Correct fewer writing errors and follow them across drafts. A page covered in corrections may look thorough, yet it gives the learner too many decisions at once: interpret each mark, recall the rule, repair the sentence, and somehow apply the same lesson next week. Focused written corrective feedback takes a narrower route. The teacher selects one or two recurring language features, marks them consistently, and checks whether accuracy changes in a fresh piece of writing. This article is for a teacher reading 18 first-year university paragraphs on a Tuesday night. The claim concerns language accuracy, especially rule-governed forms. Content, organisation, evidence, and voice still need attention on their own schedule.
Red ink hides the signal
Consider a 176-word paragraph with missing articles, verb-tense shifts, awkward word choices, comma splices, and an unclear main claim. Correcting every visible problem produces a dense map of local repairs. The learner may copy the teacher's changes into draft two and still use articles in the same way on Friday's new topic. Revision happened on the page. Transfer remains unknown. This is partly a problem of attention. Each correction competes with the others, while the teacher has little evidence about what the learner noticed. A neat rewritten sentence may come from copying. A short explanation in the margin may be read once and forgotten. To find learning, the teacher needs another occasion where the same form appears and the old sentence is unavailable.
One paragraph cannot carry six lessons.
What focused feedback found
Younghee Sheen studied 91 adult intermediate ESL learners from varied first-language backgrounds in an experiment that targeted one linguistic feature, English articles, and compared direct correction, direct correction with metalinguistic information, and a control condition across immediate and delayed measures of learning. Both feedback groups outperformed the control group on immediate measures. On delayed measures, the group receiving metalinguistic information did better than the direct-correction group. Learners' language-analysis aptitude was also associated with their gains, especially when the feedback included metalinguistic explanation. That result supports a narrow feedback target and warns against a single recipe. Ninety-one learners provide useful evidence, though the target was the English article system. The findings do not establish that every grammar feature, every proficiency level, or every writing goal responds in the same way.
John Bitchener and Ute Knoch supplied a longer view in a ten-month study with 52 low-intermediate ESL students in Auckland. The researchers targeted two uses of English articles. Learners wrote a pre-test, an immediate post-test, and three delayed post-tests. Three treatment groups received different combinations of direct correction and written or oral metalinguistic explanation. Each treatment group outperformed the control group across the post-tests. Differences among the three feedback treatments were absent in that sample. The ten-month design matters because it followed performance beyond one repaired draft. Its boundary matters too. Fifty-two learners working on two article functions cannot settle the wider debate over written correction. Together, the two studies justify a careful classroom claim: focused feedback on a limited, treatable form can improve accuracy on new writing, and extra explanation sometimes helps.
The target stays narrow.
Choose a target with a future
A feedback target deserves repeated opportunities to appear. English articles work in these studies because learners meet a, an, and the across topics. A teacher might also track past-tense endings in narratives, subject-verb agreement in reports, or sentence boundaries in email writing. A rare idiom from one essay offers fewer chances to test change. Frequency alone is insufficient. The learner needs a usable rule or decision. "Word choice" is too broad. "Use make with a decision" is teachable, though it may matter in only one sentence. "Choose simple past for a finished event with a finished time" supplies a decision that can recur in a 287-word account of a weekend, a lab report, and a GEPT writing response. I would rather mark one pattern across four assignments than circle every preposition in a single paper. That preference has a cost. Some errors remain unmarked for a while, and learners may assume that silence means approval. Tell them the feedback focus at the top of the page: "This draft tracks sentence boundaries and past-time reference. Other language has not been fully corrected." The note protects the meaning of an unmarked line.
Build a feedback cycle
The smallest useful cycle has four moves:
- Collect a baseline sample and count the chosen error in its possible contexts.
- Give a correction plus a short cue that names the decision.
- Ask the learner to explain or repair two examples without copying a full model.
- Check the same feature in a later text on a different topic.
Use rates when opportunities vary. A learner who writes "the" correctly in 17 of 21 required contexts has supplied more information than a learner with two correct uses in two short sentences. Raw error counts punish longer or more ambitious writing. The denominator keeps the comparison honest. The later text should contain enough chances for the form to appear naturally. If the target is past-time reference, a description of someone's bedroom may yield almost none. A short account of a delayed TRA train, a missed connection at Banqiao, and arrival at 8:43 p.m. gives the learner decisions to make without dictating the sentences.
Keep meaning in the room
Focused correction can become tunnel vision. A grammatically accurate paragraph may still lack a claim, evidence, or a reader. Alternate the lens. One draft may receive comments on the argument and one marked language pattern. A conference can address organisation. A later lesson can compare two introductions. Accuracy tracking should occupy a defined lane rather than consume the whole road. Peer response has a different job. Classmates can report where meaning became unclear, identify a convincing detail, or test whether the sequence makes sense. Peer feedback earns its place when a second draft follows. Focused teacher correction adds a stable language target that can be checked again after that draft. Some forms also resist short rules. Preposition choice, register, and collocation often depend on phrase, genre, or context. A corpus example or several contrasting sentences may help more than a code in the margin. Lower-proficiency writers facing an urgent real-world message may need a clean direct correction now. The long-term tracking can wait until the message has been sent.
Meaning still leads.
Record one line per draft
For the next set of papers, choose one recurrent feature before opening the first file. Record correct uses, required opportunities, and one note about the learner's decision. Repeat the count in a new text 13 days later. The useful record is small: "articles in specific reference, 9/14, then 16/19." That line can guide the next comment more reliably than a page that has turned entirely red.