Research a questionPhase 5 · Write as you go
Cut Ten Percent
Most first drafts are padded. Count your words, set a target ten percent lower, and cut the filler, but keep every fact, caveat and citation a reader needs to check your work.
Purpose
First drafts carry padding: openings that clear the throat, hedges piled on hedges, adverbs that add nothing, points made twice. Padding makes a reader work harder to find what you found, and it can hide what you actually claim.
Stephen King tells the story in On Writing. In the spring of 1966, his senior year of high school, he got back a scribbled note from an editor whose name he says he can’t remember. It called his work puffy and gave him a formula: the second draft equals the first draft minus 10%. He taped it to the wall beside his typewriter, and he says it changed how he rewrote for good. In his view, a writer who can’t find ten percent to cut isn’t looking hard enough.
George Orwell gave the same advice in more general form. His essay “Politics and the English Language” (1946) argues that vague, stale wording and muddled thinking feed each other, and it ends with six rules:
- Don’t use a metaphor or figure of speech you’re used to seeing in print.
- Don’t use a long word where a short one will do.
- If you can cut a word, cut it.
- Don’t use the passive voice where you can use the active.
- Don’t use a foreign phrase, a scientific word or jargon if there’s an everyday English word for it.
- Break any of these rules sooner than write something truly awful.
The ten percent is a target, not a law. It works because it forces you to reread every sentence and ask whether it earns its place.
When you need it
You’ve finished a draft and it reads slower than it should.
A paragraph says the same thing twice in different words.
Your claims come wrapped in “it seems that perhaps” and “may possibly.”
You’re over a word limit for a journal, a grant or a class.
You’re about to share your findings with someone who has five minutes, not an hour.
How to do it
1Count, then set the target
Get the word count from your editor (most show it in the status bar or a Tools menu). Multiply by 0.9 and write the target at the top of the page. A 2,000-word draft becomes 1,800. A number makes the job concrete: you stop when you hit it, not when you get tired.
2Mark what must stay
Before you cut anything, mark the parts a reader needs to check your work: every number, every source and citation, and every condition on a claim (who was studied, when, where, how it was measured, what it can’t show). These are not padding. In research writing, the caveats are often the most accurate part of the claim.
3First pass: whole sentences
Look for sentences that announce what you’re about to say (“It is worth noting that this section is concerned with…”), sentences that restate the one before (“In other words…”), and summaries of things the reader just read. Cut them. This pass usually gets you most of the way.
4Second pass: phrases and words
Now go line by line. Collapse stacked hedges into one: “seems to perhaps suggest it may possibly” becomes “may.” Cut adverbs that add nothing, like “really,” “actually,” “clearly” and “very.” Swap long phrases for short ones: “in order to” becomes “to,” “a total of 412” becomes “412,” “over a period of six weeks” becomes “over six weeks.”
5Third pass: who did what
Find the passive constructions and ask who the actor is. “The data were analyzed” hides who analyzed them; “We analyzed the data” tells the reader, and it’s shorter. If you don’t know who did it, that’s a gap in your research, not a style problem. Find out.
6Check against the original
Recount. Then read the cut version next to the first draft and confirm that every fact, number, caveat and citation you marked in step 2 survived. If cutting changed what a claim means, put the words back.
Try it
Your task
Here is a paragraph from an imaginary library report. It is 187 words long. Cut it by at least 10% (to 168 words or fewer) without losing a single fact or caveat. Then count your version.
It is worth noting at the outset that this section of the report is concerned with the question of whether extending the library’s opening hours had an effect on how much it was used. In order to examine this question, a survey was conducted over a period of six weeks in March and April 2025 at the Eastside branch. A total of 412 visitors were asked how often they visited. The results seem to perhaps suggest that the longer evening hours may possibly have led to a somewhat higher number of visits among people who work during the day. Visitors who reported having full-time jobs said that they came an average of 2.1 times a month after the change, compared with an average of 1.4 times a month before it. In other words, working visitors appeared to come more often once the library stayed open later. However, it should be clearly emphasized that these figures are self-reported, that they come from only one branch, and that they cannot really tell us whether the same thing would actually happen at other branches with different neighborhoods and different users.
