Click here for a summary of the main recommendations • This essay is part of the science style guide series


Figures are the lifeblood of scientific papers. Many people won’t read your paper, but will look at the figures.

Why is that? There are many reasons, ranging from being attracted to shiny colors to reducing cognitive load by activating a different part of memory. But I suspect that one of the most important reasons is that figures allow more freedom to writers than text. There are no word limits, fewer formatting constraints, no bad writing style norms. All that matters is that your figure is clear and true. Accordingly, scientists spend a lot of time on them, and they are often the best part of any scientific paper.

Because of this freedom, I would say that figures are perhaps the part of scientific publishing that works best as thing currently stand. That doesn’t mean we can’t improve on them.

Captions

One of the norms I observe around figures is that they should be “separate” from the main text. The caption should be sufficient to understand a figure even if someone hasn’t read the text. This is a good idea overall, but it sometimes leads to paragraph-length captions that are as difficult to read as the main text. Not great.

Journals also usually limit the number of figures you can have in a paper. This can lead to very dense figures, with multiple subparts (Fig. 1A, 1B, 1C, etc.), trying to cram as much information as possible. This is the visual equivalent to a long and dense paragraph. Also not great — plus it makes the captions even longer and confusing.

Here’s an example figure and caption from the platypus paper I rewrote:

Fig. 4. (A) Alignment of predicted protein sequence of transcripts containing a recombined VHδ gene isolated from platypus spleen RNA. The individual clones are identified by the last three digits of their GenBank accession numbers (JQ664690–JQ664710). Shown is the region from FR3 of the VHδ through the beginning of the Cδ domain. The sequence in bold at the top of the alignment is the germ-line VHδ and Cδ gene sequence. The double cysteines at the end of FR3 and unpaired cysteines in CDR3 are shaded, as is the canonical FGXG in FR4. (B) Nucleotide sequence of the CDR3 region of the eleven unique V(D)J recombinants using VHδ described in the text. The germ-line sequence of the 3′-end of VHδ, the two Dδ, are shown at the top. The germ-line Jδ sequences are shown on the right-hand side of the alignment interspersed amongst the cDNA sequences using each. Nucleotides in the junctions between the V, D, and J segments, shown italicized, are most likely N-nucleotides added by TdT.

I’m sure all of it is essential information, so I won’t say that it should be shortened, exactly—although if you can, yes, shorten it! But it’s a figure. It’s supposed to be visual. As much of the caption as possible should be included directly in the figure.

Then there’s the fact that Fig. 4 is in fact Fig. 4A and 4B. The “(B)” is buried in the middle of the caption, with no bolding or anything to make it stand out. Readers have to parse the caption themselves to understand the two parts separately. Plus, the text uses parentheses for normal parenthesis purposes, and even has a “(D)” that doesn’t refer to a hypothetical part D. This is asking a lot of effort from readers!

I do like the fact that the caption refers to the main text (it says “described in the text” in part B), although it would have been nice if it said where to look exactly in the text. As things stand, the reader has to spend energy to find the spot being referred to.

Also, if your caption contains the word “red” or “blue” or any color, and does not use the mentioned color, then you’re doing something wrong. Using color directly in a caption, to reference a color-coded part of the figure, helps your reader so much it would be mad not to do so. (There can be technical limitations, of course, but it’s not usually very difficult to add color to text. I wasn’t sure I could do it on this website, and it took three seconds of googling the relevant HTML snippet.)

Referencing a figure

The sacrosanct separation of text and figures means that in the main text, the commonly preferred way to refer to a figure is to put “Fig. X” in parentheses, e.g. (Fig. 2). You almost pretend that the figure doesn’t exist. When a figure is central to explaining something, that can lead to a rather silly repetition of references to the figure, like this:

Here is a sentence explaining something that can also be presented visually (Fig. 2). Of course, that sentence raises a second point about something adjacent, also included in the visuals (Fig. 2). When comparing those two ideas, we can derive some interesting results, such as this and that (Fig. 2).

There should be a way to integrate the text and figure without losing the advantages of separation. The proper path, it seems, is to make sure the figure stands on its own, but to reference it liberally in text. This suggests it’s okay to write e.g. “Figure 2 shows that …” or “Refer to figure 2 for a visual explanation.”

In the case of multipart figures, the common practice of referring to a specific subfigure (e.g. Fig. 4B) is good, but it’s possible to do better. If part B of figure 4 happens to be in the top-right of the figure, you can specify that readers should look in the top-right: “as we can see in fig. 4B (top-right).” This is especially helpful if the “B” is small in the figure. That way, your readers will immediately know where to look. You don’t want them to waste their enegy trying to match the text and the visuals.

Similarly, if you control the final layout enough to know where the figures end up being printed relative to the text, it’s helpful to add indications like “on the next page” and such. (If you don’t control the final layout, then this is in the hands of editors, who should also follow my advice.)

Sometimes, we also want to reference a figure that was shown much earlier. Ten pages into a paper, after you’ve been referencing Fig. 6 and Fig. 7, readers expect the next figure to be Fig. 8. If instead you refer to Fig. 1, it can be confusing because it doesn’t match the expectations, not to mention that Fig. 1 is probably on a distant page. You can alleviate that by explicitly saying you’re referring to something from earlier, e.g., “… as we saw in Fig. 1 from the introduction.”

And again, if you reference a color-coded part of the figure, there’s no rule against using color in the main text to ease the matching.

Abbreviations

In my post on abbreviations, I was pretty vehemently against them, or at least their proliferation. I would be more lenient in figures because spatial organization is more often a concern. If you have a plot with 30 lines and you want to label them all and they all refer to something like “Gross Domestic Product of the Democratic Republic of the Congo,” then yes, use abbreviations.

But that’s only if space and/or clutter are actually a concern! If the space you save thanks to an abbreviation remains blank, and if it doesn’t really declutter the figure, then you have made things harder for the reader, for no gain.

If you do use abbreviations, they have to be spelled in full somewhere close by. The standard way to do this is include them at the end of the caption, but this has the drawback of making the caption long and difficult to parse. A line of prose is also not the best way to format a list of abbreviations. If you can, a visual legend within the figure is better.

Other stuff

Tables follow similar conventions than figures, and most of the advice here also applies to them.

Beyond the scope of JAWWS, another interesting improvement around figures would be to move towards a system where they can be published on their own, without being embedded in a paper, and have figure-makers be recognized for their work. But that’s a bigger change than just improving the norms.

Recommendations

  • Avoid very dense figures and dense captions
  • When possible, use separate figures more than subfigures (unless it is relevant that the subfigures be together)
    • When using subfigures, make it clear where the captions for each subpart begin and end
  • Use color in captions and text when referencing a color-coded part of a picture
  • Make use of your figures in the text; reference them in a way that makes things easier for reader
  • Abbreviations are okay to save visual space, but don’t overdo it