Correcting students without discouraging them

Two failure modes, both common, both damaging.

The teacher who corrects everything produces a student who stops talking. Every sentence gets interrupted, so sentences stop being attempted, and the lesson turns into the student answering questions in three words while the teacher fills the silence.

The teacher who corrects nothing produces a student who is fluent in their own errors. Nothing impeded communication enough to be worth mentioning, so nothing got mentioned, and after two years the errors are fossilised and much harder to shift than they were in month two.

The useful question is not how much to correct. It is which errors, when, and how.

Which errors

A rough triage, in priority order.

1. Errors that impede understanding. If you had to work to parse it, correct it. This is the only category that is always worth the interruption.

2. Errors on the current target. If today’s lesson is the perfect tense, correct perfect tense errors and let the article mistakes go. A lesson with one target teaches one thing; a lesson with eleven targets teaches nothing.

3. Errors that are hardening. The third time you hear the same structure this month, it is becoming automatic. This is the argument for early correction of things that do not impede communication — a Spanish learner who diphthongises every vowel is perfectly understood, so nobody corrects it, so it sets.

Everything else: let it go. Slips the student can self-correct, one-off errors, things well above their level, and anything that would be the fourth correction in two minutes.

Fluency and accuracy compete for attention in real time. A student concentrating on getting the case ending right is not concentrating on saying anything. Choose which one this activity is for, and say so out loud: “This next bit is just for fluency — I’m not going to stop you.” Students relax noticeably when they know the rules.

When

Immediately for pronunciation and for the day’s target form. Pronunciation feedback loses almost all its value when delayed, because it is a motor skill and the motor memory is gone in seconds.

At the end of the turn for most grammar. Let them finish the sentence, then go back. Interrupting mid-clause destroys the thought and teaches the student that finishing is not expected.

At the end of the activity during fluency work. Note errors as you go, address three at the end. Interrupting a role-play to fix a preposition trades the thing you were practising for a detail.

Next lesson for patterns. If something recurs across a lesson, it belongs in the lesson notes and gets opened deliberately next time rather than patched on the fly.

How — the techniques, ranked by uptake

Corrective feedback research distinguishes several moves. They differ sharply in how likely the student is to actually notice and use them.

Recast — gentle, often invisible

The teacher reformulates correctly without comment.

S: I have went to the shop. T: Ah, you went to the shop. What did you buy?

Conversational, non-threatening, and frequently unnoticed — students routinely hear a recast as a confirmation of what they said rather than a correction. Fine as a background tool; unreliable as your main one.

You can make it land harder by stressing the corrected part and pausing: “You went to the shop… yes?”

Elicitation — highest uptake

Prompt the student to produce the correction themselves.

S: I have went to the shop. T: I have…? S: …gone?

This works better than anything else on this list, for the same reason retrieval beats re-reading: the student produced it, so it is theirs. It also confirms whether they actually know the form or were guessing.

Costs a few seconds. Worth them.

Clarification request — good for meaning failures

T: Sorry — you went where?

Signals that comprehension failed without naming an error. Appropriate when the problem genuinely was intelligibility, and it is honest: the student learns that this construction does not work on real listeners.

Metalinguistic clue — for rule-based errors

T: Gehen takes sein, not haben.

Efficient with adults, who generally want the rule. Useless if the student does not already have the concept, and easy to overuse into a grammar lecture.

Explicit correction — clear, blunt

T: Not “have went” — “have gone”.

Unambiguous, and it interrupts. Right when the student is genuinely stuck, when time is short, or when subtler moves have already failed twice.

Pronunciation is a special case

Correcting pronunciation is more face-threatening than correcting grammar. A grammar error is a knowledge gap; an accent error feels like a personal deficiency, because accent is bound up with identity in a way that verb morphology is not.

Four things make it survivable.

Correct the sound, not the person. “That /θ/ came out as /s/” is a technical observation about an event. “Your th is bad” is a verdict about a person.

One thing at a time. A learner producing an English sentence may deviate from the model in six ways. Naming all six teaches nothing and demoralises comprehensively. Pick the one with the highest intelligibility cost — usually stress placement or a vowel merger, rarely the exotic consonant they are worried about. See accent vs clarity.

Let the recording do it. Record the student, play it next to a model, and ask “what is different?” The evidence now comes from the audio rather than from your authority, which changes the interaction from a judgement into a joint observation. It is also more effective, because the student hears their own error from outside their own head — usually for the first time.

Check perception before blaming production. If the student cannot hear the contrast, correction is pointless and repeated correction is cruel. Ninety seconds of minimal pairs tells you which problem you have.

Say the rules out loud

Most of the anxiety around correction comes from the student not knowing what to expect. Fix that directly, in lesson one:

“I’ll stop you when something makes it hard to understand, or when it’s the thing we’re working on today. Otherwise I’ll let you talk and we’ll come back to it. If you’d rather I corrected more or less, tell me.”

Then ask again in month two. Preferences vary enormously — some adults want every article fixed, some want to be left alone until they finish a thought — and the mismatch between what a teacher does and what a student wants is a common reason students quietly leave.

Make the correction outlive the lesson

A correction that exists only as speech during the call has a half-life of about an hour.

The corrections block in lesson notes — two columns, what you said / better, three to five items, no rules — is what turns a moment of feedback into something reviewable. Then open the next lesson with it. Students study what gets checked; if corrections are never revisited, the student correctly infers that they did not matter.