What Is the Shadowing Technique?
The shadowing technique is a language-learning method where you listen to a native speaker and repeat what they say out loud while they're still talking, a fraction of a second behind, like an echo. It's also called language shadowing or the shadowing method. You don't wait for a pause, and you don't translate. You track the sound as it arrives and reproduce it, rhythm and intonation included.
Does shadowing work? For listening, the evidence is reasonably good. Studies in Japan, where shadowing has been researched since the late 1980s, repeatedly find that it sharpens how well you pick out words in connected speech. For speaking, the gains show up in fluency and in how easy you are to understand. On accent, the evidence is mixed.
Researchers use a stricter definition than most blogs. Tamai (2005) describes shadowing as "an act or task of listening, in which the learner tracks the heard speech and repeats it as exactly as possible while listening attentively to the in-coming information." Two parts of that matter. It's a listening task first, and the tracking happens live.
That's what separates shadowing from listen-and-repeat. When you repeat, the speaker stops and you get a silent gap to rebuild the sentence from memory, which is why Shiki and colleagues (2010) call repeating an off-line task and shadowing an on-line, immediate one. In shadowing the next words are already arriving, so you have to hear the sounds as they come.
Who Invented Shadowing? Labs, Interpreter Booths, and Polyglots
No single person invented it. Shadowing was a psychology tool long before it was a study method. In 1953, the British cognitive scientist Colin Cherry played different messages into a listener's two ears and asked them to repeat one aloud as they heard it. Repeating locked their attention onto one ear, and they could report almost nothing about the content of the other. That same paper gave us the "cocktail party problem": how we follow one voice in a noisy room.
In Leningrad in the late 1950s, Ludmilla Chistovich and Valerij Kozhevnikov used shadowing to study how quickly the brain turns heard speech into speech of its own. The fastest "close shadowers" follow at a lag of roughly a quarter of a second, and in 1973 William Marslen-Wilson reported in Nature that even at that speed they were still analyzing the grammar and meaning of what they repeated.
Interpreters adopted it next. The interpreting researcher Sylvie Lambert (1992) described shadowing as a paced, parrot-style auditory tracking task done with headphones, and it became a familiar first exercise for trainee simultaneous interpreters, who must listen and speak at once. Trainers still debate how much it helps them.
Japan is where it became a classroom method. Research on shadowing for English learners there began in the late 1980s, and Japan is still the biggest single source of studies, including 10 of the 44 in a 2024 review of shadowing for pronunciation. The psycholinguist Shuhei Kadota of Kwansei Gakuin University wrote up the theory in several books, including the 2019 English-language Shadowing as a Practice in Second Language Acquisition. In a 2019 review, Yo Hamada of Akita University noted that 25 years had passed since the technique reached Japanese classrooms and that it was gradually being recognized around the world.
The polyglot Alexander Argüelles brought it to independent learners through demonstration videos with a distinctive routine: walk briskly outdoors, keep an upright posture, hold the text up in front of you, and speak loudly and clearly. Many online guides call him the inventor. He wasn't, but his videos did a lot to popularize the technique outside classrooms.
Why Shadowing Works: Four Effects of One Echo
Kadota's 2019 book sorts shadowing's benefits into four effects, which make a good map of what you're training.
| Effect | What it trains | What you'd notice |
|---|---|---|
| Input effect | Listening comprehension, through faster and more accurate sound perception | Connected speech stops sounding like one long word |
| Practice effect | Subvocal rehearsal in phonological working memory, used to learn new words, phrases, and constructions | Phrases stick after a few sessions |
| Output effect | The later stages of speaking: finding the words and their sounds and articulating them, not deciding what to say | Your delivery gets smoother |
| Monitoring effect | Metacognitive monitoring and control of your own performance | You catch your own errors as you make them |
The input effect is bottom-up listening. Listening runs in two directions. Top-down processing uses context and background knowledge to guess what was said. Bottom-up processing builds meaning from the sounds themselves: phonemes into words, words into phrases. Learners usually bring plenty of top-down skill from their first language. What they lack is bottom-up speed, the ability to pick words out of connected speech before the next sentence arrives.
