Write the topic as one sentence, then look for the word doing the hiding — usually ’and’, ’often’ or ’many’. That word is where two or three separate claims got folded into one. Pick the single claim you can support with a specific case, number or example, and write only that. If the sentence still needs several examples to hold together, it is still a topic, not a claim.
Flow diagram of 4 steps: Write the topic as one sentence, Circle the joining word, List what sits on each side of it separately, Keep the half you can support with one case.
Most stalled drafts share the same root cause, and it is not lack of material — it is that the writer picked a subject rather than a claim, and a subject cannot be defended or disproved. Ask ’why is this true’ of a topic like ’volunteering builds community’ and there is no single answer, because the sentence is really three or four different claims stacked under one heading: that regular contact reduces isolation, that shared tasks build trust faster than shared conversation, that visible commitment changes how neighbours see each other afterwards. Any one of those is arguable on its own. Held together as ’volunteering builds community’, none of them has to survive an argument, because the sentence can retreat to whichever part is easiest to defend the moment one part is challenged.
That retreat is also why broad topics rarely reach a finished draft. A single claim gives the rest of the piece something to build around — every paragraph either supports it, complicates it, or answers the obvious objection to it. A broad topic gives no such anchor, so each paragraph has to re-decide what it is actually arguing, which is a slower and more exhausting way to write than most people realise while they are stuck inside it.
The fastest way to locate the claim buried inside a broad topic is to write the topic as one sentence and look for whichever word is doing the joining work — ’and’, ’often’, ’many’, ’various’. That word marks the seam where two or three separate claims were folded together so the sentence could sound complete without committing to any single one of them.
Not a heading or a working title — an actual sentence with a subject and a claim, however rough.
Find the ’and’, ’often’, ’many’ or ’various’ that lets the sentence cover more ground than it can defend in one breath.
Write each half as its own sentence. They will usually turn out to need different evidence and different examples.
Pick whichever half you could back with a specific number, example or firsthand detail, and delete the other half from this piece entirely — it can become a second piece later.
Take the broad version: ’small acts of kindness change communities’. It sounds finished, but it is not a claim — it is a category with no size, no timeframe and no mechanism attached. Run it through the joining-word test: the hidden seam is ’communities’, which quietly stands in for everything from a street to a city to an entire online audience, at whatever scale makes the sentence hardest to challenge.
Narrowed, the same idea becomes: ’a five-minute favour repeated weekly changes how two specific neighbours behave toward each other, not how a whole street behaves.’ That version names a scale (two people, not a community), a mechanism (repetition, not a single act), and an explicit limit on what it is not claiming. It is smaller than the original sentence and considerably easier to write eight hundred honest words about, because every paragraph now has one job: support, complicate, or qualify that one claim.
Three signals are reliable even before a single paragraph is written. First, the sentence still needs two or three examples to make sense on its own — a genuinely narrow claim can usually stand on one strong example, because it is not trying to cover several situations at once. Second, a stranger’s obvious follow-up (’but what about X’) has no ready answer, which means X is still living inside the sentence unaddressed rather than deliberately excluded from it. Third, deleting any single clause from the sentence would not change what it claims — if a clause is doing no defensive work, it was added to sound thorough rather than to say something specific.
None of these signals require a second opinion to check. They can all be tested by rereading the one-sentence version of the topic and asking what it would take for that exact sentence to be wrong. A claim narrow enough to write about is a claim specific enough to be wrong about.
Narrowing is not always the answer. A genuine overview piece — a reference page, a glossary, a ’here is the landscape’ explainer — is legitimately broad, and forcing it into a single claim would misrepresent what it is for. The difference is intent stated up front: an overview says plainly that it is surveying a field rather than arguing a point, so a reader never mistakes its breadth for an unfinished, un-narrowed claim. The failure mode this guide is about is the opposite case — a piece that is trying to argue something specific but has not yet noticed that its subject line is doing the arguing’s job for it.
When the one-sentence version of the claim needs only one strong example to hold up, rather than two or three to cover its different cases. If you can state clearly what would make the sentence wrong, it is specific enough to write about.
No. A title is about wording and how it reads; narrowing is about which claim the piece is actually going to defend. A punchy title can still sit on top of an unnarrowed, uncommitted topic, and the piece will stall regardless of how the title reads.
Write three separate pieces rather than forcing one. Each half of a joining word like ’and’ or ’many’ usually needs its own evidence and its own examples, and combining them under one heading only weakens all three by giving each less room than it needs.
No — narrowness and length are different things. A tightly narrowed claim can support a long, detailed piece precisely because everything in it serves one argument. A broad topic produces a short piece just as easily; it simply skims several claims instead of developing one.