Peer feedback can consume 25 minutes and leave a page covered in “good job” and comma corrections. The activity earns that time when every writer must produce a second draft, decide which comments to use, and explain one rejected suggestion. For a 19-year-old Taiwanese sophomore who has passed GEPT Intermediate and is preparing a 350-word English report, revision is the engine of the task. The conversation around the first draft supplies evidence for the decisions.
Untrained peer response often disappoints. That failure tells a lesson designer to teach the reviewing move, narrow the question and reserve time for action.
A comment is only a possibility
Feedback has no direct line into a learner’s future writing. The writer has to understand the comment, compare it with the intended meaning, and alter the text. A note such as “unclear” identifies a reader’s reaction but gives little help with diagnosis. A replacement sentence fixes the page while allowing the writer to accept it without thinking. A better exchange asks for evidence. The reviewer underlines the sentence where the argument became hard to follow, writes what they expected next, and names one detail they still need. The writer then chooses among moving a sentence, adding evidence, cutting a claim or leaving the passage unchanged. Those are writing decisions rather than proofreading chores.
This is assessment for learning at close range. The draft reveals current performance, a peer acts as a real reader, and the revision shows what the writer could do with the information after weighing a reaction from somebody who encountered only the words on the page. Choc’s article on spoken correction and learner repair follows a related principle. The learner performs the repair. Writing simply leaves a visible record of the before and after.
Taiwan supplied a demanding test
Hui-Tzu Min studied 18 second-year university students in southern Taiwan.
Sixteen were women, two were men, and their average age was 19. All spoke Mandarin as their first language and had passed the intermediate General English Proficiency Test. This was one intact EFL writing class, which makes it relevant to Taiwan and also limits how widely the result can travel.
The preparation was substantial: a four-hour classroom demonstration followed by a one-hour reviewer-teacher conference for each student. After training, peer-triggered changes accounted for 90 per cent of the revisions reported in the study, and more of the revisions improved text quality than before training. The 2006 Journal of Second Language Writing paper supports trained review followed by revision. It offers no support for putting pairs together with a generic checklist and hoping that useful feedback appears. An older classroom study makes the warning sharper. In two groups of first-year ESL students, students revised frequently, yet few changes came directly from peer-group response. Writers who changed more of the underlying text tended to improve more than writers who stayed at the surface. The authors raised questions about group formation and modelling. Their 1994 study is valuable precisely because the result was mixed.
Giving feedback can teach the reviewer
Writers also learn while reading another person’s draft. Cho and MacArthur randomly assigned 61 undergraduates to review papers, read the same papers without commenting, or receive no treatment. Everyone then wrote a new paper in the same genre on another topic. The reviewers produced higher-quality reports than both comparison groups. The Learning by Reviewing experiment examined university writing across disciplines, not an English-language class, so the transfer to EFL teaching remains an informed application.
Why might reviewing help?
A reviewer has to apply a criterion to a concrete paragraph. Finding a missing warrant in someone else’s argument may make the same gap easier to spot in one’s own. Distance helps. The other draft carries none of the private intentions that make a writer read what they meant instead of what reached the page.
The criterion also gives the reviewer practice in naming quality. “Add evidence” stays abstract until the reader has to distinguish a statistic that supports a claim from a colourful detail that merely sits nearby, explain the difference in plain English, and then face the same test in a new piece of their own. Reviewing can make a rubric operational one paragraph at a time. Twenty-eight undergraduates in another study received comments from one expert, one peer or several peers. Multiple-peer feedback produced more comments of every type, and non-directive comments predicted more complex repairs. Complex repairs were associated with better quality. The sample was small, and quantity alone can swamp a writer. The study of revision paths supports giving a writer more than one reader when the task includes a way to sort the responses.
Design the review backwards from revision
Start by choosing one quality that the second draft can visibly improve.
For a 350-word opinion paragraph, it might be support for the main claim. Give reviewers two prompts:
- Underline the claim you think the writer most wants you to accept.
- Put a bracket around the evidence for it, then write one sentence explaining what remains unsupported.
Two prompts are enough. A 14-item rubric invites box-ticking, especially when learners are still reading slowly. Grammar can have a separate pass after the argument has been revised. Mixing article use, paragraph unity, citation format and reader impact in one exchange sends attention in four directions. Keep the writer’s response sheet equally small. Copy one comment, mark it “use”, “adapt” or “decline”, and give a reason. Then paste the original passage beside the revised version. For a declined comment, explain the intended reader or evidence that justifies keeping the text. Disagreement is allowed. A reviewer reports a reading. The author remains responsible for the page.
Keep the teacher in the system
Peers should not become unpaid substitute teachers. They can report meaning, organisation and reader experience well because they are actual readers. They may confidently recommend faulty grammar or miss an error that a trained teacher catches. The teacher chooses examples, models two reviews, watches the first exchange, and intervenes when advice would teach the wrong form.
Pairing also changes the task.
Close friends may avoid criticism. A confident student may rewrite a quieter partner’s paragraph. Anonymous online review can reduce some social pressure, though it also removes the chance to ask immediately, “Which sentence gave you that impression?” No arrangement removes the need for ground rules. My preference is one short peer-review cycle with a required second draft over three rounds of comments on a text that never changes. On Tuesday, collect the first draft. On Thursday, review one named feature. Before the next class, submit the new paragraph with three sentences explaining one accepted, one adapted and one declined comment. The teacher can now assess the decisions, not the politeness of the discussion.