Shadowing forces your attention onto the sound rather than the content (Kadota, 2019; Hamada's 2016 book). Kadota describes the improvement as a sequence: first you repeat more accurately, then your articulation speeds up, and only then does listening comprehension improve. George Miller described a similar climb in a radio operator learning Morse code, who first hears every dit and dah separately and eventually hears whole phrases, an idea we unpack in our guide to the chunking technique. The same order explains why your first week of shadowing can feel like it's only training your mouth.
The practice effect is a two-second buffer. Working memory holds speech in a short-lived phonological store. Baddeley, Thomson, and Buchanan (1975) found that people remember about as many words as they can say in roughly two seconds, so the faster you can articulate a phrase, the more of it fits. Shadowing pushes articulation speed up, and Miyake (2009) found that multiword expressions learners could say within two seconds after repeated shadowing were recalled significantly better than ones that took longer.
The output effect is half of speaking. Speaking starts with deciding what to say, then finding the words and their sounds, then articulating them. Kadota argues that shadowing doesn't touch that first stage but may simulate the later two. Expect smoother delivery before you expect new things to say.
The monitoring effect is self-correction in real time. Shadowing trains you to monitor your own output while you listen to someone else's, and recording a pass makes that monitoring explicit.
One caveat matters for everything that follows. Shadowing can run without meaning. Van Paridon and colleagues (2019) describe two routes through the task: one where you process meaning, and one where you reproduce the sounds without ever retrieving what the words mean. That's why the good protocols build in a meaning check.
What the Research on Shadowing Shows
Much of the evidence comes from Japanese students shadowing English, and most of the studies are small and short. The before-and-after results line up in a fairly consistent pattern, even though head-to-head comparisons with other listening activities are mixed.
| Study | Who | Training | Result |
|---|---|---|---|
| Tamai (1992) | 94 Japanese high school students | Shadowing vs dictation, 13 lessons over about three and a half months | Only the shadowing group's listening scores improved |
| Nakanishi and Ueda (2011) | 89 Japanese university students | Extensive reading with or without shadowing, vs two translation-based classes, one year | All groups improved, but none outperformed the others in reading comprehension |
| Hamada (2016) | 43 Japanese university EFL learners | Nine shadowing lessons | Phoneme perception improved in both proficiency groups; basic listening scores rose only in the lower group; advanced listening didn't move |
| Foote and McDonough (2017) | 16 ESL learners | Shadowing dialogues on iPods for eight weeks, 4+ times a week, 10+ minutes a time | Comprehensibility and fluency improved, as rated by 22 English speakers; accentedness didn't change significantly |
| Whitworth (2024) | Systematic review of 44 studies (Oxford master's dissertation) | Shadowing for pronunciation | Support for comprehensibility, intelligibility, accentedness, fluency, and prosody; evidence on individual sounds inconclusive |
| Hashizaki et al. (2025) | 30 Japanese university students | 11 weekly classes, plus at least 10 repetitions of each passage as homework | Bottom-up processing and repetition speed improved; listening and phrase-memory results tracked how many repetitions students logged |
The results fall into three patterns.
Listening gains come from the bottom up. Hamada's study is the cleanest example. Phoneme perception, measured with a dictation-cloze test, improved in both groups, while the comprehension gains were narrower. Shadowing improves how well you hear words before it improves how well you follow ideas.
Speaking gains show up in fluency and intelligibility before accent. Foote and McDonough's learners became easier to understand and more fluent, but their accent ratings didn't improve significantly, and the study had no control group. Whitworth's review is more positive about accent, but it flagged a methodological problem: many studies measured controlled tasks like reading aloud rather than spontaneous speech.
Shadowing hasn't been shown to help reading. In a year-long study, Nakanishi and Ueda found no reading-comprehension difference between extensive reading with shadowing, extensive reading alone, and two translation-based classes. All four groups improved by about the same amount.
Does Shadowing Work Best for Beginners?
