How to Practise Shadowing with a Partner and Give Useful Feedback
Practise shadowing with a partner using one target, one evidence-based comment, a model check and one retry—then swap roles.
Agree on one target, let one learner shadow while the other listens only for that target, verify uncertain feedback against the model, retry once, then swap roles.
“Sounds good” feels nice. “Your pronunciation is wrong” feels awful. Neither tells you what to do next.
Useful feedback: “I heard ___ at ___. In the model, I hear ___. Let’s replay it. On the retry, try ___.”
When you are unsure: “I’m not sure. Let’s compare the model again.”
Giving feedback is practice too
The observer is not taking a break. They are doing a different pronunciation task: listening deliberately for one feature and deciding what evidence is strong enough to mention.
That role can be useful in its own right. In a study of 96 L2 learners of German, Martin and Sippel compared teacher feedback, giving peer feedback, receiving peer feedback and a control condition after pronunciation training. The teacher-feedback and peer-feedback-provider groups improved more than the peer-feedback-receiver group on comprehensibility ratings, with a slight provider advantage in that specific study. See the study.
That does not mean giving feedback is always better than receiving it. It does mean the observer role can involve real learning. So switch roles instead of letting the more confident speaker become the permanent unpaid professor.
A much smaller 2022 study of five mixed-proficiency EFL dyads also reported improvement for both feedback givers and receivers while showing that peer-feedback interactions can unfold in different ways. See the study. The practical lesson here is modest: both roles deserve attention.
Step 1: choose one listening target before the take
If the observer is listening for “pronunciation,” they can find twenty things. If they are listening for one contrast word, one thought-group boundary or one place where the performer falls behind, the feedback can actually lead somewhere.
Useful partner targets include:
- Did the learner stay with the model or fall behind at one specific point?
- Did the intended contrast word sound more prominent?
- Did two thought groups stay connected?
- Was there an extra pause inside a phrase?
- Did the request or question broadly carry the intended communicative shape?
- Was one key word hard to understand on first listen?
Keep technical diagnosis off the target list unless one of you genuinely has the relevant expertise. “Does the tongue need to move three millimetres?” is not a good ordinary partner-shadowing job.
A 2026 study of learner pronunciation feedback found that feedback specificity and task alignment varied with the pronunciation target, feedback type and learner proficiency. See Hua’s study. That supports giving the observer a narrow job instead of inviting unrestricted correction.
Original practice example
I didn’t say Friday; I said Thursday.
The correction depends on a clear contrast between Friday and Thursday. “I didn’t said Friday” is wrong because after didn’t, use the base form say.
Use this pattern when correcting dates, names or other details.
Partner target: listen only for the contrast. Useful comment: “I heard Thursday clearly, but Friday was less prominent. Let’s replay the model.”
Explore more language-learning guides in Media-Based Language Learning.
Step 2: observe before you advise
The observer’s first job is to report what they heard, not why they think it happened.
- Observation: “I heard a pause between go and over.”
- Unverified diagnosis: “You pause there because your mouth is too tense.”
The first comment gives the performer something checkable. The second jumps from sound to cause without evidence.
Evidence before advice.
Original practice example
Could you go over that last point again?
Go over can mean review or explain. “Could you go again over that last point?” is interpretable but unusual for the neutral request; go over … again is more natural.
Use it in meetings and classes when you want something explained again.
Observer target: did a pause split go over? Say what you heard before suggesting a repair.
Step 3: verify uncertain feedback against the model
Your partner’s confidence is not evidence. Luckily, you both have the model.
If the feedback depends on what the source speaker did, replay that exact moment. If you recorded the learner’s take, replay only enough to compare the target. Do not turn the session into a courtroom argument about who remembers the audio correctly.
- “I think the model stresses this word more. Let’s replay it.”
- “I’m not sure whether that pause is in the source.”
- “I heard the model keep these words together — do you hear that too?”
- “Your wording is different from the model, but let’s first check whether it is still natural and whether exact wording was even the target.”
Original practice example
I thought the deadline was next Tuesday.
In a correction, next Tuesday may carry the important prominence. “I thought deadline was next Tuesday” is non-standard for this intended singular countable use because deadline needs a determiner.
Use it when correcting a schedule misunderstanding.
