Here is a #ritaunsolicited very long thread on writing good figure legends. These tips might not match everyone’s style, and opinions will certainly vary among editors and journals. This feedback is based on my experience with methods papers, and I hope it helps.

What is the point of a figure legend? Given that they exist as a part of a larger whole, I don’t think they are meant to be mini versions of your paper.

In essence, what they should do is complement the figure by orienting the reader to what exactly they’re looking at. The figures and legends, in turn should clearly illustrate the main points of the paper.

Should the figure stand alone without the legend and vice versa? I would argue that the best figures are self-contained in the sense that you don’t need the legend to tell you what the red line on the graph is showing, etc.

Making figures that tell a clear story outside the context of the paper is a challenge that requires its own thread and is a boon to clear scientific communication.

Reading a figure legend alone, a reader should be able to make an imaginary version of the figure in their head, not perfect, but they should get a sense of what is being shown.

Try this experiment. Ask someone to mock up a figure based on the legend. If they don’t match, why? Very often a panel description never actually tells the reader what they’re looking at in favor of describing the result and/or how the result was accomplished – this is a no no.

In our published papers, figure legends should be clear, concise and contain no ‘results’, ‘discussion’, or ‘methods’. These are the things I most frequently remove from figure legends when I line edit, with the comment “please make sure what I deleted is somewhere in the paper”.

The figure legend shouldn’t be a place to squeeze in more info because the editor gave you word limits on the main text. Describe your findings in the results section. Describe how you did it in the methods. Describe your interpretation in the discussion.

As a journal, we have strict requirements for figure legends for style, consistency, and length. We also think it’s important for things to be in the place readers will be expecting them, rather than buried in a legend. We also use figure legends for relevant stats reporting.

That being said!!! Not all of this holds for an initial submission. And honestly, as an author I always thought there would be two versions of my figure legends, the version I initially submitted and the unsweetened condensed version that made it into the published paper.

The reason for this is…convenience. In an initial submission, the goal of the figure legend should be not only to say what is in each panel, but also to state clearly the main takeaway point of the figure. What point is this figure making? State it up front.

An example: This super fluorescent protein is a versatile tag for mammalian proteins. This would be followed by the concise descriptions of what cell types and proteins are being shown in the figure.

Longer legends can offer some nice hand-holding to the reader (have in mind editors and refs).

If you have an outline of your method as Fig 1a, we are going to make you shorten the description to ‘Schematic of CUTE ACRONYM’ if we publish it. We will expect the detailed description to be in the main text. But for initial submission, go ahead and describe the steps.

You can also add a little interpretation. Instead of (i.e.) ‘Shown is a 10-point DNA origami with spacing of 10 nm imaged using COOLDUDE’. You might say ‘Shown is a 10-point DNA origami with spacing of 10 nm. COOLDUDE achieved 2 nm axial localization precision.’

I think for an initial submission this is OK, good even. It makes my life easier, because I interact with the figures first and any help I can get situating myself with the data is useful!

This doesn’t mean I don’t or won’t read the whole paper—I do! It just helps me immediately get immersed in what you’re actually showing me (rather than what you’re saying you’re showing me in the text). This is not strictly necessary, but it is nice.

This also doesn’t mean that figure legends for initial submissions should be huge and bloated and simply reiterate what is in the main text. But, you can feel free to include what you feel are major take home points.

I would at all stages avoid too many methods details in the legends, unless some difference in the method is the key takeaway from your panel (i.e.) this was done with 1000x less reagent or in 1000x less time.

Speaking of handholding and being kind to editors and reviewers, when you submit, keep your figures and their legends together. It’s a small kindness.

Finally, what makes for the best legends? Much like abstracts, they require clarity of thought, not a last minute brain dump.

Think about what you are conveying in each panel before writing, and then look at what you wrote again and say, can I write this more clearly? The PI and first author should each write them and then compare!

Before I end, I will say I've never rejected a paper because of bad figure legends, but I have rejected nonsensical or completely unclear papers, and if I had to bet they probably had useless figure legends.

This is the end for now, but I would love to see (friendly) discussion and will try to answer specific questions.