How to Give and Use Peer Feedback on Writing

Feedback fails when it addresses commas in a draft whose argument does not hold together. Work top-down: does the piece have a clear claim, does the structure support it, do the paragraphs each do one job, and only then, do the sentences work? Marking surface errors on a draft that will be restructured wastes both people's time and gives the writer a false sense that the piece is nearly finished.

The most useful comments report effects rather than issue instructions. "I lost track of who was speaking here" is more actionable than "fix the dialogue", because it tells the writer what went wrong without prescribing a solution they may not want. Writers can dispute an instruction; they cannot dispute that a reader was confused at a particular line, and that confusion is the data.

Praise is not politeness padding — it prevents the writer from deleting their strongest material during revision. Point at the exact sentence or move that landed and say why: the concrete detail, the pacing, the turn in the argument. Vague encouragement gives no protection, since a writer cannot preserve something they cannot identify.

Read once without a pen, to get an honest first impression, and write two sentences about what you think the piece is doing. Read again for structure, marking where the argument advances or stalls. Read a third time for sentences only if the first two passes found nothing structural. Most workshop comments are made during a single distracted pass, which is why they cluster on the first page.

Individual comments are unreliable; patterns across readers are not. If three people stumbled at the same paragraph, that paragraph has a problem, even if all three proposed different fixes and none of the fixes appeals to you. Sort the feedback into repeated observations and one-off preferences, then act on the repetitions first.

Feedback identifies problems more reliably than solutions. A reader who says a section drags may be right about the effect and wrong about the cause — the fix might be cutting, or it might be adding the stakes that were missing three pages earlier. Treat every suggestion as a flag, diagnose it yourself, and be willing to reject a proposed change while accepting the observation behind it.

After a workshop, resist the urge to apply every note at once, which produces a draft with contradictory pulls. Pick the largest structural change, make it fully, then re-read to see which smaller notes still apply. Many of them will have disappeared, because they were symptoms of the structural problem rather than independent faults.