One of the most repeated claims about shadowing is that it helps lower-proficiency learners the most. It comes from real studies. When Tamai reanalyzed his 1992 high-school data by proficiency level (reported in Tamai, 2005), shadowing worked better for the low and middle groups than for the top group. Hamada (2016) found the same shape: only the lower group's basic listening scores rose, and by the end of training they had approached where the intermediate group started.
A 2025 study in JALT Journal by Ryotaro Hashizaki and colleagues raises a statistical problem with that kind of result. Both analyses grouped learners by their pre-test scores. People who score unusually low on a first test tend to score closer to average on the second, whatever happens in between, a pattern called regression toward the mean. Sort a class by pre-test score and the bottom half can look like it improved more even if the training did nothing special for them. Hashizaki's team plotted their own data and saw that shape: of the learners below the median at pre-test, only three scored lower afterward, while 12 of those above the median did.
That doesn't make shadowing useless for beginners or for advanced learners. It means the "works best for beginners" finding was never as solid as it sounded. Hamada's own 2019 recommendation is more useful than the rule: beginners should start with shadowing for listening and then move on to shadowing for speaking. If you're past the beginner stage, aim shadowing at the speaking side, and mix harder clips in with easier ones.
The material matters too. The standard advice, from Kadota and Tamai (2005), is material slightly below your current level, or easier, and Kadota (2007) suggests no more than two or three unknown words per 100. That's deliberately easier than the "i+1" input Stephen Krashen recommended for acquiring new language, which we cover in how to learn a language with AI. Easy material isn't the whole story, though: in a 2012 classroom study, Hamada found that learners who alternated between two difficulty levels improved their listening more than those who stuck to one.
Types of Shadowing, Compared
"Shadowing" covers a family of techniques, and there's no single agreed set. As Hamada (2012) put it, how shadowing is used varies from researcher to researcher. These are the variants worth knowing:
| Type | How it works | Best for |
|---|---|---|
| Mumbling | Shadow quietly, focusing on the incoming sound rather than your own pronunciation | First contact with a new clip |
| Parallel (synchronized) reading | Read the transcript aloud in sync with the audio | Learning rhythm and linking before going script-free |
| Prosody shadowing | Shadow without the script, copying stress, rhythm, and intonation | Pronunciation and natural rhythm |
| Content shadowing | Shadow while consciously following the meaning | Listening comprehension; the final stage |
| Silent shadowing | Shadow in your head, without voicing | Public places and low-energy sessions |
| Part shadowing | Repeat only the stressed words or the last word of each phrase | Beginners and very fast speech |
| Conversational shadowing | Echo parts of a conversation partner's speech, sometimes adding a question or comment | Practice with real people |
Kadota and Tamai (2005) described the first four, which form the standard sequence. Kurata (2007) described silent and part shadowing, and Tim Murphey (2001) studied conversational shadowing, in which learners and native speakers echo parts of each other's speech as they talk.
The sequence hands you crutches and then takes them away. Mumbling removes the pressure to sound good, parallel reading lends you the script, prosody shadowing takes it back, and content shadowing removes the option of ignoring meaning. The routine below climbs through all four.
How to Shadow: A Step-by-Step Routine
This is a self-study version of the lesson procedure Hamada (2012) used in his classes, which followed Kadota and Tamai's (2005) sequence, with one change taken from Hamada's later advice: understand the content before you start shadowing it. Budget about 15 to 20 minutes per clip.
- Pick a 30- to 60-second clip: one you can mostly follow, with only a few unknown words. Alternate it with a harder one from session to session, as Hamada's 2012 learners did.
- Listen once and jot down what you catch: Hamada opened and closed every lesson with the same dictation-cloze exercise so learners could see their own progress. A rough note of what you heard is a self-study stand-in.
- Check the meaning (about three minutes): read the transcript, look up unfamiliar words, and make sure you understand every sentence. Hamada's learners checked both an English script and a Japanese translation.
- Mumble along twice: quietly, without the transcript. Focus on the sound, not on your pronunciation.
- Parallel read two or three times: read the transcript aloud in sync with the audio.
