Most intermediate learners I talk to have the same complaint. They understand English movies, can read articles without a dictionary, and write reasonably well, but the moment a real conversation starts, the words won’t come. Then they ask what to study to fix it, and the honest answer is “less.” You don’t have a knowledge problem. You have a retrieval problem. The only thing that fixes a retrieval problem is using language under time pressure, which means speaking, with all the discomfort that involves.
Below are four protocols I’d actually recommend, in rough order from least to most socially demanding. You don’t need to do all four. Two, done consistently for a few months, will outperform reading thirty more articles about how to learn languages.
Shadowing
Shadowing was popularized by polyglot Alexander Arguelles, who used it to internalize the rhythm and intonation of dozens of languages. The technique sounds trivial and is harder than it looks.
Take a 30-to-60-second clip of a native speaker. A podcast interview works well. Movies are too fast at first. Listen once for content. Then play it again and speak along with the speaker, half a beat behind, copying not just the words but the stress, the pauses, and the melody of the sentences. Don’t stop to look anything up. If you fall behind, jump back in.
The first few sessions feel impossible. Your mouth lags, you mumble through phrases you don’t catch, and you can’t tell whether you’re getting better. You almost certainly are. After two weeks of daily ten-minute sessions, learners typically report that their internal sense of English rhythm has shifted. They start to feel where the stress wants to go in a sentence rather than placing it deliberately.
What shadowing fixes: stilted intonation, robotic stress patterns, and the lag between thinking a sentence and producing it. What it doesn’t fix: vocabulary gaps or grammar errors. It’s a fluency tool, not a knowledge tool.
Self-narration
This is the protocol with the lowest social cost and the highest weirdness factor. You narrate your own actions and thoughts to yourself, in English, throughout the day. Out loud when you’re alone. In your head when you’re not.
Okay, I need to make coffee. Where did I put the filters? They were on the counter yesterday. Maybe in the cabinet. Yes, here. Now the water…
It feels stupid. It works because it forces you to find words for things you actually do, in real time, with no one to bail you out. You’ll discover that you don’t know the word for “filter” or “to rinse” or “to come to a boil.” Unlike a textbook list, these gaps are personal. They’re the words you need.
Keep a running list (in your phone) of words you reached for and couldn’t find. Look them up later, not during. Interrupting the narration to look things up defeats the point.
Recorded retell
Pick a podcast episode, an article, or a YouTube video you found interesting. Consume it once. Then, without looking at it, record yourself summarizing it in English for two or three minutes, as if explaining it to a friend.
Listen back to your recording. This part is genuinely unpleasant the first few times. You’ll hear all your filler words (“uh,” “like,” “you know”), the moments where you abandoned a sentence midway, the pronunciations that came out wrong. You’ll also hear what’s working: moments where the language came easily, sentences that flowed.
Two or three retells per week, kept in a folder, become the most honest progress tracker you have. The recording from three months ago will surprise you.
A more demanding variation: retell the same source after a 24-hour gap, then a week later. The first retell tests comprehension and short-term retrieval. The later ones test what’s actually consolidated.
This builds directly on closing the comprehension-production gap. The recording forces you to convert what you understood into what you can produce.
Structured conversation with a tutor
iTalki, Preply, and similar platforms let you book one-on-one sessions with native or near-native speakers, often for less than the price of a movie ticket. The platforms aren’t the point. The structure is.
Unstructured conversation, the kind where you say “let’s just chat for an hour,” is where most learners’ tutoring sessions go to die. You’ll talk about your weekend, the weather, what you do for work, and you’ll do it with the same vocabulary you’ve used a hundred times. You’ll feel productive. You won’t be making progress.
A protocol that works better:
- Before the session, pick a topic with some substance. What do you think of remote work? Why is your country’s birth rate falling? Anything that requires you to argue, not just describe.
- Spend ten minutes preparing, not by writing a script, but by listing five vocabulary items you’d want to use and two grammatical structures you’ve been working on. The point is to bring tools to the session.
- Tell your tutor at the start: “I want to discuss X. Please correct me on grammar and on word choice, but only after I finish a thought. Don’t interrupt mid-sentence.”
- After the session, write down three things the tutor corrected. Those become your study list for the next few days.
The tutor isn’t there to be your friend. They’re there to give you the resistance that ordinary conversation doesn’t.
A couple of things people get wrong
One: you won’t feel ready. Or rather, the feeling of readiness only arrives after a few months of speaking while not feeling ready. The willingness to sound bad is the practice. There isn’t a level of preparation that turns speaking into something comfortable in advance.
Two: treating speaking as the final exam after a long study phase is backwards. Speaking is what reveals which of your “learned” words are actually usable, and which sit dormant in passive memory. It should run alongside everything else, from the first week of learning a language to the day you stop.
If you can only commit to one of the four protocols above, do self-narration. It’s free, requires no scheduling, and will quietly produce more progress in three months than most learners get from formal courses in a year.