Revising, and finding out whether it worked
85 min
Two hosts talk the lesson through. The voices are synthetic; the script was written from this lesson and checked against it, and asserts nothing the lesson does not.
- Separate revising from editing, and say what each one is for and when it happens
- Run a paraphrase test on a page of your own writing with a real reader
- Say what a readability score does and does not tell you about whether you were understood
You have taken a document through eight lessons. You know where its point should go, what its sentences should look like, which of its words are doing nothing, and how it should be shaped for someone who is scanning.
You still don't know whether anyone understands it.
That sounds like a small remaining gap and it is the whole thing. Every judgement you have made so far has been made by the one person in the world who cannot read the document fresh: you know what it was supposed to say, and your own knowledge fills every hole as you reread. Lesson 4 used that fact to explain why you can't feel a long wait between a subject and its verb. It applies to everything else too.
So this last lesson does two things. It separates revising from editing, which get treated as one thing and aren't, and then it asks the question almost nobody asks: did it work?
Revising is not editing
These get used interchangeably, and treating them as one thing is the commonest way a draft gets worse instead of better.
Revising is structural. What is this document for, is that still what it is doing, is the argument in the right order, does that whole section need to exist. Revising can delete a third of a draft.
Editing is local. The sentence, the word, the comma. Everything lessons 3 to 6 taught you.
Revising comes first, and the reason is not taste. If you edit first, you'll spend an hour polishing sentences in a section you're about to cut. That work isn't merely wasted; it makes the cut harder, because nobody likes deleting a paragraph they've just perfected.
You have an hour with a draft, and a feeling that its third section shouldn't be there. What do you do first, and what do you leave until last?
Show the answer
First, decide about the section. That's revising: what the document is for, whether the section serves it, whether the argument is in the right order. If it goes, it goes now, before you've spent any of the hour on it. Last, the sentences and the words, which is editing, and which lessons 3 to 6 taught you. Polish the third section first and you'll have made it harder to delete and spent time on prose nobody will read.
Nancy Sommers made the same complaint about how the subject was taught. In her 1982 report to the National Institute of Education she examined fifteen composition textbooks and found they "yield 23 terms used synonymously with revision". What those books teach, she found, equates revision with cleanliness, "to groom, to polish, to order, and to tidy-up one's writing", and places it structurally as "that interlude after you finish writing the first draft and before you type the paper."
Both halves are wrong, and between them they describe what most people actually do.
What experienced writers do differently
Sommers studied this directly. Eight first-year students and seven experienced adult writers each wrote three compositions, rewrote each of them twice, suggested revisions to a composition by an anonymous author, and were interviewed three times. The students wrote in class; the adults wrote at home or in their offices.
Fifteen writers can't establish a fact about students and experts in general, and this lesson doesn't claim one. What a study this size gives you is a distinction worth having and a vocabulary for it, which you then check against yourself. Take the finding below as something to look for in your own drafts rather than as a measured difference between two populations.
Her central finding, in her own words:
"The student writers have operational procedures for revising and reasons to explain their procedures, but ... they have not codified or synthesized the procedures into a theory of the revision process. The experienced writers, on the other hand, have a codified set of principles about how their revision process works."
Read that carefully, because it is not the finding people expect. The students weren't lazy and they weren't without method. They had procedures and they could explain them. What they lacked was any account of what revision is for, which meant they could not tell which procedure a given draft needed.
The second half is sharper still:
"For the experienced writers revision was not a stage but rather a process that occurred throughout their writing. Their first drafts were already the results of an elaborate revision process."
So the neat sequence, write then revise then edit, describes neither group well. Sommers says the evidence "calls for a rejection of the linear stage theory", and proposes instead that writers "sense dissonance, tolerate dissonance, and resolve dissonance".
That sits awkwardly next to this lesson's own "revise before you edit", so be clear about what each claim is for. Sommers is describing how composing actually happens: skilled writers don't wait for a finished draft to start revising, and their first drafts are already the product of revision. "Revise before you edit" is not a rival theory of composition. It's advice about what to do with a draft that is already in front of you, and its whole content is: don't polish sentences in a section you're about to cut. Those two are compatible, and holding both is the point. If you notice you're revising while drafting, you're not doing it wrong; you're doing what the experienced writers in the study did.
