Classroom skills · Article

Error Correction in ELT: Correct It Now or Let Them Finish?

Practical guidance on immediate and delayed error correction, recasts, prompts, learner uptake and accurate, encouraging feedback in ELT lessons.

By Ahmed Bin Nasser, CELTA Main Course Tutor · · Last updated

The real question isn't whether to correct. It's what the correction is supposed to achieve.

A learner is telling a great story about her first trip abroad.

When I arrived, I didn't knew where to go…

You hear the error at once. You're a careful teacher, so you jump in: "Not didn't knew. Didn't know."

She repeats it. Then she stops. Then she tries to remember where her story was going.

Another learner makes the same error in a controlled past-simple exercise. This time, one quick prompt helps him fix it, and the class moves on.

Same error. Different moment. Possibly a very different decision.

That's why I don't like slogans such as "always correct" or "never interrupt fluency". Teaching doesn't work well when we replace judgement with slogans.

Before you read on: in the story above, was the teacher wrong to interrupt?

Not necessarily wrong, but probably costly. The learner was speaking to communicate. The interruption fixed one form and broke her story. In a fluency task, noting the error and coming back to it later would usually serve the aim better. In controlled practice of the past simple, the same interruption would make perfect sense.

What is corrective feedback?

Corrective feedback is any information or interaction that draws attention to a possible problem in a learner's language and may help them revise what they said or understood.

It's much more than saying "That's wrong." You can give the correct form, reformulate the sentence, ask for clarification, signal that something needs attention, or give a grammar clue.

In a classic study, Lyster and Ranta (1997) described the types of corrective feedback teachers use and how learners respond straight away. That immediate response is called uptake. Later reviews and meta-analyses (Lyster & Saito, 2010; Lyster, Saito & Sato, 2013) looked at effects on development in different contexts. Their findings are more complex than a simple "best to worst" ranking of techniques.

Start with the lesson aim

Before you correct anyone, ask: what is this task meant to develop?

Stage or taskWhat matters mostCorrection usually…
Controlled practiceAccuracy of the target formImmediate and quick: a prompt or short correction
Fluency speakingMeaning, interaction, willingness to talkSelective on the spot, mostly delayed
Pronunciation workSounds, stress, intonation while they're activeImmediate modelling so learners can adjust
WritingPatterns, register, organisationFocused comments, then learner revision

These are tendencies, not rules. If a misunderstanding stops communication in a fluency task, a quick clarification request may be exactly what's needed.

Five ways to respond to an error

Take this sentence: She go to work yesterday. Tap each card to see what the teacher says and what it depends on.

1 · Explicit correction

Teacher: "We say: She went to work yesterday."

Strength: clear and fast; nobody misses it.

Watch: learners repeat rather than retrieve.

2 · Recast

Teacher: "Oh, she went to work yesterday?"

Strength: keeps the conversation going.

Watch: learners may hear it as agreement, not correction.

3 · Clarification request

Teacher: "Sorry, when did she go?"

Strength: can be a real meaning check or a push to rethink the form.

Watch: the learner may just answer "Yesterday!"

4 · Elicitation (prompt)

Teacher: "Yesterday… she…?"

Strength: the learner does the repair.

Watch: works only if they already know the form.

5 · Metalinguistic clue

Teacher: "Yesterday is finished past. Which verb form do we need?"

Strength: links the error to a rule.

Watch: needs the terminology; slows the interaction.

+ Repetition

Teacher: "She go?" (rising intonation)

Strength: points to the exact spot.

Watch: tone matters; it can sound mocking.

Tap a card. In real classrooms these moves overlap, and tone, timing and context change how they're heard.

Lyster and Ranta grouped the learner-repair moves (elicitation, metalinguistic clues, clarification requests, repetition) as prompts, and contrasted them with moves that give the answer: recasts and explicit correction.

Think first: a learner has never met the past of go. You prompt three times; he keeps saying goed. What now?

Stop prompting. Give went clearly, model the pronunciation and let him use it in a meaningful sentence. Prompting can't work if the learner doesn't have the information. Later practice and retrieval will help it stick.

Recasts: useful, but easy to miss

A recast keeps communication going because you give a better version of what the learner meant.

L: He explained me the rules.

T: He explained the rules to you?

L: Yes! And then he said…

Illustrative exchange.

