The intermediate plateau is the stage, usually somewhere around B1, where measurable progress slows sharply even though nothing about your study habits has changed. It is not a motivation failure and it is not a ceiling on your aptitude. It is arithmetic. The first thousand words you learned covered most of the words on every page you opened, and every thousand you learn from here covers progressively less. The plateau is what that curve feels like from the inside.
Why the beginning felt so fast#
Word frequency in natural language is extremely lopsided. A small set of function words and very common verbs and nouns accounts for a huge share of everything anyone writes or says, so the first few hundred words you learn appear constantly, get reinforced automatically, and immediately change how much of a text you understand.
Paul Nation's work on vocabulary size and text coverage gives the standard picture. The figures below are approximate, vary by corpus and by how word families are counted, and should be read as a shape rather than a set of targets.
| Word families known | Approximate coverage of general written text | Unknown words per 100 |
|---|---|---|
| 1,000 | ~80% | ~20 |
| 2,000 | ~87% | ~13 |
| 3,000 | ~90% | ~10 |
| 5,000 | ~95% | ~5 |
| 8,000–9,000 | ~98% | ~2 |
Read the increments rather than the totals. Going from 1,000 to 2,000 words buys around seven percentage points of coverage. Going from 5,000 to 9,000 buys about three, and costs four times as many words. Yet the gap between 95% and 98% is the difference between reading that takes effort and reading that does not, which is why the last stretch feels both slowest and most valuable. The same math for one language appears in how many words it takes to be fluent in Spanish.
The second mechanism: rarity kills repetition#
Coverage explains why progress feels smaller. Frequency explains why it also gets harder to make.
Below roughly 3,000 words, almost everything you need to learn appears often enough that ordinary reading supplies the repetitions on its own. Meet a word four or five times across a month and it sticks without deliberate effort. Above that threshold, the words you still need appear once every few tens of thousands of running words. You can read consistently for weeks and meet a given mid-frequency word exactly once, which is not enough for it to survive.
This is where a method that worked stops working. Passive reading was sufficient at A2 and is insufficient at B1, not because reading got worse, but because the statistical support it quietly provided has run out.
Diagnosing your plateau#
Plateaus have distinguishable causes, and the fix depends on which one you have. Three checks separate them.
The coverage check: take a hundred consecutive words of native text at a fixed difficulty and count how many you do not know. Repeat monthly with comparable material. If the number is not falling, vocabulary is your bottleneck.
The range check: read something from a domain you never read in. If a familiar-topic text feels fine and an unfamiliar one collapses, your problem is breadth rather than depth.
The output check: record yourself speaking for two minutes on any topic, then transcribe it. Most plateaued learners find they cycle through a few hundred words while understanding many thousands more.
| Symptom | Likely cause | The relevant fix |
|---|---|---|
| Comfortable on familiar topics, lost outside them | Narrow input range | Widen domains deliberately |
| Understand almost everything, produce the same small set of words | No output pressure | Structured speaking and writing |
| Reading feels identical to six months ago | Rare words not repeating enough to stick | Targeted mid-frequency vocabulary |
| Comprehension fine, but everything is exhausting | Sitting at 90–93% coverage rather than 98% | More volume at a controlled difficulty |
Way out 1: widen the range, not the volume#
The instinct at a plateau is to read more of what you already read, which mostly re-teaches words you know. What breaks the pattern is variety of domain and register, because different subject areas draw on largely non-overlapping mid-frequency vocabulary.
Pick three domains you actively avoid and read one text from each per week. Cooking, sports coverage, law, local politics, product reviews, technical documentation. The discomfort is the signal that new vocabulary is present. This is also why an interest-driven approach beats a curriculum here: you are not looking for difficulty but for unfamiliarity, and the two are not the same thing. The trade-off between reading widely for volume and reading closely for detail is covered in intensive versus extensive reading.
Difficulty still has to stay in the productive band, and an unfamiliar domain at native density drops you well below the comprehension threshold. Reading at a controlled target-language share solves this: LingoBlend lets you paste an article from a domain you know nothing about and set what proportion of its words appear in your target language, so the topic can be new while the difficulty stays honest. That is the diglot weave technique Robbins Burling described in 1968; the features page covers the reader and word-saving side.
Way out 2: deliberate output#
Reading and listening build recognition. Recognition does not convert into production on its own, and at the plateau the gap between the two is usually enormous. Closing it requires producing language under some constraint that forces you past your comfortable vocabulary.
Three exercises do this cheaply. Summarize something you read out loud, using the text's words rather than your usual paraphrase. Write a short argument for a position you disagree with, which forces hedging and concessive structures. Keep a running list of things you wanted to say and could not, then look up how a native writer would have said them.
The point is not fluency practice for its own sake. Attempting production reveals precisely which words you have only ever recognized, turning a vague sense of being stuck into a specific list.
Way out 3: targeted vocabulary rather than more of the same#
Since rare words no longer repeat enough for reading to teach them, the repetitions have to come from somewhere else. That is what spaced repetition is for, and it is more important at B1 and above than it ever was at A1.
Two selection rules make this efficient. Mine from your own reading rather than generic lists, so every word you review is one you already met in a remembered context. And prioritize mid-frequency vocabulary over rare specialist terms: a word in the third-to-eighth thousand band pays off across everything you read, while a technical term repays only inside its field.
A standard SM-2 ladder starts a word at short intervals and stretches them with each success, the practical answer to the forgetting curve Hermann Ebbinghaus documented in 1885. Saving words from the text you are reading and letting the scheduler surface them is the whole loop; spaced repetition for language learning covers the mechanics, and the B1 to B2 reading plan structures this stage week by week.
What progress looks like once it stops feeling fast#
The plateau is partly a measurement failure. At A1 progress was visible daily because coverage climbed steeply. At B1 it is real but slow enough to be invisible over any window shorter than a month, which is why learners who are genuinely improving conclude they have stopped.
Replace the feeling with three numbers, tracked monthly: unknown words per hundred in comparable native text, the number of domains you can read without preparation, and the share of your speaking not drawn from your usual few hundred words. All three move slowly. The learners who get past the plateau are rarely the ones who found a better method; they are the ones who kept going through the stretch where the method stopped producing visible feedback. What B2 level means sets out the descriptors you are heading toward.