# 400 Million People Learned English for 12 Years. Fewer Than 5% Can Speak It.

_The Unscript team · September 9, 2026_

A 580 on the national college English exam. Day one abroad, three minutes at a Starbucks counter, and not one coffee ordered.

Not because the words were missing. Because the head was translating and the mouth couldn't keep up. Twelve years of English class, and not one lesson on the thing that mattered at that counter: getting a complete sentence out while someone waits for the answer.

### A 580, and no coffee

580 out of 710 on the CET-6, the national English exam that Chinese university students take. Two reading questions wrong, a few points off on translation, listening slightly above average. By the school's standard, this is a student whose English is "pretty good".

First day in another country, standing at the Starbucks counter. The barista says "What can I get for you?" and the brain starts its procedure: translate what was said into Chinese, confirm it was understood, translate the intended reply back into English, check the grammar, arrange the word order, and then—

The people behind have started to form a line.

In the end a finger points at a picture on the menu, and one word comes out: "This."

This is not a joke. Search "first day abroad" in any Chinese English-learning forum and there are enough of these posts to fill a book. The details vary; the scene never does. Someone who did fine on exams discovers, in a real situation that requires speaking, that their English is read-only.

### Twelve years of English class trained the eyes, not the mouth

Counting from third grade, when English class begins for most Chinese students, to university graduation, that is at least twelve years of English. Counting from first grade or kindergarten, as many families do, it is more than fifteen.

Nobody has counted how many English exam papers a student completes in those years, how many words they memorise, how many grammar rules they learn. But one thing is close to certain: across the entire twelve years, **the number of times a student is required to say a complete stretch of English to another person is, in all likelihood, in the single digits.**

The reason is not complicated. English education in China is designed around exams, and exams need standardised grading. Reading comprehension has a right answer. Grammar fill-in-the-blanks has a right answer. Cloze passages have a right answer. Speaking doesn't.

A teacher can grade a reading question in under a minute. A spoken answer has to be listened to in real time, judged for grammar and meaning in real time, scored in real time. Fifty students in a class, two minutes each: the listening alone takes nearly two hours.

**It is not that teachers don't want to teach speaking. The system never left room for it.** Exams decide what is taught, teaching decides what happens in the classroom, and the classroom decides what students practise. End to end, this system is optimised for grading efficiency, not for the ability to communicate.

This is where cause and effect got flipped. The story we were told: grammar and vocabulary are the foundation of a language, so we teach them first. The actual story: grammar and vocabulary are the easiest things to turn into multiple-choice questions, multiple-choice questions are the easiest things to grade at scale, so the whole system has revolved around them for decades. The result was repackaged as the reason. The means was promoted to the goal.

### 400 million people, 5 percent

In 2014 a widely cited figure put the number of people in China who had studied English at around 400 million, with the country's English proficiency ranking second to last among Asia's major economies. Those able to hold a conversation in English without difficulty: no more than 5 percent.

Five percent of 400 million is 20 million. Put another way, the other 380 million learned, but did not learn to do it. At least not to speak.

Of those 380 million, how many can read an English news article? Probably quite a few. Follow an English podcast? Some. But open their mouth in a real situation and say a complete sentence while someone waits for it?

The people who can do that are rare enough to be praised as "good at English".

In 2023 China's English training industry was worth more than 100 billion yuan, roughly 14 billion US dollars. That figure does not include the school system itself: public-school English teachers' salaries, textbooks, the cost of running the exams. Those are paid by the state and are not counted in the market size.

**An industry that takes in over 100 billion yuan a year, whose core output, "can this person talk to someone in English", passes fewer than 5 percent.** If this were a product, it would have been recalled long ago.

### You think you're missing words. You're missing the exit.

Most people blame "not enough vocabulary" for not being able to speak. That is another reversed cause.

Linguistics has a basic pair of terms: receptive vocabulary and productive vocabulary. You know a word, you see it and understand it: that is receptive. You can bring it out yourself while speaking: that is productive.

A student who has passed the CET-6 has a receptive vocabulary of roughly 5,000 to 6,000 words. Productive? Possibly a third of that, or less. And everyday English conversation does not actually need much. Research has suggested that the 2,000 most frequent word families cover more than 90 percent of ordinary speech.

The problem was never that you don't know enough words. The problem is that the words you know have never been through the step of being said out loud, under pressure.

Cognitive psychology calls this retrieval practice. If you only look at something repeatedly (recognition), the memory stays shallow. If you are required to actively recall it and produce it (retrieval), the memory goes much deeper. A vocabulary app's "see the English, pick the meaning" trains recognition. "It's your turn in the conversation, pull the word out of memory and say it" trains retrieval.

In twelve years of English education, we did recognition practice countless times and retrieval practice almost never.

### The 580 and that cup of coffee

Back to the Starbucks counter.

The student with the 580 knows how to spell latte, knows that large means the big one, knows that "Can I have…" is a legitimate sentence frame. All of it is in there, stored intact.

What is missing is not knowledge. It is the path. Between knowing and saying there is a stretch of road that has never been walked. In twelve years, not one class practised this one thing: with someone waiting and no prompt to look at, get a sentence out of your mouth.

There is a way to design it the other way round. Not "finish learning, then speak", but speak from day one: start by reading the line off the page, then see a little less of the script each time, until you can say it with nothing to look at. Language teaching calls this scaffolding fading. Give the learner full support, then remove it piece by piece until they can stand on their own.

This is not a new theory. Pimsleur was doing it in the 1960s. But in China's English classrooms it has almost never been tried, because you cannot turn it into a multiple-choice question.

The coffee did get ordered in the end. Reportedly, a Chinese student further back in the line stepped up and did the talking.

---

Originally published at: https://hintlib.com/blog/12-years-of-english-5-percent-speak/
Unscript, English speaking practice for iPhone: https://apps.apple.com/app/id6803985536