A learner has just met the English past tense and begins a story with “Yesterday I go to Tainan.” If the lesson is designed to practise past narration, waiting until next week gives that form several chances to become the rehearsed version. Feedback belongs early in the practice cycle, close enough to the first attempts that the learner can use it in the next sentence. This claim has boundaries. A teacher should not interrupt every error in an open discussion. The strongest evidence concerns a selected grammar target inside focused communicative tasks, where learners know which form they are practising.
Timing needs a target.
Early means early in the cycle
Teachers use “immediate feedback” for several timings. It can mean an interruption within a learner's turn, a prompt after the sentence, a note between two task rounds, or feedback at the end of the same lesson. Those moments carry different costs. An interruption may break the message. A correction between rounds preserves the first performance and still leaves another attempt. Fu and Li studied 145 seventh-grade EFL learners learning English past tense. The learners were assigned to Immediate Feedback, Delayed Feedback, Task Only, and Control groups. The three experimental groups completed six focused communicative tasks over three treatment sessions. The immediate group received correction in the first session, while the delayed group received it in the final session. On an untimed grammaticality judgement test and an elicited imitation test, early feedback produced larger learning gains.
The timing contrast in that experiment spans an instructional cycle. It does not prove that a teacher should cut into every sentence after 0.8 seconds. The learners were adolescents working on one target structure, and the outcome measures sampled that structure. Free conversation, pronunciation, vocabulary, and adult professional English may behave differently. A 2024 research timeline on feedback timing makes the mixed evidence visible. Several studies favour immediate feedback, especially when learners can apply it soon. Other comparisons find no timing difference, including a 2023 study of immediate and delayed prompts. Written feedback follows another pattern in some studies. “Correct everything now” is much broader than the evidence permits.
Keep the boundary narrow.
Preserve one more attempt
The practical unit is a task round. Imagine a pair activity in which one learner tells a 64-second account of a missed train and the partner reconstructs the sequence. The target is past tense. The teacher listens during Round 1 and records two examples. Before Round 2, the teacher says, “You said ‘I miss the train.’ Give me the past form.” The learner repairs “missed,” then tells the account to a new partner. Feedback now has somewhere to go. The second telling matters because a correction at the lesson's last minute can be understood and then left unused. An immediate prompt inside every turn creates the opposite problem: the learner may hand control of the story to the teacher. Between-round feedback gives the speaker a complete first attempt, a narrow correction, and another communicative reason to speak. Choc's article on leaving another speaking attempt after correction covers that repair step. Timing sharpens the lesson plan further. Place the correction before the remaining attempt, while the target form and communicative situation are still active.
One target is enough for a round. If the lesson focus is regular past endings, comments on articles, word choice, /r/, and eye contact can wait. A learner who receives five directions has little chance of implementing any one of them during a 64-second retell.
Match the move to the error
Feedback can become lighter as learners gain control. The teacher might move through four responses across a course:
- Give the form with a short contrast: “Today I go; yesterday I went.”
- Repeat up to the error and pause: “Yesterday I...”
- Point to a timeline or a written “yesterday” cue.
- Mark the error privately and ask for a repair between rounds.
Teachers can select among these four responses according to the learner and error. An irregular form such as “went” may need the answer. A familiar regular verb may only need a pause. If the learner cannot explain the story because the topic is unfamiliar, adding a grammar prompt may diagnose the wrong problem. The social setting also changes the choice. Public correction can be cheap in a rehearsal where everyone expects it. During a personal account in front of twelve peers, the same move can narrow what the speaker is willing to say. Small classes give a teacher room to notice who welcomes a prompt and who needs a note after the turn.
My preference is to tell learners the correction contract before the task: “I will listen for past tense. I may pause you once, and I will give you one note before the second telling.” The agreement removes surprise and protects the message from constant inspection.
Agreement matters.
Record patterns across lessons
One early correction reveals little about durable learning. A learner may repair “go” today and produce it again next Tuesday. Track the same feature across at least three dates, using comparable tasks. The record can be small: target, first attempt, prompt used, repair, next-lesson sample. Do not turn that log into a count of every mistake. It should answer whether the chosen form is becoming easier to retrieve during meaning-focused speech. If “went” appears independently in Round 2 and again six days later, feedback has moved beyond short-term imitation. If it only appears after the teacher supplies it, the learner still needs retrieval practice.
There is also a clean stopping rule. When a form appears accurately in several spontaneous samples gathered across different partners and topics, stop hunting for it and choose another pattern, since repeated correction after stable independent use merely teaches the learner to expect surveillance. Learners deserve stretches of conversation in which yesterday's target has returned to ordinary language.
Then leave it alone.
For tomorrow's lesson, take one focused speaking task and draw a line between its two rounds. Write the target grammar above that line and one likely prompt below it. Correct no more than one instance before the learner speaks again.