- Shadow without the transcript three times: expect to drop words, and keep going rather than stopping.
- Check the details (about three minutes): go back to the transcript for the sounds you couldn't hear or reproduce and the meanings you missed.
- Content shadowing: shadow once more, following the meaning as you speak.
- Record one pass and compare it: listen back against the original, then repeat your note from step 2.
How many passes? Accuracy, meaning how many words you reproduce correctly, tends to plateau after four or five passes on the same material (Shiki et al., 2010; O'ki, 2014). Hamada advises five or six shadowing passes per text in total (steps 6, 8, and 9 above make five), which don't have to happen in one sitting, before moving on to a new one. Hashizaki's team goes further: in their 2025 study, students who logged more repetitions did better on the end-of-course listening and phrase-memory tests, and they recommend going beyond five. That's a correlation from a study without a control group, so treat it as a reasonable bet rather than a rule. A middle path is to drill a clip until your accuracy levels off, then come back to it on a later day.
How long per session? The recent studies used short blocks. In Hamada's 2016 study, two 10- to 15-minute sessions a week for a month produced a noticeable change. Hashizaki's students shadowed for 10 to 15 minutes in class, and Foote and McDonough's learners practiced at least four times a week for at least 10 minutes. Hamada calls shadowing demanding and recommends using it intensively for a short stretch, such as a month, and then moving on, because learners find it hard to stay motivated through such repetitive work.
Shadowing with YouTube Videos
Shadowing needs native-speed audio, an accurate transcript, and an easy way to replay the same 20 seconds over and over. YouTube has all three, in almost every language and on whatever topic you like. Choose the topic with care: you'll hear each clip 10 or more times, so pick something you want to hear.
Choose videos with human-made captions. Open the captions menu: machine-generated tracks are labeled "(auto-generated)." Automatic captions often mangle names and technical terms, and a wrong transcript teaches you wrong sounds. TED talks and learner programs such as BBC Learning English's 6 Minute English publish accurate transcripts, and many interviews and podcasts do too.
Open the transcript. On desktop, expand the description and click "Show transcript" to see the full text with timestamps. If you'd rather have the script as text you can print and mark up, paste the link into Glasp's free transcript tool. Our guide to getting a YouTube transcript covers desktop, iPhone, and Android.
Get the gist first. Understanding the content before you shadow is faster when you already know what the video is about. YouTube Summary gives you a summary and the transcript with timestamps before you start, so context does the top-down work and your attention is free for the sounds.
Slow down, then speed up. If a clip feels too fast at first, use 0.75x playback for mumbling and parallel reading, then return to normal speed for the script-free passes. Keyboard shortcuts make the replay loop painless: K pauses and plays, J jumps back 10 seconds, and the left arrow key jumps back 5.
Bookmark your chunk. The annoying part of shadowing from YouTube is finding your place again tomorrow. With Glasp's YouTube highlighter, the transcript appears beside the video, and highlighting a line stores it with its timestamp, so one click on the highlight takes you back to that exact second. Highlight the first line of each chunk you're working on, and the highlights become your shadowing queue. Highlight the expressions you want to keep, too, and the same list doubles as a phrase bank.
Record and compare. Use your phone's voice recorder for one pass and listen back against the original. Foote and McDonough's learners recorded themselves while shadowing, and it's the cheapest feedback loop there is.
For the wider system of learning from video, from choosing videos to taking notes and building a knowledge base, see how to learn from YouTube.
Common Shadowing Mistakes (and How to Fix Them)
Choosing material that's too hard. If you can't follow a clip even with the transcript open, you're practicing failure. Fix: drop to material with only a few unknown words, and save hard clips for an occasional stretch.
Skipping the meaning. Remember the two routes: you can shadow fluently and understand nothing. Hamada (2012), noting Shiki and colleagues' (2010) finding that accuracy stops rising after four or five passes, built meaning checks into every lesson. Fix: understand the clip before you shadow it, check the details afterward, and finish with a content-shadowing pass.
Stopping after one or two passes. Accuracy takes four or five passes to level off. Fix: decide on a number of passes before you start, not a number of minutes.
