What Should the Conclusion of a Research Paper Actually Do?

The proven thesis: not a copy, an upgrade

The conclusion's first job is restating the thesis — but the restatement must carry the paper's proof with it, which is what makes it an upgrade rather than a photocopy. The introduction's thesis argued on credit; the conclusion's thesis has been through the evidence and returns with receipts. Where the introduction said the 2023 feed redesign worsened body-image self-report because the algorithm amplified comparison content, the conclusion says the two cohort studies support exactly that mechanism. Same claim, different epistemic state — and that difference is what the reader should feel in the final paragraph's opening lines.

The mechanics are simple: take the introduction's thesis sentence and add its strongest support. One clause of evidence converts assertion into establishment. What dies in the upgrade is the because-clause's tentative phrasing — suggests becomes supports or demonstrates, calibrated to what the body actually showed. This move also polices the paper: if you cannot attach any evidence clause to the thesis at the end, the body did not do its job, and the problem is upstream of the conclusion. The restatement is honest precisely because it is earned.

The numbers at a glance

The four conclusion moves, in order — the range, and where each row actually shows up in coursework. Skim it first; the sections below explain why the range exists and which row applies to you.

The strongest finding, named and located

The second move picks the paper's best result and gives it a sentence of its own. Strongest means most persuasive to a skeptic — usually the largest effect, the cleanest comparison, or the finding that survived the harshest check you ran. Name it concretely: the fourteen-point gap between cohorts, the two-thirds reduction after the policy change. Vagueness here (the results were quite supportive) wastes the conclusion's most quotable sentence. This is the sentence a skim-reader takes away; make it carry a number or a named contrast.

The move does double duty as a hierarchy signal. By elevating one finding, you implicitly rank the others — and graders read that ranking as judgment, the thing that separates analysis from report. If two findings genuinely tie, name both and say why they anchor the claim from different directions. If the strongest finding is also the most counterintuitive, better — say so in the sentence (against expectation, the effect was strongest where usage was lowest). The strongest-finding sentence is the conclusion's proof of life: it shows the paper's evidence was weighed, not just accumulated.

Limits stated honestly, without self-sabotage

The third move is one sentence of limits, and its craft is calibration. Too little and the paper reads overconfident; too much and the conclusion undoes itself — a paragraph of caveats ending in the reader's doubt. The working form names the single most consequential limit and binds it to a boundary rather than a apology: the findings hold for one state's public schools; private and rural districts remain untested. A boundary sentence admits the limit while keeping the claim standing inside the territory it proved, which is the honest position every study occupies.

Choose the limit the grader would name. Every assignment has one soft rib — sample size, single site, self-report measure, short follow-up — and the conclusion that names it first controls it. The sentence pattern gives the limit its scope and then points forward, which is the bridge to move four: the limitation is exactly what the next study should test. Handled this way, the limit paragraph is not a confession but a design specification for the research that follows, and it reads as rigor rather than weakness.

The forward point: one implication, not a symposium

The final move points outward once. Either a concrete implication — what a practitioner, policy-maker or researcher should do differently given the claim — or the next study, specified enough to be checkable: testing the mechanism in remote cohorts would separate the delivery effect from the selection effect. One implication, two sentences at most. The failure mode is the symposium ending: implications for policy, practice, pedagogy and future research, each a clause of fog. Graders read the symposium as a paper that ran out of argument and switched to horizons.

The forward point is also the conclusion's only legitimate place for the larger world. Here — and only here — the paper's specific claim may connect to its field-level consequence: if the mechanism holds, feed-design choices are a public-health lever, not just an interface question. That widening is earned by the four moves before it. A conclusion that opens by widening (in today's society) is the cliche opener relocated; one that widens last, from a proven specific claim, is a landing. Order is the difference between the two, and the grader feels it as authority.

The three failure modes and their repairs

The replay conclusion summarizes the paper section by section — it wastes 200 words restating what the reader just read, and its tell is the phrase in this paper, followed by a table of contents in prose. Repair: run the four moves; a proven thesis plus the strongest finding is what a summary was pretending to do. The echo conclusion restates the introduction nearly verbatim — the tell is identical sentence openings in paragraphs one and last. Repair: the evidence clause; the claim must return changed. The new-idea conclusion introduces a source or argument in its final lines — the tell is a citation or the word another. Repair: delete and relocate; anything that needs proving belongs in the body.

All three repairs share a diagnosis: the conclusion was written as an afterthought, in the submission hour, from the top of the paper instead of the end of the argument. The five-minute version that prevents all three: proven thesis, one sentence; strongest finding, one sentence; limit and boundary, one sentence; forward point, one sentence. Four sentences, 150 words, and the form is done. Write it in the editing pass when the body's real results are known — a conclusion drafted before the evidence exists can only guess at what turned out to be persuasive.

The format check graders run before reading

Graders check mechanics first and content second — a two-minute scan that decides the paper's first impression. The list: correct font and size, double spacing throughout (no stray single-spaced paragraphs), one-inch margins, page numbers in the right corner, the required title-page fields, heading levels used consistently, and references formatted with hanging indents. Each item is trivial; missing two costs measurable points in most rubrics.

Do this scan on the exported PDF, not the working document — export quirks (shifted margins, dropped page numbers, orphaned headings) only appear there. The style switcher below shows the same source formatted in APA 7, MLA 9 and Chicago side by side, which is the fastest way to catch the format you've been half-remembering from the wrong course.