He explained me the rules. He explained the rules to me.

Did the learner hear the correction? Maybe. Or maybe she just heard you confirming her story and carried on. That's the problem with recasts: the corrective intent can be invisible.

That's not a reason to ban them. It's a reason to ask three questions:

  • Is the corrective intent clear enough here (stress on the changed word, a short recast, a pause)?
  • Does this feature matter for today's aim?
  • Will the learner get another chance to use it?

And be careful with numbers. Don't quote a percentage from one classroom study as proof that one technique works everywhere. Age, setting, the language feature, the way learning was measured and the interaction itself all matter.

Prompts: only when repair is within reach

Prompts make learners retrieve what they know instead of repeating your sentence. But a prompt is not magic. If the learner doesn't know the form, "Try again" becomes a test they can't pass. It's the same line our eliciting guide draws: don't turn it into a guessing game.

CaseWhat the learner knowsBetter response
B1: I didn't went.Has practised didn't go successfullyRaised eyebrow + "After didn't…?"
Beginner: I have seen him yesterday.Hasn't studied present perfect vs past simpleGive the form, or teach the contrast later if it's not the priority

Correct now or later?

I like to think about the cost and the value of interrupting. Walk through these questions:

  • Does the error block the meaning?Yes → deal with it now: clarify, briefly.No → next question.
  • Is it the target language or skill of this stage?Yes → next question.No → note it, or let it go.
  • Is this an accuracy stage, and can the learner repair it now?Yes → quick prompt now.No (fluency stage, or form unknown) → note it for delayed feedback; give the form later.
  • Is it widespread in the class?Yes → plan a shared feedback slot after the task.No → a quiet word with that learner.
Correct now or later? A thinking aid, not a fixed algorithm.

The answer can also differ by learner. One confident speaker enjoys a quick prompt mid-sentence; another loses the thread completely. The goal isn't to avoid challenge. It's to make correction useful. Notes you take for later come from good monitoring.

Delayed feedback: make it worth the wait

Illustrative scenario. Learners role-play customers changing a booking. While monitoring, you note these (anonymously):

  • Could we change the booking to Friday?
  • I am agree with you.
  • I want you change the date. (missing to, and may sound too direct)
  • Would it be possible to move it to the afternoon?
  1. Content firstWhat did the pairs agree? What problems did they solve?
  2. Show the languageGood and not-yet-good examples on the board, no names.
  3. Learners judgePairs decide which work and how to improve the others.
  4. ClarifyExplain only where needed.
  5. Use it againA short new exchange with the improved language.
A five-step delayed feedback stage.

Step 5 is the one teachers skip. Without it, correction may stop at recognising an error on the board. For more on this stage, see feedback after tasks.

Error lab: the aim is making polite requests. Which of these do you prioritise? A. I want you to send me the document. B. Would you mind sending me the document? C. I goed to the office yesterday.

A deserves attention: it can sound too direct, depending on the relationship and urgency. B is a great model of the target language: praise it and use it. C is a real past-tense error, but unless it blocks communication or is widespread, it's a lower priority today.

Good feedback is selective, not indifferent.

Why not correct every error?

Because correction competes with attention, time and the purpose of the task. Learners need space to communicate, take risks and try new language. (Our teacher talking time guide explains why purposeful teacher talk matters more than simply talking less.)

And more corrections don't automatically mean more learning. A board with fifteen isolated errors can be less useful than three well-chosen examples learners understand and then use.

This isn't permission to ignore accuracy. It's a reminder that teaching means prioritising.

One more caution: not every unusual sentence is an error. It may be normal in a variety of English you don't know well. Check what the learner meant before you decide what needs fixing.

Written feedback: don't edit every sentence

A learner hands in a 250-word email. You correct every article, preposition, collocation and comma. The page comes back covered in red.

You worked very hard. Did the learner?

Correct everything yourself and hand it back. Pick two or three priority patterns, give an example, and ask the learner to revise.

In a lesson on register, comment on the opening, the request forms and the closing before small spelling slips. A correction code helps if learners understand it and can act on it, but it shouldn't become another terminology test. Some errors need a direct explanation.

Feedback, noticing and uptake

Correction can help learners notice the gap between what they said and a more target-like form (see noticing in ELT). But noticing isn't acquisition. And uptake, the immediate response, doesn't prove lasting change.

