Yes, for some things. In 2025, a systematic review searched six databases and included 44 studies on shadowing and second-language pronunciation (Whitworth & Rose, 2025). Shadowing helped with rhythm, intonation and how easy learners were to understand. For individual sounds like th or r/l, the results were inconclusive.
That split is useful, because it tells you what to aim shadowing at and when to use something else.
Where the evidence is solid
The review found shadowing training can improve:
- Comprehensibility: how much effort a listener needs to understand you.
- Intelligibility: whether they actually catch your words.
- Accentedness: how strong your accent sounds to listeners.
- Suprasegmental control: rhythm, stress, intonation and fluency.
Put simply, shadowing trains the sound of the whole sentence: where the beat falls, which words shrink, where the voice rises and falls.
Where the evidence is weak
The same review says findings on segmental pronunciation (individual sounds such as th, the r/l difference, or vowel length) were inconclusive. Whether those improved seemed to depend on explicit phonetic feedback and on how the task was designed.
So if your goal is a clean th, shadowing alone may not get you there. You'll probably need someone or something to explain where your tongue goes and what that one sound is doing wrong. For that sound specifically, see how to pronounce th.
What learners often do, and what shadowing fixes
Most learners who say "my pronunciation is bad" are worried about individual sounds. When you listen to them, the bigger problem is often that every word gets the same weight.
- I WANT TO GO TO THE STORE AFTER WORK, every word equally loud → I wanna go to the STORE after WORK. English puts stress on the words that carry meaning and squeezes the rest. Equal stress makes listeners work harder to find what matters.
- Can you SEND me the FILE BY FRIDAY? with a flat voice at the end → Can you send me the file by FRIday? with the voice going up on the last stressed word. A flat ending can make a request sound like a demand.
- What-do-you-think? as four separate words → Whaddaya THINK? Small words link and reduce. Saying each one fully is correct, but it sounds like reading aloud.
These are rhythm and intonation problems, which is the area where the review found support for shadowing.
What learners said about it
The review also reports how learners felt: they generally found shadowing interesting, enjoyable and effective. That matters more than it sounds. Most people don't fail at speaking practice because they picked a slightly worse method. They stop. A routine you're happy to repeat every day is worth a lot.
A routine that matches the evidence
- Listen for the beat first. On the first listen, ask which words are stressed before you worry about how any single word is pronounced.
- Shadow out loud. Pick one short sentence, 5 to 12 words, and speak along with the recording at the same speed.
- Compare. Repeating along and knowing exactly where you fell behind are different activities. Record yourself or use a tool that shows the difference.
- Repeat the same sentence three times before moving on. Short and frequent beats one long session a week.
- Treat problem sounds separately. Look up the mouth position, practice that sound in a few words, then put it back into a sentence.
For the full daily routine, see how to practice English speaking alone. For what happens to words in fast speech, see connected speech.
Where MirrorSay fits
MirrorSay covers step 3. You shadow the reference audio for a sentence and get a word-by-word rhythm comparison, so you can see which word you dragged and which you swallowed. Tap Mine to hear your own take. It compares rhythm, the layer where the evidence is reasonably solid. It does not score individual sounds.
Reference
- Whitworth, B. N., & Rose, H. (2025). A Systematic Review of Research on the use of Shadowing for Second Language Pronunciation Teaching. doi:10.1080/29984475.2025.2546827
MirrorSay is a free English speaking app: say a sentence, see how a native speaker would say it, then shadow it with word-by-word rhythm feedback. No sign-up; speech is processed on-device whenever possible.