Two of her subjects, in their own words. What is the difference between them?
Show the answer
Michael, a student. He rewords when a word is "too childish" or when he has repeated it, and his "strategy is to use a thesaurus to improve the 'sophistication' and thus the appropriateness of his writing".
Debra, an experienced writer. "Rewriting means the deep structural changes, the content. Revising means the surface changes, word and phrases. Rewriting is the global process, revising the local process. In rewriting, I am more audience oriented."
Michael is working at the level of the word, and his test is how the word sounds. Debra has two named activities with different scopes, and the larger of them is oriented to a reader. She is not more talented in that quotation. She has a distinction he does not have, and the distinction tells her what to do next.
Note also what Michael's thesaurus habit costs, which lesson 6 warned about. Swapping a repeated word for a synonym leaves the reader working out whether the two words mean the same thing.
To analyse what her writers did, she classified every change by two things at once: the operation, which is "deletion, substitution, addition, and reordering", and the level, which is "word, phrase, sentence, and thema", a thema being "an extended statement of one idea".
Use it as a description rather than a checklist. Go through a revision you have made and mark each change. If nearly all of them are substitutions at the word level, you have edited and called it revising, and that is Michael. If a draft needed structural work and got a hundred word-level substitutions, the scheme shows you that in about two minutes.
The step almost nobody takes
Now the part most writing advice skips, and it is the only one that produces evidence rather than judgement.
Ask a real person what they think it says.
The US federal plain-language guide sets this out as paraphrase testing, and it's deliberately cheap. The guide's own scoping, which this course has read: "With only 6 to 9 participants, paraphrase testing will not take a lot of time." It's recommended for short pages and survey questions; for a long form or a whole service, the same guide points at usability testing instead, which is a bigger undertaking than this lesson teaches.
How to run one:
- Find between six and nine people, which is the number the guidance works with.
- Give them a section and let them read it. Not the whole document. One section at a time.
- Ask them to tell you in their own words what that section means.
- Write down what they say, in their words, and don't correct them.
- Ask what they'd do next, and what they think the writer wanted. Never ask "was that clear?", because a question answerable with yes or no gets you a yes and tells you nothing.
Steps 4 and 5 are this course's own wording rather than quotations. The guide says more about both; what's above is the short version that fits a working writer with one afternoon.
The guidance works with six to nine, and it's right that you learn more from six than from one: three readers making the same mistake points at a sentence, where one reader making it might be one reader.
This course's advice, and it's a judgement rather than a finding: do it with one rather than not at all. Nothing in the guidance says one is sufficient, and this lesson isn't going to pretend otherwise. What it says is that the step people actually skip is the first one, and a single real reader will change how you write in a way that no amount of rereading will. Start with one, and get to six when the document is worth it.
That fourth step is the one people can't do. Someone misreads your sentence, and every instinct says explain it. The moment you explain, the evidence is gone: you now have a reader who understands your document because you told them what it says, which isn't a thing you can ship.
The misreading is the finding. It is the entire point of running the test.
One test, written out
Here is a section of a letter about a council tax reduction, and what happened when one person read it.
If your circumstances change in a way that affects your entitlement, you must tell us within one month of the change. Where a change is not reported within that period, any resulting overpayment is recoverable, and a penalty may be applied.
The reader was asked what that section meant. Her answer, written down in her words:
"So if something changes I should let them know, and if I don't they might fine me. I'd probably ring them when I got round to it. It sounds like the sort of thing where they'd understand if you were a bit late."
Read that and you can see the finding, and it isn't that she failed. Her paraphrase is fluent, confident, and about eighty percent right. What she has lost is the one month, which is the only number in the passage and the thing the whole section exists to communicate. She converted a deadline into a general obligation.
Now the useful question, which is not "why didn't she read it properly". Which sentence did that? Before you read the diagnosis, look at the passage again with lessons 4 and 5 in mind and find the mechanism yourself.
What in the passage's construction lost the month, and what kept the penalty?
Show the answer
The deadline arrives only after an eleven-word condition that opens on "If your circumstances change", so by the time she reached "within one month" she had been holding an unresolved condition for half the sentence. That's lesson 4. And the paragraph's stress position, the place lesson 5 says a reader puts their weight, is occupied by "a penalty may be applied", which is why the penalty survived her reading and the deadline didn't.
