Early in my career, an editor returned a draft of mine with one kind of edit: a pencil line drawn across the page in four places. No words changed. No sentences cut. Just four clean slashes, and beside the first one, in the margin, the word "two."
I had written one paragraph where there should have been two, four times over. Each block was carrying a pair of ideas, and the slash showed where the second idea started. It remains the most useful edit I have ever received, because it taught me the rule that hides under every question about paragraph length.
The real rule: one idea per paragraph
A paragraph is a unit of thought, not a unit of length. "How long should a paragraph be?" is a malformed question. The honest version is "how much room does this idea need?" A paragraph should run exactly long enough to deliver one idea and its support, then stop.
The test takes ten seconds. Read a paragraph and state its point in a single sentence. If you can, check that every other sentence serves that point; the ones that do not are either your next paragraph or a cut. If you need two sentences to state the point, you have two paragraphs that were merged in the dark.
Length follows automatically. A simple idea with one piece of support might take two sentences. An idea that needs an example, a complication, and a consequence might take seven. Neither is wrong in isolation. The sin is the paragraph with a third idea hiding behind the second because you were not ready to leave the first.
Why screens want shorter paragraphs
Print and screens reward different paragraphing, and the reason is physical: line length and visible page height. A 150-word paragraph on paper is a tidy block. The same paragraph on a phone is one unbroken column of gray with no white space to rest the eye.
Eye-tracking research from the Nielsen Norman Group famously found that web readers scan in an F-shaped pattern: they read the opening lines, then sample less and less as the page descends. Dense blocks suffer most, which means your carefully developed eighth sentence is being read by almost nobody. Online, short paragraphs are not a stylistic tic; they are a delivery mechanism. The white space creates the entry points a scanning eye needs, and it is half of what makes web posts readable.
None of this argues for one-line paragraphs everywhere. It lowers the ceiling. A paragraph that sits comfortably on paper often wants one split for the screen, at the point where the supporting evidence begins.
The topic sentence tradition, and when to break it
School taught a sturdy model: topic sentence first, support after, optional wrap-up at the end. The model survives because it works where it works. In business writing, documentation, and anything a reader will skim, put the point in the first sentence, every time. A skimming reader harvests first sentences like grain, and a buried point is a lost point.
The tradition breaks in two established places. Feature journalism uses the delayed lede: the writer opens with a scene or a detail and withholds the point, letting tension build before telling you what the story is about. Narrative paragraphing does the same in fiction, where a paragraph can be built like a small joke or a small horror, payoff last. End-weight is a real effect; readers remember what a paragraph finishes on.
The rule of thumb: when the reader is hunting for information, lead with the point. When the reader has surrendered to the ride, you can make them wait. A thriller that summarizes first and a report that saves its finding for the last line are both misusing the form.
The one-sentence paragraph, and its abuse
A one-sentence paragraph is an emphasis device. Set alone, a sentence gets promoted: the white space around it raises its voice. "Then the power went out." Alone on its line, that sentence lands like a beat in a film. Buried mid-paragraph, it is just another thing that happened.
The device has legitimate jobs: a turn in an argument, a thesis stated once, a breath after a long build. Used twice in a piece, both instances hit. Used in every paragraph, the promotion cancels out, because if everything is emphasized, nothing is. A draft that is all single-line paragraphs is usually notes pretending to be prose.
Diagnose your draft: the wall and the machine gun
Paragraph problems arrive in recognizable shapes. These are the four I see most:
| Symptom | What it looks like | Why it happens | The fix |
|---|---|---|---|
| The wall | One 150+ word block, no breathing room | Fear of fragments; ideas merged to look developed | State each idea in one sentence, then split at the seams |
| The machine gun | Ten one-liners in a row | Notes mistaken for paragraphs | Merge the related ones; give each idea one piece of support |
| The wanderer | Changes subject halfway through | Drafting momentum; nobody checked the load | The pencil slash: draw a line where the second idea starts |
| The orphan | A lone sentence doing no emphasis work | A transition cut loose from both sides | Attach it to the paragraph it serves, or cut it |
The diagnosis is mechanical enough to half-automate. Run the draft through a word counter and a paragraph counter, divide words by paragraphs, and you have your average. Far above the norm for your genre means look for walls; far below means look for machine guns. Averages hide outliers, so also find your longest paragraph and your longest run of one-liners, because the extremes are where the reader suffers. The average will not show up in a score, by the way. The classic formulas count sentences and syllables and are blind to paragraphing, which is one of the quiet gaps in readability scores: two texts can score identically while one is scannable and the other is soup. And once the blocks are fixed, the pacing inside them is its own project, which is where sentence length and rhythm comes in.
Frequently Asked Questions
How many sentences should a paragraph have?
There is no correct number, which is why the question keeps being asked. Two to six sentences covers most web writing, but that range is descriptive, not prescriptive. The paragraph is finished when the idea and its support are delivered; a new idea is always a better reason to break than any sentence count.
Does paragraph length affect SEO?
Not as a direct ranking factor anyone has demonstrated. It affects the things around ranking: whether visitors stay, finish, and return. Search engines keep telling us to write for humans, and humans bounce off walls of text, so the practical effect points one direction.
Should I use subheadings instead of splitting paragraphs?
Different jobs. A subheading announces a new topic; a paragraph develops one. A section whose subheading sits on top of a single unsplit block has solved the wrong problem, because readers inside the section still have no entry points.
Was the five-sentence rule from school wrong?
It was never a rule about writing; it was a rule about proving you could develop an idea, in a context where a teacher needed a fast way to see structure. The training wheels served their purpose. Professional writing keeps the development and drops the quota.
Can a paragraph be one word?
Yes, sparingly. A one-word paragraph is a struck gong: "No." It costs the reader a full stop and pays them a beat of silence. Spend it once per piece at most, at a moment that has earned the quiet.
Count ideas, not sentences
Print one page of your current draft tonight. Draw a line wherever a new idea starts, exactly as my editor did, and split the paragraph at the line. Keep going until every paragraph can answer one question: what is this about? If the answer takes two sentences to state, you are not done cutting. Paragraph length stops being a mystery the moment you start counting ideas instead of words.