Useful partner language: “I think the model gives next Tuesday more weight, but I’m not sure. Let’s compare.”
Step 4: make feedback end in one retry
Useful feedback should make the next attempt simpler, not give the performer a lecture to remember.
- one observation;
- one model check if needed;
- one small retry instruction;
- one retry.
If the retry still does not change the target, that is useful evidence. Simplify the line, choose a different repair or leave the problem for a more knowledgeable source. You do not need to keep giving increasingly imaginative diagnoses.
Then transfer the target beyond the copied line
Before swapping roles, let the performer make one small change to the language. This keeps the pair session from ending at exact imitation.
Original practice example
If you’re free after lunch, we could talk then.
The opening condition can form one thought group and the suggestion another. We could talk presents an option. “We could to talk” is wrong because could takes the base form talk.
Use it for low-pressure scheduling.
After shadowing and feedback, transfer the pattern: If you’re free tomorrow, we could talk then. Your partner listens for the same group boundary, not a new list of errors.
Step 5: swap roles
Do not let the stronger or more confident learner become the permanent evaluator.
Swap after one short round. The new observer gets practice in selective listening. The new performer gets the speaking turn. This also helps keep feedback reciprocal rather than hierarchical.
Research does not justify saying peer feedback is universally as reliable as teacher feedback. In fact, a 2026 study of 11 pre-service English teachers found that pronunciation-pedagogy instruction was associated with more balanced segmental and suprasegmental feedback and clearer explanations. See the study. If trainee teachers can benefit from feedback training, ordinary learner partners should be comfortable saying, “I don’t know.”
Is this useful partner feedback?
Listen like a witness, not a judge
“Your pronunciation is wrong.”
Not useful. There is no target, location, evidence or next action.
“I heard a pause between go and over.”
Useful observation. It identifies an exact moment without pretending to know the cause.
“Your tongue is too far back.”
Usually avoid this in ordinary peer practice. Unless you genuinely have the expertise and evidence to observe the articulation, report what you heard and compare the model instead.
“Your accent sounds strange.”
Do not use this. It is broad, identity-loaded and gives no repair target.
“I think Friday was less prominent than Thursday, but I’m not sure. Let’s replay.”
Useful. It contains an observation, uncertainty and a verification step.
“You used different wording from the model, so it’s wrong.”
Not automatically. First ask whether the alternative is natural and whether the target required the exact phrase or only the communicative meaning.
“I couldn’t understand the key word on the first listen.”
Useful listener evidence. Report the communication problem, but do not jump straight to an articulatory diagnosis.
“Everything was better.”
Too broad. Name the target that changed.
“On the retry, keep could you lighter and together like the model.”
Potentially useful if the pair just verified that feature against the model.
What your partner should not try to diagnose
Do not ask an ordinary practice partner to decide:
- whether your accent is “good,” “bad” or native-like;
- what exact tongue, jaw or larynx movement caused a sound;
- whether you have a speech, voice or hearing disorder;
- whether one legitimate accent or pronunciation variant is wrong merely because the model differs;
- every sound-level difference in one take.
A 2025 systematic review of shadowing research reports generally positive evidence for broad outcomes such as comprehensibility, intelligibility, fluency and aspects of prosody, while evidence for individual segmental sounds is less conclusive. See the systematic review. That is another reason a partner session can profitably focus on observable delivery and listener evidence without pretending to solve every phonetic detail.
Be a second set of ears
Your partner does not need to know everything about English pronunciation. They need to know what they are listening for, what they actually heard and when to verify instead of bluff.
Choose one target. Observe. Replay when uncertain. Give one actionable comment. Retry. Transfer. Swap.
Evidence before advice.
Sources
- Martin and Sippel, “Is giving better than receiving? The effects of peer and teacher feedback on L2 pronunciation skills”
- Dai and Wu, “Mobile-assisted peer feedback on EFL pronunciation: Outcome effects, interactional processes, and shaping factors”
- Sonsaat-Hegelheimer, “Pre-Service English Teachers' Peer Corrective Feedback Practices in L2 Pronunciation”
- Hua, “Understanding the Nuances: The Quantity and Quality of Learner Feedback in Second-Language Pronunciation”
- Whitworth and Rose, systematic review of shadowing for L2 pronunciation teaching