The rewrite follows from the diagnosis rather than from taste:
Tell us within one month if your circumstances change. If you tell us later than that, you may have to pay back anything you were overpaid, and you may be charged a penalty.
The deadline is now in the first six words, it's the only thing in the opening sentence, and the penalty has moved to where it belongs, which is after the thing the reader has to do.
The second reader paraphrased it as: "I've got a month to tell them, and if I miss it I pay it back and maybe a fine." That's the test passing, and it took eleven minutes.
A third reader paraphrases the rewritten version perfectly, then says "I'd wait and see if it makes a difference to my bill first." What have you found, and is it a writing problem?
Show the answer
You've found something real, and it isn't a writing problem. She understood the document, which is what the test measures, and she then told you what she intends to do, which the document has failed to change.
That's worth separating carefully, because it's the commonest way a test result gets misread. If she'd misparaphrased, you'd have a sentence to fix. She didn't. What she's revealed is that the letter gives her no reason to act inside the month, and she's weighing the deadline against the effort of ringing up.
Some of that is out of your hands. Some of it isn't: a sentence saying what happens to people who tell them late, with a number in it, is information the letter currently doesn't carry and she'd have used. That's a finding about content, not about wording, and it's the kind you only ever get by asking someone what they'd do next rather than only what the words mean.
A reader paraphrases your section fluently, confidently, and wrong. Which of these is the right response, and why are the others wrong?
Show the answer
(a) "Ah, no, what it means is..." Wrong, and it's the natural one. You've just destroyed the finding and taught your next question's answer.
(b) Write down what they said, verbatim, and move to the next section. Right. You have the finding. Analysis comes later, when you look at your notes and find that three of your six readers made the same mistake, which points at a sentence rather than at a person.
(c) "Was that section confusing?" Wrong twice. It's answerable with yes or no, and people who've just confidently misunderstood something don't report being confused, because they weren't. Confusion is what a reader feels when they know they are lost.
(d) Ask them to read it again more carefully. Wrong. Your document doesn't get a second attempt in the world. The first reading is the one you're testing.
What a readability score cannot tell you
So why do people reach for a readability score instead?
Readability formulas count surface features: syllables per word, words per sentence. That's what they're made of, so a better score means those counts went down. It doesn't mean a reader understood more, and you can improve a score by splitting a sentence at a comma while changing nothing about what anyone takes away.
Lesson 1 gave the deeper reason. Gooding, Berzak, Mak and Sharifi's study of 518 readers found that reader background shapes both how people read and what makes a text feel difficult to them. Difficulty is a relation between a text and a reader. A formula sees only one side of that relation, so it cannot in principle report the thing you wanted to know.
None of which makes formulas useless. A score that suddenly jumps is a signal worth investigating, in the same way sentence length was a signal in lesson 4, and a signal is not an outcome.
What people get wrong
"Revision means finding better words." This is Michael's position and it is the misconception this lesson exists to break. Word substitution is editing, and it's the last thing you do.
"I read it through again and it was fine." You're the one reader who can't test this document. Rereading your own work tests your memory of what you meant.
"Getting a colleague to read it is the same thing." Better than nothing and not the same. A colleague on your team shares your assumptions, your jargon and your knowledge of the project, which are exactly the things you are trying to test for.
"I don't have six to nine people." Use one, knowing what one buys you and what it doesn't. Six readers let you tell a bad sentence from an unusual reader; one reader can only tell you that something happened. That's still more than you had.
"Testing is for websites and forms." Paraphrase testing suits anything short enough to read in sections, which covers most of what you write. Long forms and whole services want usability testing, which the same guide covers and this lesson doesn't.
"If it needs testing, it was badly written." Every document is written by someone who already knows what it says. That isn't a flaw in the writer; it's the condition of writing.
Where this leaves the document you have carried
You've had one document open since lesson 1. Here is what the nine lessons did to it.
- Lesson 1 made you measure how much of it a reader was ever going to take in, and name who that reader was.
- Lesson 2 moved its point to where they'd be when they stopped.
- Lesson 3 put a character in the subject of its sentences and the action in the verb.
