you still need to show a large number of data series, consider splitting them up in two half-size plots.
4. Determine an appropriate axis intercept
Play around with the axis scales to find the section that captures your message most accurately. For example, if you need to display an absolute magnitude, showing the whole scale might be most suitable. If you want to show the difference between various data points or series, a zoom into the section where this difference becomes visible would be the way to go. When you adjust the axis ranges, always make sure to not distort the data to mislead the reader. To not waste any space available in the figure, let your data fill the whole plot.
The simpler your plot, the easier and quicker your reader will grasp it. In some data visualisation programs, the default options are full of clutter. So, remove everything you can from your plot while still telling the same message. That holds true for 3D effects, background grids, background shading, excessive tick marks, frames around legends etc. If you only have a few data series, it might be appropriate to skip the legend and put labels right next to the graphs. Another thing to bear in mind is to choose a 2D over a 3D plot whenever possible – perhaps a second y-axis would do the trick?
6. Choose informative titles for your axes and legends
Try to find titles and descriptions for your plot that one can understand quickly, and that match the rest of the manuscript. If you use many different acronyms in your manuscript, however, it might be better to spell them out in the axis titles. Someone looking at your figure might not have read the full paper. You’ll quickly lose them if they have to skim through several pages to find out what an uncommon acronym stands for. Is it needless to say that units always have to be included? In your legend/key, I suggest you leave out any unnecessary information. For example, instead of repeating “current voltage curve 1st run”, “current voltage curve 2nd run” and “current voltage curve 3rd run” in your legend, you can simply label the data series “1st run”, “2nd run” and “3rd run”.
7. Use a consistent design
A clear and consistent design in your figures will make it easy for your reader to gather the presented information. For this, I suggest to use the same colour and symbols for each variable throughout all your figures. Sample 1 is displayed as red triangles in Figure 1? Make sure it is in Figure 5 too. Do also use your font, font size, marker size, line size etc. consistently. Try to match the font size in the figure to that of the main text in the final article format. For example, Nature recommends a font size of 8 pt and Science of 9 pt in the final figure. If you show different panels in your figure, use a design trick: Imagine an invisible grid in the background (or make it visible in the program you use) so that all panels are aligned.
8. Decide on a colour palette
There are better and worse colours you can use in your plot. Firstly, colours that are hard to see on the screen or in print, such as yellow or beige, should be avoided. In order for colour blind people to read the figure, don’t use red and green together. If you need many different colours, for example for a heat map, experts don’t recommend the rainbow colour scale. More intuitive are the different colour schemes provided by the website Color Brewer that distinguish between sequential, diverging and qualitative data. Another point to keep in mind is that we often connect different meanings to different colours. For instance, we typically interpret darker colours as “more” and blue as more abstract.
9. Add necessary extras
No, this isn’t a contradiction. Although we aim for an as simple figure as possible, we still want to assist the reader with interpreting it. Perhaps a curve fit or a line as a guide for the eye would be helpful to see trends? Or does your graph show different ranges that could be divided by using dashed vertical or horizontal lines? Do the ranges correspond to different states in your model system that can be visualised with a simple graphic or scheme? Include them then! Your reader might also need arrows, an insert showing the measurement conditions or a zoom into a specific region to get your key message.
10. Provide informative figure captions
Once the visual part of your figure is done you should spend some time on the captions. As important as a great figure is a caption that gives your reader all the necessary information in the shortest way possible. As a point of reference, Nature restricts their caption to 250 words, and Science to 200 words. Start your caption with one title sentence that captures the main result in the figure. Then explain everything depicted in the figure, such as all panels, variables and any methodological information necessary to understand the data. Ideally, your reader can understand the result in your figure without reading the main text. The figure caption is, however, not the ideal place for interpretations or the conclusions you have drawn from your data.
This guide focusses on data visualisation because this is the most common type of figure in the papers of my clients, who work mainly in Physics and Chemistry disciplines. However, many of the mentioned aspects also apply when your figures consist of schemes and images.
Whatever your figure will show, there are three key points to an excellent figure:
1) Have a clear message for each figure
2) Make it as simple as possible while still conveying the message
3) Show best ethical practice when you present your data
What do you struggle with most when you make your figures? What other topics would you like to see in this blog? Let me know in a comment or write me a message. Do you want a free checklist to make sure your figures are in great shape before you submit your paper? Click on the image below.
If you want help to get the story right in your manuscript, just give me a shout to learn about the different editing and writing consulting packages I offer.
Elegant Figures Guide on Earth Observatory Blog by NASA
“Ten Simple Rules for Better Figures” by N. P. Rougier et al. doi.org/10.1371/journal.pcbi.1003833
“Ten guidelines for effective data visualization in scientific publications” by C. Kelleher et al. dx.doi.org/10.1016/j.envsoft.2010.12.006 (it seems like the magic number is 10 when it comes to making figures)
About the Author:
I’m Anna Clemens, a scientific editor and writing coach for scientists. I give workshops about scientific writing, offer strategy calls and edit papers and proposals. I’d love to work with you, please click here for more information.
I also regularly blog about scientific writing and write articles about science for magazines and websites. I hold a PhD in Chemistry/Materials Science.
When I’m not at my desk, I’m probably out and about with my dog and assistant Zuza.
Twitter handle: @annacle_science