Most intermediate learners describe the same wall. They can follow a podcast at near-normal speed, read a novel without too much friction, and understand almost everything spoken to them in person. Then they open their mouths and produce a halting, simplified version of themselves. The gap between input and output yawns wider than they expected, and the conventional advice — "you just need to speak more" — turns out to be only partly true.
It is partly true because nobody ever gets fluent without speaking; the production system needs the work. It is mostly wrong because the bottleneck for most people in this state is not lack of practice but the speed at which words can be retrieved. You know the word. You have heard it a thousand times. You can recognise it in a fraction of a second when someone else says it. But when you go to summon it, there is a small, embarrassing pause, and by the time the word arrives the conversation has moved on. Multiply this lag across an evening and what comes out of your mouth is not what you would say if you were thinking in your own language.
The standard prescription — more conversation — does work eventually, but slowly, and at a cost most people underestimate. Talking with a tutor twice a week gives you maybe ninety minutes of speaking time and a great deal of polite waiting while you reach for words. Most of those minutes are spent in conversations whose vocabulary you already had locked down weeks ago. The ones that would stretch you require words your retrieval system is still rehearsing, and rather than do the rehearsing you tend to substitute simpler words you can reach faster. The result is the experience many learners report: months of speaking practice without much sense of progress.
What helps more, for less time, is anything that forces retrieval under low stakes. Shadowing — listening to a recording and speaking along about half a second behind — trains the same circuits that fire in conversation but without anyone to perform for. You can do it on a walk and look reasonably normal. Self-narration is the same idea, less structured: out loud, in your target language, describe what you are doing as you do it. Most people skip this because it feels childish; ten minutes of it a day, sustained for a month, will close more of the retrieval gap than any amount of polite tutoring.
The other thing that helps, almost as much, is narrowing the input. Instead of listening widely — a different podcast every day, varied accents, varied topics — pick one source you genuinely enjoy and listen to the same episodes more than once. The first listen catches the meaning. The second catches the phrasing. The third lets you steal whole turns of phrase that will surface, weeks later, when you need them. Variety feels productive; repetition is what actually moves words from the recognition system into the production system.
None of this contradicts the case for real conversation. It just clarifies what conversation is for. Real conversation is where you find out what you cannot yet do — what gaps still exist, what topics make you freeze, what register you can’t manage. It is poor practice in itself but priceless as a diagnostic. The work of closing the gap happens mostly in private, in the cheap, unwitnessed reps that no one would think to recommend because they sound too humble.
It is worth saying that the gap is also slightly self-correcting if you allow it. People who have lived in their target country for a year speak with a fluidity they could not have produced after a year of classes, even though the classes contained more deliberate practice. The reason isn’t immersion in the romantic sense. It’s that ordinary life forces thousands of low-stakes retrieval events — at the bakery, on the bus, paying a bill — and most of them work because the script is constrained. You retrieve "a small loaf" enough times and the next time it comes up the lag is gone.
If you cannot move countries, you can manufacture a smaller version of the same conditions. Read a recipe out loud while you cook. Watch a film in your target language and pause every ten minutes to summarise aloud what just happened. Keep a voice diary on your phone — not for anyone to hear, just for the practice of speaking in continuous sentences without preparation. These are the things that work, and the reason most learners don’t do them is that they feel beneath the seriousness of the project. They are the project.
The understand-but-can’t-speak problem is real and it is durable. It also resolves, in nearly every case, given enough cheap reps in the right direction. The mistake is to assume the resolution lies in difficult, expensive practice with another person. Most of it lies in easy, lonely practice with yourself.
