What forty years of marking maths papers taught me about giving feedback on writing
I marked mathematics papers for the better part of four decades — first as a teacher, later as an assistant principal who still couldn’t leave a red pen alone. When I moved into editing other people’s writing, colleagues assumed I was starting over. I wasn’t. Marking a proof and editing a paragraph are the same job at the same moment: a person has shown you their thinking, and what you say next determines whether they can use it.
Forty years of handing work back to its makers taught me a few things that the feedback research later confirmed, and a few things I had to unlearn. Both lists are useful to anyone who edits, teaches, reviews, or supervises writers.
The mark is not the message
A student who gets a 4 out of 10 learns almost nothing from the 4. I learned this early, from watching what students did the moment papers came back: they looked at the number, compared it with their neighbour’s, and put the paper away. The corrections — the substance — went unread. John Hattie and Helen Timperley’s landmark review of feedback research put data on what every marker eventually notices: feedback works when it answers three questions — where am I going, how am I going, and where to next — and grades answer none of them.
Writing has its own version of the naked 4 out of 10. “This chapter isn’t working” is a grade wearing a sentence’s clothing. So is “I love this,” which surprises people. Praise without location is as unusable as criticism without location. The writer can act on neither.
Correct the work in front of you, not the person
The most robust — and most ignored — finding in the feedback literature is Avraham Kluger and Angelo DeNisi’s 1996 meta-analysis, which found that in over a third of studies, feedback made performance worse. The pattern behind the failures: feedback that drew attention to the self rather than the task. “You’re careless with algebra” degrades the next paper. “The sign flips in line three” improves it.
Editors live on the right side of that line or lose their authors. “You have a comma problem” produces a defensive writer; “this sentence has three clauses and no commas — a reader will take the wrong fork” produces a fixed sentence and, over a manuscript, a writer who starts hearing the forks themselves. The task-level comment teaches. The self-level comment merely judges, and judgment, the data says, does not compound.
Specificity is a kindness, not a cruelty
Teachers new to marking often soften their comments into uselessness because they think precision is unkind. Maths cures you of this: a student cannot be vaguely wrong about a derivative, and pretending otherwise wastes their evening. Writers deserve the same respect. The kindest comment I can put on a manuscript is the exact location and mechanism of a problem — page 41, paragraph two: the argument assumes the survey result you don’t introduce until page 60.
Vagueness, by contrast, is quietly brutal. A writer told “the middle sags” must now re-litigate every page of the middle, including the ones that were working. The imprecise editor outsources the diagnosis to the person least equipped to make it — the author, who cannot see their own blind spots, which is why they sought an editor at all.
The ratio that matters is not praise to criticism
Popular advice fixates on how many compliments should cushion each correction. Marking teaches a different ratio: the proportion of comments the receiver can act on. A paper returned with one actionable note beats a paper returned with ten observations, however accurate. Working memory is finite; a writer revising against forty margin notes retains perhaps six. So the marker’s discipline: rank what you noticed, deliver what matters most, and hold the rest for the next round. There is always a next round.
This is also why the best feedback is staged. In a thesis, structural problems come first — there is no sense line-editing a chapter that should not exist. Editors formalise this as separate passes (structural edit, copy edit, proofread), but the underlying principle is a marking principle: feedback should arrive in the order the receiver can use it.
What the red pen is actually for
Somewhere in my second decade of teaching I realised the corrections were not the product. The product was the next paper — the one I would never mark, written by a student who had internalised the questions I used to ask. The goal of feedback, done properly, is its own obsolescence.
Editing has the same endpoint. A good editor works towards not being needed: the author who once needed every fork flagged starts flagging their own. The feedback research calls this self-regulation, the highest level at which feedback operates. Markers and editors just call it the job — you correct the page in front of you, so that one day the writer corrects it before it reaches you.
Receiving it is a skill too
Markers see something writers rarely admit: most feedback fails at the point of reception, not delivery. The student who argues with every red mark and the student who accepts all of them without reading are making the same mistake — neither is treating the comments as information.
Carol Dweck and Claudia Mueller’s study on praise and setbacks maps the difference: feedback lands differently depending on whether the receiver hears a verdict on their ability or a report on their current draft.
Writers can stack the deck. Read feedback twice, a day apart — the first reading is emotional and inaccurate. Sort comments into agree, disagree, don’t understand and interrogate only the third pile first; misunderstanding is cheaper to fix than disagreement. And when multiple readers flag the same passage for different reasons, believe the location, not the diagnoses — readers are excellent detectors of where a problem is and unreliable narrators of what it is. Forty years of marking says the crossed-out line in the margin is pointing at something real, even when the suggested fix is wrong.
None of this requires a maths degree. It requires believing what the evidence — from classrooms, meta-analyses, and every marked-up manuscript I have ever returned — keeps saying: feedback is not an event that happens to writing. It is the mechanism by which writers are built.