A teacher writes twelve errors on a slide after a 25-minute speaking task, and learners copy the corrections, close their notebooks and leave even though the lesson has removed the one event that could turn accurate feedback into speaking: another attempt. For a selected grammar or pronunciation target, correction should arrive while the learner can still use it in the next turn.
This article is for a teacher running an online class of four Taiwanese B1 learners. The aim is precise: correct one recurring form during a task cycle and leave enough task for the learner to try it again.
Timing includes the next action
“Immediate feedback” can describe several schedules. A teacher may respond seconds after an error, interrupt after a task round, or give feedback in the same lesson after explicit instruction. “Delayed” feedback might mean five minutes later, at the end of three tasks or 24 hours after a video call. These timings create different learning conditions, so a study supporting one definition cannot settle every classroom choice.
Fu and Li studied 145 seventh-grade EFL learners working on the English past tense. The learners completed six focused communicative tasks across three treatment sessions. One group received corrective feedback in the first session and then had later opportunities to practise, whereas another received feedback in the final session after its chances to apply the correction had already passed. On an untimed grammaticality judgement test and an elicited imitation test, the earlier-feedback group showed greater development. The authors explain their design and analyses in the Studies in Second Language Acquisition article.
Fu and Li’s design compares feedback at different points in a sequence, rather than a teacher interrupting versus waiting until the end of one conversation. Its strongest practical feature may be the opportunity structure. Learners who received feedback earlier could apply it in later tasks. The final-session group could not.
Feedback needs somewhere to go.
Choose the error before choosing the moment
Correcting everything during conversation will fracture the task. Waiting until next week will separate a correction from the memory of saying the sentence. A teacher can avoid both problems by selecting a target before the activity.
Suppose the task elicits past-tense questions about a weekend trip, and the teacher listens for Did you + base verb? while ignoring article slips because one narrow target can remain active across the full exchange. When a learner asks, “Did you went by train?”, the teacher can say, “Did you go?” with a rising prompt, let the learner repair the question and keep the exchange moving. If the same pattern returns, pause after the pair completes its 90-second round, show two anonymous examples and ask everyone to correct them.
Round two must still require past-tense questions. Change the partner, hide two details or add a new destination card. The learner now has a reason to retrieve Did you go...? under communicative pressure. Copying the corrected form offers recognition. Asking it to a new partner requires production.
The sequence builds on Choc Education's article arguing that spoken correction should make the learner finish the repair. Timing adds another layer: after the repair, the lesson should contain another use.
Use an interruption ladder
One timing will not suit every error. A four-step ladder gives the teacher room to protect meaning while keeping feedback close enough to use.
During the turn, intervene only when the error blocks understanding, concerns the chosen target or can be repaired in a few seconds. Between turns, recast or prompt a recurring form that deserves attention but would have derailed the sentence. Between task rounds, display two or four collected examples and run a 3-minute repair. After class, send a short record for patterns that need explanation or individual work.
This ladder is a decision rule, not a quota. In a fluency task about an emotionally difficult topic, even a quick correction may damage willingness to speak. In controlled practice of a new past-tense form, frequent prompts may be exactly what the task promises. The communicative purpose changes the cost of interruption.
I prefer feedback between two short rounds for most classroom speaking. The first round supplies real learner language. The pause allows a focused correction without talking over anyone. The second round tests whether the correction can enter speech.
Research gives a direction with caveats
Li, Li and Qian’s 2026 research timeline maps studies of feedback timing across face-to-face tasks, text chat, video chat, writing and drill activities. Their open review of corrective-feedback timing reports a general advantage for feedback during task performance over feedback after a communicative task, while also identifying studies with no timing difference. Prior instruction, target salience, ceiling effects, sample size and feedback type can alter the outcome.
A warning from the review is one warning especially useful for teachers: researchers have used the same labels for unlike schedules, so a correction 15 seconds after an error and a correction in the first of three lessons may both be called immediate. A worksheet after one role-play and a video overlay 24 hours later may both be called delayed. Classroom claims need the actual sequence, not just the label.
One reasonable claim survives that variation. Feedback close to performance lets learners compare their own form with a correction while the attempt remains active in memory. A following task then creates a chance to apply the difference. This does not guarantee acquisition. It gives the correction a better job than sitting in a notebook.
Make the second attempt meaningfully different
Repeating the identical dialogue can turn into recitation. Keep the language demand and change the information. In round one, learner A interviews learner B about a trip to Tainan. After feedback on past questions, switch to a lost-luggage report with times of 7:43, 11:18 and 16:26. The grammar returns because the information has changed.
Pronunciation feedback can use the same cycle. If sentence stress is the target, a learner first leaves a voice message changing a meeting from Tuesday to Thursday. The teacher marks the contrastive word, models one line and asks for a repair. A second message changes room 402 to 407. The sentence frame stays familiar while the focus moves.
Keep a small tally. For each learner, record target opportunities, successful uses before feedback and successful uses after feedback. Five opportunities with a shift from 1 accurate form to 3 gives more information than “understands correction.” An immediate repair followed by another error also tells the teacher that recognition has not yet become stable production.
Cases for waiting
Some feedback needs delay. A learner giving a 4-minute presentation deserves a coherent audience, especially when the goal is organising an argument. Sensitive errors may be better discussed privately. A pattern that requires a diagram, several examples or a contrast with 華語 can wait until the task ends. Delayed feedback also allows the teacher to group common problems and protect speaking time.
Waiting works best when the lesson later reopens production. After the presentation, choose one feature, rehearse a 20-second section and record it again. If the next attempt will happen on Thursday, send one exact cue and begin Thursday's lesson with the relevant line. Delay becomes weak when the corrected form never returns.
For tomorrow's class, prepare two rounds of the same information-gap task. Pick one target that the task will naturally elicit at least seven times. During round one, note errors and intervene only for a quick repair. Spend 3 minutes between rounds on two anonymous examples. Change partners and information cards, then count what happens. The useful evidence sits in the second round, in the first sentence a learner produces correctly without waiting for the teacher's face to change.