Writing your first paper, section by section

Writing · 4 min read

A paper is not an essay about your project. It is an argument that one specific claim is true, arranged so a busy expert can check it. Every section has a job. If you know the job, the writing gets much easier.

Write them in this order

Not the order they appear. Write:

Methods → Results → Introduction → Discussion → Conclusion → Abstract → Title

Methods and results are facts you already have — start where the writing is mechanical. The introduction is the hardest section, and you cannot write it properly until you know what you found. The abstract is last because it summarises everything else.

Methods: could a competent stranger repeat this?

That is the whole test. Not "did I explain what I did" but "could someone else get the same numbers."

Include: data source and size, how you split or sampled it, every parameter that a reader would otherwise have to guess, the hardware or instruments if they affect results, and the version of anything you didn't write yourself.

Write it in past tense, passive or active depending on your field's convention. Do not justify choices here — justification belongs in the discussion. Methods is a recipe, not an argument.

A useful trick: write the methods section, then hand it to a labmate who wasn't involved and ask them what they'd have to guess. Every guess is a hole.

Results: what you found, with no interpretation

Results states outcomes. It does not explain them, does not compare to other papers, and does not say whether they are good.

  • Lead with the number that answers your research question. Not the setup, not the sanity checks — the answer.
  • Every figure and table needs a sentence in the text pointing at what to see in it. A figure that is never discussed will be read as filler.
  • Report what you measured, including the parts that went the wrong way. Reviewers find omissions, and an omission costs more credibility than a negative result.
  • Give effect sizes and uncertainty, not just significance. "p < 0.05" without a magnitude tells the reader nothing about whether it matters.

Introduction: four paragraphs, and only four

The structure that works almost everywhere:

  1. The area matters — one paragraph, with citations, no hype. Two or three sentences is often enough.
  2. Here's what's known — the current state, compressed. This is where your reading file pays off.
  3. Here's what's missing — the specific gap. One sentence should be able to state it: "however, no existing work has X."
  4. Here's what we did — your contribution, stated plainly, plus a list of what the paper delivers.

If you can't write paragraph three in one clear sentence, you haven't found your gap yet, and no amount of writing will hide that from a reviewer.

Discussion: what it means, and where it breaks

This is where interpretation lives. Three jobs:

Explain the result. Why did it come out this way? Offer a mechanism, not just a restatement.

Reconcile with the literature. Where do you agree with prior work, where do you disagree, and why? Disagreement is interesting — do not bury it.

State the limitations honestly. Every paper has them. Naming your own limitations makes you look competent; having a reviewer name them for you makes you look careless. Say what your design cannot show, then say what would be needed to show it.

Avoid the two failure modes: a discussion that merely repeats the results in longer sentences, and one that speculates far beyond the data.

Abstract: the only part most people read

Roughly 200 words, and almost always five moves:

  1. Context — one sentence.
  2. Problem — one sentence.
  3. What you did — one or two sentences.
  4. What you found — with the actual number. This is the sentence people remember and the one most beginners leave vague.
  5. Why it matters — one sentence.

No citations, no abbreviations you don't define, no "we discuss the implications." Say the finding.

Title: specific beats clever

The title is a search query as much as a name. "A Novel Framework for Enhanced Analysis" tells a reader nothing and matches no search. Include the thing studied, the method if it's distinctive, and the finding if it fits.

Before you submit

  • Does the abstract's claim match what the results section actually shows? These drift apart during revision more often than you'd think.
  • Does every figure have a caption that stands alone without the body text?
  • Is every citation in the text in the reference list, and vice versa?
  • Have you actually read the venue's formatting rules — page limit, required sections, citation style, anonymisation?
  • Has one person who is not on the paper read the whole thing?

That last one catches more problems than any checklist. Find someone outside your project, buy them a coffee, and ask what confused them.