The gap between comprehension and speech is not an accident and it is not a personal failing, because input built it, week by week. Input is available at any hour, costs nothing, requires nobody else and produces a feeling of progress. Output requires another person, has to be arranged, and produces a feeling of failure. Given those two options every week for a year, anybody rational chooses input, and the gap widens by design.
Which is why telling somebody to speak more does not work. Nothing is wrong with their judgement; each individual choice was the better one, and the sum of fifty of them is a person who follows a documentary and cannot order lunch.
Eight things about the wall, and what gets over it
- B1 to B2 is the longest stretch in most people's experience, and the one where the majority of learners stop.
- the early gains were the frequent words. The first thousand words appear constantly, so learning them pays immediately.
- the next words are ten times rarer, so each one is met far less often and takes far longer to fix. The slowdown is arithmetic rather than psychological.
- switch from courses to real material. Courses are built for the first two levels and there is very little above B1 worth following.
- narrow your domain. Go deep in one subject for two months, where the vocabulary repeats, rather than broad across many.
- start producing daily, in writing if speaking is not available, because this is the skill that has stopped moving.
- find a corrector, a tutor or an exchange partner who will actually correct rather than encourage.
- measure differently at this stage. Words learned stops being a useful measure; tasks completed in the language starts being one.
Two skills on one scale
Understanding and producing are not one ability measured at two strengths. Comprehension and production run on different routes, and building one does very little for the other, so a large gap between them is not a paradox needing an explanation.
It does mean a single number for your level is close to meaningless once the gap is wide. Somebody reporting a level is almost always reporting the higher of the two, and the one other people will experience is the lower.
It also means the gap does not close on its own, ever, and it does not close from the input side however much of it there is. Input is what made it.
Asynchronous output is the cheap first step
The reason output gets postponed is that it is expensive in a specific way: it needs another person, at the same time as you, prepared to wait. Every one of those conditions can be relaxed except the last.
A voice message needs another person and not the same moment. You can record it four times, you can leave a long pause, and nobody is watching you assemble the sentence. That removes the real-time pressure and keeps the production, which is exactly the right first trade for somebody whose problem is that they never produce anything.
It does not build speed, and speed is the second problem. But a person who has never produced anything cannot practise speed, because there is nothing to speed up.
Where it stops: choosing the comfortable hour, correctly
The stall is a series of reasonable weekly decisions. Tonight there is an hour. A podcast is available, enjoyable, at the right level, and will teach you something. Arranging a conversation would take three messages and a scheduling negotiation, and the conversation itself would be uncomfortable.
Nobody chooses the second option on a Tuesday. Over a year, that is fifty hours of input and none of output, and the person is genuinely surprised by the gap, because they were working the whole time.
What breaks it is not resolve. It is making the output hour cheaper to start than the input hour, which means it has to be already arranged: a standing appointment, a person who expects a message on Sundays, a booked slot that would have to be cancelled rather than arranged.
What it looks like when it goes wrong
The recognisable failure is somebody who watches an hour of television in the language every evening and cannot introduce themselves without rehearsing. Both halves of that are real and neither is going to fix the other.
The second is measuring by comprehension. Progress is easy to see on the input side, so that is where people look, and the number that moves is the one that was never the problem.
Why every method stops working at some point and how to read a stall as a fact about the method rather than about you is the stalls every learner hits, and the way through each, and the arrangement that supplies the missing half, along with what it costs, is finding someone to speak to, and what to do in the hour.
The open question is what happens to the input side once output starts, because the two are not independent: an hour of listening after a conversation you struggled through is aimed at something, and the same hour on its own was not: material just past your level, and why that is the sweet spot.