If a learner repeats "She went" after you, that's one piece of evidence. If she uses went on her own in a new context next week, that's another. Neither tells the whole story, but together they guide your decisions.

That's why corrections deserve a follow-up: a later task, a retrieval activity, or simply listening for it again.

CELTA and DELTA: the questions that matter

CELTA: choose appropriate moments and strategies for responding to learner language, while protecting the aim and the atmosphere. The CELTA syllabus explicitly includes this.

DELTA: go beyond labelling "recast" or "prompt". Explain why that technique suited this learner, this error, this stage and this goal. Discuss the limits of the evidence and what you'd do if learners didn't respond.

I'd rather read one well-reasoned account of a selective correction than a paragraph proudly saying the teacher corrected every mistake.

Try it tomorrow

During one speaking task, write down two strong learner expressions and three priority language issues. Choose the issues against the aim. Then design a five-minute feedback stage with learner reflection and a chance to reuse at least one improved expression.

Afterwards, ask yourself: did I help learners develop their language, or did I mainly show that I knew the answers?

Quick FAQs

Are recasts ineffective?

No. Their effect varies, and learners may or may not recognise them as correction. Make the correction clear and give learners a chance to use the form again.

Should I correct fluency mistakes immediately?

Sometimes, when communication breaks down or a quick intervention really helps. Often, delayed feedback is less disruptive.

Is learner uptake proof of acquisition?

No. Uptake is an immediate response to feedback, not a measure of long-term learning.

Can I use direct correction?

Yes. Direct correction can be efficient and appropriate, especially for language the learner hasn't met yet.

Keep learning with Lumina

References

  • Lyster, R., & Ranta, L. (1997). Corrective feedback and learner uptake: Negotiation of form in communicative classrooms. Studies in Second Language Acquisition, 19(1), 37–66. https://doi.org/10.1017/S0272263197001034
  • Lyster, R., & Saito, K. (2010). Oral feedback in classroom SLA: A meta-analysis. Studies in Second Language Acquisition, 32(2), 265–302. https://doi.org/10.1017/S0272263109990520
  • Lyster, R., Saito, K., & Sato, M. (2013). Oral corrective feedback in second language classrooms. Language Teaching, 46(1), 1–40. https://doi.org/10.1017/S0261444812000365
  • Cambridge English. CELTA syllabus and assessment guidelines. cambridgeenglish.org

Check your understanding

6 short questions. No account needed, no score, and nothing you choose or write is saved or sent anywhere.

  1. 1 During a fluency discussion, a learner makes a minor past-tense error that doesn't affect understanding. Which response may be appropriate?
    Show the answer (no automatic checking)

    Answer: B. It protects the discussion while retaining evidence for selective feedback.

    By Lumina Trainers

  2. 2 Which teacher response is a recast? Learner: “She go yesterday.”
    Show the answer (no automatic checking)

    Answer: A. The teacher reformulates the intended message in a more target-like form.

    By Lumina Trainers

  3. 3 Why does a learner's immediate repetition after correction not prove acquisition?
    Show the answer (no automatic checking)

    Answer: B. Later use and transfer provide additional evidence.

    By Lumina Trainers

  4. 4 True or false: A direct correction can be the best response when the learner does not know the required form.
    Show the answer (no automatic checking)

    Answer: A. Prompts cannot reliably elicit information that is unavailable.

    By Lumina Trainers

  5. 5 Match each teaching situation with the most appropriate response.
    • A Metalinguistic feedback
    • B Recast
    • C Prompt/elicitation
    Show the answer (no automatic checking)

    Answer: 1–B, 2–C, 3–A. Compare the learning need and the purpose of the response; there is no benefit in using one routine for every situation.

    By Lumina Trainers

  6. 6 You have monitoring notes from a booking role-play: one strong phrase, three overly direct requests, and a minor unrelated tense slip. What will you address and when?

    Show a suggested answer

    First respond to what pairs achieved, then share the strong phrase and focus language feedback on directness and suitable requests because it relates to the aim and recurs. The tense slip may be deferred unless it blocked communication. Include a short opportunity to reuse improved expressions.

    Suggested answer by Lumina Trainers

Ahmed Bin NasserCELTA Tutor & Teacher Trainer | Lumina