The skeleton before the prose, outline, and build your figures first

By now you have a sharp question (Part 1) and the right design with its reporting checklist (Part 2). The temptation is to open a blank document and start writing the Introduction. Resist it. The single slowest, most painful way to produce a paper is to write it one paragraph at a time from the top, hoping a structure will emerge. It rarely does. What emerges instead is a draft that wanders, repeats itself, and has to be torn down and rebuilt, usually more than once.

There is a faster way, and it feels counter-intuitive: build the skeleton and the figures first, and write the prose last, into a structure that already holds together. This part is about that skeleton.

Why the figures come before the sentences

A scientific paper is, at heart, an argument made of evidence. The evidence lives in your figures and tables. The text exists to walk the reader through that evidence in order. So the figures are not decoration you add at the end, they are the paper. If you design them first, you have designed the argument; the writing becomes narration.

This is why experienced authors often build what is sometimes called the “figure deck” before writing anything: lay out, in order, every figure and table the paper will contain, each with a one-line caption stating what it shows. When you can flip through that sequence and watch the story build, here is the problem, here is what we did, here is the key result, here is the control that rules out the obvious objection, here is the effect in detail, you have a paper. If the sequence has a gap, a detour, or a figure that proves nothing the reader needs, you have found it now, on one page, instead of after writing three thousand words around it.

Let the one message decide every figure

Remember the single sentence from Part 1, this study shows that ______. That sentence is the constitution of your paper, and it governs the figures with a simple, ruthless rule:

Every figure must earn its place by advancing the one message. If a figure does not help the reader believe or understand that sentence, it does not belong in the paper.

This rule kills the most common figure disease: the data dump. Trainees include a panel because they made it, because it was a lot of work, because it looks impressive. None of those are reasons. The reason a figure exists is that the argument needs it. Everything else goes to the supplement or to the bin.

Apply the rule and a clean structure usually falls out:

• A first figure or table that establishes the starting point, often “Table 1,” the characteristics of your sample, so the reader knows who or what was studied.
• A central result figure, the one that, on its own, carries the one message. If a reader saw only this, they should grasp your finding.
• Supporting figures that strengthen, qualify, or explain the central result, a dose-response, a subgroup, a mechanism, a sensitivity analysis.
• Controls and counter-arguments shown, not hidden, the figure that pre-empts the reviewer’s “but couldn’t it just be X?”

Order them so each one answers the question the previous one raises. That order is the order of your Results section. You have just written it without writing a sentence.

Make each figure stand alone

A figure and its caption should be legible without the main text, because that is how readers actually move through papers, they look at the figures first. So before you write the body, make every figure self-contained: axes labelled with units, groups identified, the test and sample size stated in the caption, the take-home point makeable in one read. If you cannot write a caption that says plainly what the figure shows, the figure is not finished, and often the analysis behind it is not finished either.

Now the text skeleton: one line per paragraph

With the figures fixed, outline the prose, but not in prose. Write the skeleton as a list in which each bullet is one sentence describing one paragraph’s single job. Not the paragraph itself; its purpose.

The standard frame is IMRaD, and each section has a known shape you can outline before you write:

• Introduction, a short funnel: what is known, what is not known (your gap from Part 1), and the question. Three or four paragraphs, each with one job: the field, the problem, the gap, the aim. End on the one sentence.
• Methods, already largely written, because in Part 2 you built the study against its reporting checklist. Outline it in the order a reader needs to reproduce you: design, participants, what you did, what you measured, how you analysed it.
• Results, the figure order you already fixed, turned into one bullet per figure: “Figure 2 shows the primary outcome differed by X.” No interpretation here; just what the data say.
• Discussion, a mirror of the Introduction, in reverse: what we found (the one message restated), what it means, how it fits the literature, the limitations stated honestly, and what follows. One paragraph each.

When this outline is done you should be able to read the bullets top to bottom and hear the whole argument. If it doesn’t hold together as a list, it will not hold together as prose, and now is the cheap time to fix it: reorder bullets, not paragraphs.

A worked example, from my own work

For our review of the lymphatic system in musculoskeletal disease, I did not start by writing the introduction. I started by deciding what the reader would need to see: a schematic of how lymphatic drainage works in a healthy joint; then a figure assembling the evidence of how that drainage fails as disease progresses; then a figure of the therapeutic targets that failure implies. Three images, in that order, were the entire argument, problem, evidence, consequence. Once those were drafted, the text had a job that was almost mechanical: introduce the schematic, walk through the evidence, lay out the implications. The writing went quickly not because I write fast, but because the structure had already been built and tested before a paragraph existed. The paper was published in Bone Research. The skeleton did the hard part.

The traps at this stage

• Writing top-to-bottom from a blank page. It feels like progress and produces a draft that must be demolished. Build the skeleton first.
• The figure that proves nothing. Kept because it was hard to make. The reader does not care how hard it was; cut it or supplement it.
• The data dump Results. Every panel you have, in the order you made them, with no argument. Order by the question, not by your lab notebook.
• Captions that need the text. If the figure can’t be understood on its own, it isn’t done.
• Outlining in full sentences. It tricks you into writing prose before the structure is sound. Keep the skeleton to one line per paragraph until the logic holds.

Your turn, an afternoon, no prose allowed

1. Lay out every figure and table the paper will contain, in order, each with a one-line caption of what it shows.
2. Test the sequence against your one message. Cut anything that does not advance it; move detail to the supplement.
3. Write a caption for each figure that stands entirely on its own.
4. Outline IMRaD as a list, one sentence per paragraph, stating that paragraph’s single job. No paragraphs yet.
5. Read the bullets top to bottom. If the argument holds as a list, you are ready to write. If it doesn’t, fix the list, not your weekend.

Only when the skeleton holds do you earn the right to write sentences, which is exactly where Part 4 begins.

---

Next in this series, Part 4: Writing IMRaD, section by section. Now that the skeleton holds and the figures are built, how to write each section well, the Introduction funnel, the reproducible Methods, the Results that report without interpreting, and the Discussion that interprets without overreaching.