[reference] — built around Simon Peyton Jones’ “How to Write a Great Research Paper,” The Craft of Research, and Zobel’s Writing for Computer Science. Opens Part B (writing). The standard sections of a paper, what each one must do, and the argument that runs underneath them — because a paper is not a lab diary, it’s a persuasive case for one claim.
The kid version first
A research paper is not a diary of “here’s everything I did all summer, in order.” It’s much more like a lawyer making one argument to a jury:
“Here’s my claim. Here’s why you should care. Here’s my evidence. Therefore, believe me.”
Everything in the paper serves that one argument. The lawyer doesn’t tell the jury about the dead ends, the coffee breaks, or the three weeks fixing a bug — only what builds the case. And courtrooms have a standard order (opening statement, evidence, closing) so everyone knows what’s coming; papers do too (abstract, intro, method, results, conclusion) so any reader knows exactly where to look for what they need.
So two ideas run this whole note: (1) a paper is an argument, not a chronology, and (2) it has a standard structure where each section has one job. Master both and writing gets dramatically easier, because you stop asking “what do I say next?” and start asking “what does this section need to do for my argument?”
The paper is an argument, not a chronology
The single most important mental shift for a new writer: you are not reporting your journey; you are convincing a skeptical expert of one contribution. Peyton Jones’ framing — “don’t describe your work, sell your idea.” This means:
- The structure follows the logic of the argument, not the timeline of the work. You might present your method cleanly even though you actually stumbled into it backwards.
- Everything that doesn’t serve the argument gets cut — the failed approaches (unless instructive), the incidental details, the ego.
- One paper, one contribution. If you have three ideas, that’s often three papers. A paper trying to argue three things argues none of them well.
Nail down your one-sentence storyline first — “We show that [X]” — and let every section serve it.
Section by section — and each one’s job
The conventional structure (with variations by field/type). What each must accomplish:
Title
Findable and honest — it should say what the paper is actually about (the contribution), using terms people will search for. Not clickbait, not cryptically clever.
Abstract (~150–250 words) — a mini-paper
The most-read part of your paper by far — many people read only this and decide whether to go further. It’s a compressed version of the whole argument: problem → gap → what you did → key result → why it matters. Every sentence earns its place; no filler, no “in this paper we will discuss.” Write it to stand completely alone.
Introduction — the most important section
If the abstract earns a read, the intro earns the rest of the paper. Peyton Jones is emphatic: state your contributions explicitly, early, and concretely. The reliable structure — a funnel:
- The problem — what’s the setting, in accessible terms.
- Why it matters — the stakes; why a reader should care.
- The gap — what’s missing in what exists (lit review in miniature): “but no one has…”.
- “In this paper, we…” — your approach in a sentence or two.
- Contributions — often an explicit bulleted list (“Our contributions are: (1)…, (2)…”). This is a gift to reviewers — it tells them exactly what to evaluate.
- Roadmap — a one-line “the rest of the paper is organized as…”.
A reader should finish page 1 understanding what you did and why it matters. If they don’t, nothing else you write can rescue the paper.
Related Work — situate, don’t list
Show you know the field and how you differ. Organize by theme, synthesize (not “Smith did X, Jones did Y…” — that’s the book-report smell). It’s often placed after the intro (or even near the end) so a wall of citations doesn’t block the reader’s momentum before they even understand your idea. Its job: position your contribution in the landscape and make your novelty undeniable.
Method / Approach — precisely, reproducibly
What you did, in enough detail that a competent reader could reproduce it. Define notation once and use it consistently. Present it in the clear logical order, not the messy order you discovered it. Diagrams help enormously here.
Experiments / Results — the evidence
First the setup (datasets, baselines, metrics, protocol — enough to trust and repeat), then the findings. The figures and tables carry the story — a reader should get the headline result from a glance at your main figure. Explicitly answer the questions you posed in the intro, and report results honestly (variance, not just means).
Discussion — what it means
Interpret: why did it come out this way, what does it imply, and — crucially — limitations and threats to validity. Stating your own limitations reads as strength and honesty, not weakness; it also pre-empts the reviewer who’d otherwise raise them.
Conclusion — recap the contribution (not a summary)
Restate what you contributed (not “in this paper we discussed…” — remind them what you showed), and point to future work (open questions — a gift to future-you and to others). Short.
References
Complete, correctly formatted, original sources (citations & tools).
Write it in a non-linear order
You do not write top-to-bottom. A common, effective order:
- Figures/tables and results first — they’re the skeleton of the evidence.
- Method — you know what you did.
- Related work — while it’s fresh.
- Introduction — now that you know the paper’s real story.
- Abstract and title last — you can’t summarize a paper you haven’t written.
The abstract and intro are the hardest and most important, so writing them last (when you actually know what the paper says) is far easier than agonizing over them first.
Point-first, everywhere (the reverse pyramid)
At every scale — the paper, each section, each paragraph — lead with the point, then support it. Abstract states the result up front; each section opens by saying what it’ll establish; each paragraph’s first sentence is its claim. Readers are busy and skim; if they read only the first sentence of each paragraph, they should still get your argument. This one habit does more for readability than any other (writing craft).
Key insight
A paper is a lawyer’s argument for one contribution, not a diary of your summer — structure follows the logic of the case, not the timeline of the work, and everything that doesn’t serve the argument gets cut. Each section has a job: the abstract is a stand-alone mini-paper (the most-read part), the introduction must make the reader understand-and-care by the end of page 1 and state contributions explicitly (bulleted), related work situates by theme, method enables reproduction, results present evidence via figures, discussion interprets and owns limitations, and the conclusion recaps the contribution. Write it non-linearly (results → method → intro → abstract last), and lead with the point at every scale.
Related
- Scientific Writing Craft — sentence- and paragraph-level clarity
- Paper Types & Writing Styles — how this structure flexes by field/type
- Literature Review — the discipline behind Related Work
- Figures & Tables — the evidence that carries Results
Software engineering
Infrastructure & security
Computer science
Hardware & engineering