Translating in your head. Hamada has warned that some learners try to translate everything they hear on the spot, and the failures erode their confidence. Fix: do your translating during the meaning check, then let it go and stay in the target language.
Freezing when you lose the thread. Dropping words is normal. Keep going and rejoin at the next phrase, or you've turned shadowing back into listen-and-repeat.
Running marathon sessions. An hour of shadowing turns into an hour of mumbling on autopilot. Keep the shadowing itself to 10 to 15 minutes a session, as Hamada's and Hashizaki's classes did.
Expecting an accent transplant. Accent is where the evidence is weakest. Judge your progress by how easily people understand you, which is what conversations run on.
Frequently Asked Questions
What is the shadowing technique in language learning?
Shadowing is a technique where you listen to native speech and repeat it out loud while it's still playing, a fraction of a second behind the speaker, like an echo. Researchers define it as a listening task: you track the speech and reproduce it as exactly as you can. Along the way it also builds fluency, rhythm, and memory for phrases.
Does shadowing actually work?
Mostly, yes. For listening, studies of Japanese learners of English find better phoneme perception and bottom-up listening after a few weeks, though shadowing hasn't always beaten other listening activities head to head. For speaking, studies find gains in comprehensibility and fluency; the evidence on accent is mixed, and on individual sounds it's inconclusive. Most studies are small and short.
How long should I shadow each day?
Ten to fifteen minutes of shadowing itself per session is what recent classroom studies used. In Hamada's 2016 study, two 10- to 15-minute sessions a week for a month made a noticeable difference, and Foote and McDonough's learners did at least 10 minutes at a time, at least four times a week, for eight weeks. With meaning checks and replays, a full session takes about 15 to 20 minutes.
Is shadowing good for beginners?
Yes, with easy enough material. Early studies found the biggest listening gains in lower-proficiency learners, though a 2025 study argues those results may be inflated by regression toward the mean. Hamada recommends that beginners start with shadowing for listening and move on to shadowing for speaking later.
Should I shadow with or without a transcript?
Both, in that order. Use the transcript to understand the content and for parallel reading, then put it away for prosody and content shadowing. Shadowing only with a script turns into reading aloud. Shadowing only without one can leave you reproducing sounds you never understood.
Can shadowing improve my pronunciation and accent?
It can make your pronunciation easier to understand, but don't expect your accent to change quickly. Foote and McDonough found that after eight weeks of shadowing, learners were rated more comprehensible and fluent but not significantly less accented, while a 2024 review of 44 studies, written as a master's dissertation, was more positive about accentedness. Stubborn individual sounds may need targeted pronunciation practice.
What's the difference between shadowing and listen-and-repeat?
Timing. In listen-and-repeat, the speaker pauses and you reproduce the phrase from memory. In shadowing, you speak while the audio keeps running, a split second behind. Researchers call repeating an off-line task and shadowing an on-line one, and that live pressure is what forces you to process the sounds instead of reconstructing them.
What are the best materials for shadowing?
Short clips of natural speech, 30 to 60 seconds long, with an accurate, human-made transcript and only a few words you don't know. Interviews, podcasts that publish transcripts, TED talks, and learner programs such as BBC Learning English all work well. Pick topics you care about, because you'll hear each clip many times.
Conclusion
Shadowing is better supported than most language-learning hacks, and more modest than its fans claim. The best-supported finding is that it trains you to hear words in connected speech, and in the few speaking studies it made learners more fluent and easier to understand.
Don't count on it for your accent or your reading, and don't skip the meaning checks. Keep sessions short and the material just below your level, with a harder clip now and then.
Start today with one clip. Find 60 seconds of a video you'd want to watch anyway, read the transcript until you understand it, and run the routine: mumble, read along, shadow, check, record. Save the line you started on with Glasp's YouTube highlighter so tomorrow's session starts in one click. When a show publishes its transcript as a web page, as TED and BBC Learning English do, mark the lines you're drilling with Glasp's web highlighter. Your ear will notice the difference before your mouth does.