- Lesson 4 closed the gaps between those subjects and their verbs.
- Lesson 5 put old information at the front of each sentence and new at the end, so the paragraphs held together.
- Lesson 6 cut what was left, by named kinds, without cutting the joins lesson 5 built.
- Lesson 7 gave it headings a skimmer could aim at, and turned the parts that weren't prose into lists and tables.
- Lesson 8 is the reason you can say why you did any of that, and which of the rules you were taught you can ignore.
- Lesson 9 is the only one that told you whether it worked.
That document is now the course's project. project-one-document.md asks for it, along with the record of what you changed at each stage and what your reader said, which is the part that proves the rest. Finish it, and then send the document to the person it was always for.
Practice
Allow forty minutes for this. It's the last exercise in the course and the only one that needs another human being. Don't substitute imagining what they'd say.
Take one page of the document you have carried since lesson 1, in the state the last eight lessons have left it. Split it into two or three sections.
Find one person who does not know the subject and did not help you write it. Tell them you are testing the writing, not them, and that there are no wrong answers because you are the one being tested.
One section at a time, let them read it, then ask them to tell you in their own words what it means. Then ask what they would do if they received this document.
Write down what they say in their words. Do not correct them. Do not explain. Do not defend a sentence. If this is hard, that difficulty is worth noticing, because it is the same instinct that made you sure the document was clear.
Afterwards, for every gap between what they said and what you meant, find the sentence that caused it. Usually it is one sentence, and usually it is not the one you would have guessed. Diagnose it with the tools from lessons 3 to 6: is the character missing, is the point in the wrong place, did a cut remove a join?
Fix those sentences and nothing else, then write two or three lines recording what you learned about your own writing rather than about this document. That is the part that transfers.
If nothing needed changing, say so, and say how you know. A test that always finds a problem is not a test. But before you record that result, check that you asked for a paraphrase and not for a verdict.
Where this leaves you
The course ends here, and it ends deliberately with the only tool in it that does not depend on trusting the course.
Everything else you have been given is a claim about readers, from research this course read for you and named so that you can check it. Lesson 8 showed you how to check a claim like that, including these ones. But paraphrase testing does something the rest cannot: it tells you about your writing and your readers, this week, and it will go on telling you long after you have forgotten which principle was number four.
The project pulls all of it together. One real document, taken through diagnosis, rewriting and a paraphrase test, with a record of what the reader actually said, including the parts that were inconvenient.
Go deeper
- Sommers' 1982 report, free and complete at ERIC. Her coding scheme in the middle of it is the most useful thing in this lesson's sources, and reading fifteen writers' actual revisions is worth more than any summary of them.
- The digital.gov plain-language guide's testing section, which covers paraphrase testing alongside the usability testing this lesson doesn't teach, and is public domain.
- Gooding et al. (2021), for the study behind the claim that difficulty is a relation between a text and a reader. This course has read its abstract.
- Then the thing that beats all three: run the test again next month, on something that matters, with someone who isn't a colleague.
Sources
- Nancy Sommers, Revision Strategies of Student Writers and Experienced Adult Writers, report to the National Institute of Education, 12 June 1982, ERIC ED220839. The method and sample, the central finding about procedures against a codified theory, the rejection of the linear stage theory, the dissonance model, the four operations and four levels, the textbook analysis, and the quotations from Michael and Debra. A note on this citation: Sommers' better-known 1980 article of the same title has not been read by this course, because the freely available copy is a scan with no extractable text. Secondary summaries of that article commonly say the study used twenty students and twenty experienced writers. The 1982 report's own abstract says eight and seven, so this course reports the figures it has read and not the ones in circulation.
- Digital.gov, paraphrase testing, part of the US federal plain-language guide's "Test for understanding" section. All quoted guidance on how to run a test, how many participants, and what not to do.
- Sian Gooding, Yevgeni Berzak, Tony Mak and Matt Sharifi, "Predicting Text Readability from Scrolling Interactions" (2021), 518 participants, on reader background shaping perceived difficulty. Read at abstract level only.
Check your understanding
This lesson has a 6-question quiz. Pass it and the questions come back on a schedule in Review, so what you learned stays learned. Your progress is saved in your browser; no account needed.