Show a hint
Before you cut, list the facts and caveats that must survive. There are about ten.
Show the answer and the path
One possible cut, at 102 words (down from 187, a cut of 85 words or about 45%):
This section asks whether longer opening hours changed how much people used the library. Over six weeks in March and April 2025, a survey asked 412 visitors to the Eastside branch how often they came. Visitors who reported full-time jobs said they came an average of 2.1 times a month after the change, up from 1.4 before it. The longer evening hours may have led to more visits from people who work during the day. But these figures are self-reported and come from one branch, so they cannot tell us whether the same would happen at branches with different neighborhoods and users.
Both counts split the text on spaces, so “2.1,” “self-reported” and “full-time” each count as one word, as they do in most word processors.
What was cut, and why:
- “It is worth noting at the outset that this section of the report is concerned with the question of whether” became “This section asks whether”: throat-clearing.
- “In order to examine this question,” “a total of” and “over a period of”: long phrases with short equivalents.
- “seem to perhaps suggest… may possibly have led to a somewhat higher number”: four hedges stacked on one claim. One “may” keeps the uncertainty. The numbers in the next sentence replace “somewhat higher.”
- “In other words, working visitors appeared to come more often…”: the whole sentence repeats the one before it.
- “clearly,” “really,” “actually,” “it should be emphasized”: adverbs and emphasis that add nothing.
- “different neighborhoods and different users” became “different neighborhoods and users.”
What stayed: the dates, the branch, the sample of 412, the averages of 2.1 and 1.4, “reported” (the jobs are self-described), “may” (the survey can’t prove the hours caused the change), and all three caveats at the end: self-reported, one branch, and no basis for generalizing.
One thing the cut couldn’t fix: “a survey was conducted” hid who ran the survey, and the paragraph never says. The cut version stays neutral (“a survey asked”) rather than invent an answer. In your own writing, name who did it.
Common mistakes
Cutting the caveats. Deleting “self-reported, from one branch” saves four words and turns an honest finding into an overclaim. The conditions on a finding are part of the finding.
Cutting citations to save words. A claim without its source can’t be checked. Shorten the citation format if you must, never remove it.
Cutting every hedge. One honest “may” or “about” is accurate. The mistake is stacking four of them, not using one.
Swapping precise words for short vague ones. “Median household income” is longer than “income,” and it means something different. Short words only win when they mean the same thing.
Cutting before the draft is done. Trimming while you write slows the first draft. Finish it, then cut.
Limits
Some length is load-bearing. Technical terms, defined once and used consistently, are more precise than everyday synonyms, and Orwell’s fifth rule bends when no everyday word means the same thing. The same goes for hedges: a claim that rests on one small study needs its qualifiers, and stripping them makes the writing cleaner and less true.
Readers differ. People reading in a second language, or reading quickly, often benefit from a little redundancy: a term repeated rather than replaced by a pronoun, or a one-line summary at the end of a long section. Cut for your actual reader.
Genres have conventions. Some fields expect the passive voice in methods sections, and abstracts and grant forms have fixed structures. Follow the convention where your readers expect it, and cut the padding inside it.
And ten percent is a rule of thumb from fiction. A tight draft may only lose five; a padded one may lose nearly half, as the exercise above did. The point is the reread, not the number. Orwell said as much about his own rules: you could obey all six and still write badly.
Sources & further reading
- George Orwell, “Politics and the English Language,” Horizon, April 1946 (full text, The Orwell Foundation)
- Timothy Taylor, “Stephen King: The Editor Is Always Right; To Edit Is Divine,” Conversable Economist, 2019 (quotes King’s account of the 10% formula)
- Stephen King, On Writing: A Memoir of the Craft, official page
End of page · Back to top