The science of learning a language through stories, and what it means for Chinese

Does learning a language through stories actually work?

Yes. Three meta-analyses, which pool many studies, agree that reading a lot of easy, interesting text improves a second language. We read all three; here's what each found:

We'd call the headline effect, 0.57, moderate: real, but not magic.

Two details from the newest review matter if you learn on your own. Effects were larger when there was some accountability, such as a quiz or a summary. They were also larger when learners chose from a limited set of texts, rather than anything at all.

One caveat: most of the studies we read are of English learners. And Chinese adds a problem English doesn't have: you can't sound out a character you haven't learned.

The Chinese research we found is small, but it points the same way. In one study, 30 intermediate learners at an American university read two stories. The group whose stories had glosses in the margin learned significantly more of the new words.

In another, teachers added short stories they'd written to a first-semester Chinese course. Students reported better attitudes to reading, more vocabulary and faster reading, though that study has no numbers to compare.

Stories stick better than explanations

Stories are easier to understand and remember than explanations of the same material. A meta-analysis of 78 samples, covering 33,078 people, found that narrative texts beat expository ones by an effect size of 0.55 overall.

The edge was smaller for comprehension, at 0.48, and bigger for memory, at 0.72.

Those studies are mostly of people reading in their first language. But the mechanism is easy to see: a plot gives every sentence a reason to exist, so you predict what comes next and notice when you're lost.

Stories also repeat words naturally. A rabbit that's "fast", 快(kuài), in the first chapter is still fast in the fourth.

Repetition is the strongest predictor of picking up a word while reading. Across 26 studies, the number of encounters correlated with learning at 0.34.

Graded-reader publishers build on this. Mandarin Companion says learners need to meet a word 10 to 30 times, and Du Chinese writes repetition into its stories on purpose.

How many words do you need to know to read a story?

About 98% of the words on the page, if you want to read without help. In the study that set this figure, 66 learners read a story with different shares of unfamiliar words:

Source: Hu & Nation, 2000.

A later study proposed 98% as the optimal threshold and 95% as the minimum, Laufer and Ravenhorst-Kalovski wrote.

We treat 98% as a guide, not a law, and you should too. A replication with 104 adult learners in Sri Lanka couldn't fully reproduce the original result. Comprehension falls off gradually as unknown words pile up; there's no cliff.

Still, the arithmetic is sobering. At 98%, one word in fifty is new. At 90%, it's one in ten, which in a story means several per sentence.

For Chinese, we see two extra hurdles:

• Words vs characters. Chinese is counted in both. You might know "day", 天(tiān), and "air", 气(qì), but not "weather", 天气(tiānqì), as a word. Or you might know a word by sound but not its characters.
• Native text is far beyond the textbook lists. Even the full HSK vocabulary leaves about one word in ten unknown in film and TV subtitles, as the chart below shows.

Our HSK 3.0 guide matched the official word lists against 33.5 million words of Chinese film and TV subtitles, from SUBTLEX-CH. Here's how much of what's said on screen each list covers:

Source: SayMei's HSK 3.0 analysis of SUBTLEX-CH. Word boundaries in Chinese are fuzzy, so the counts are approximate.

That's why native material feels like a wall for years. Graded stories exist to lower that wall.

We built a tool to show what that means. Pick the HSK level you know and one of our stories, and it counts the words you'd already know. It shows where the passage falls against the 95% and 98% lines:

Here's what that looks like in one sentence: the opening of our Mulan story, for a reader who knows the 500 words of HSK 1–2.

很久很久以前，中国北方有一个姑娘，叫木兰。她每天都坐在门口织布。

Hěn jiǔ hěn jiǔ yǐqián, Zhōngguó běifāng yǒu yí gè gūniang, jiào Mùlán. Tā měi tiān dōu zuò zài ménkǒu zhībù.

"Long, long ago, in the north of China, there lived a young woman named Mulan. Every day, she sat weaving in her doorway."

Six of the 18 counted words are new to that reader, and Mulan's name isn't counted:

That's 67% coverage: readable with help, and nowhere near "reading for pleasure".

One story teaches less than you'd hope

You pick up only a small share of the new words in any one text, which is why volume matters more than anything else. A meta-analysis of 24 studies with 2,771 learners measured how much new vocabulary sticks. Here it is, set against deliberate study:

Sources: Webb, Uchihara & Yanagisawa, 2023 for the first three rows; Webb, Yanagisawa & Uchihara, 2020 for deliberate study.

One classic study tested 25 new words after learners read one graded reader. Three months later, the average reader remembered the meaning of only one, and none of the words met fewer than eight times.

That sounds discouraging. We don't think it should be. A story isn't a vocabulary lesson: it consolidates words you half-know, teaches a few new ones and builds reading speed.

The pickup comes from reading a lot. Researchers who analysed a graded-reader series suggested reading about one graded reader a week, so new words come back before you forget them.

Stories and flashcards work best together. Stories give you meetings with a word, and flashcards make sure some of those meetings stick.

Listening while you read speeds you up

Listening while you read teaches about as many words as reading alone, as the table above shows. It also has an advantage reading alone lacks: it speeds reading up.

In one study, 64 beginners each read 20 graded readers over half a year, either silently or with the audio. Both groups improved, but the audio group improved substantially more in reading speed and comprehension, and it kept the gains three months later.

For Chinese, there's a further reason: audio carries the tones. Pinyin tells you that "young woman" is 姑娘(gūniang); a recording tells you how the neutral tone on niang actually sounds in a sentence.

Slowing it down is fine. Our story reader starts at 0.8× speed, and Mandarin Companion recorded its audiobooks at about 120 characters a minute.

Pinyin helps at first, then partial pinyin wins

Pinyin gets a beginner reading real sentences from week one. But beginners also tend to read the pinyin instead of the characters.

In an eye-tracking study, all 10 beginner and lower-intermediate learners looked at the pinyin to some degree when it sat above the characters.

The most useful finding is about timing. One experiment compared full pinyin, partial pinyin and none. For learners with little prior knowledge, it made no difference; for more experienced learners, partial pinyin worked best.

Here's the rule we'd suggest:

• Beginners: keep pinyin on, but read each line twice, the second time looking only at the characters.
• Around HSK 3: switch pinyin off and tap only the words you don't know. Many apps can show pinyin over new words only.
• Always: keep a gloss within reach. Glosses helped word learning in the one Chinese study that tested them.

Graded stories cap their vocabulary

A graded story limits its words to a list, so a learner at that level knows most of them. Series grade in different ways, and some apps show your own coverage before you start:

Level labels vary from series to series, so we'd check your own coverage, not the label.

Our stories are built for assisted reading

We measured our own library the same way, and the result is humbling. The chart shows the median share of words a reader would know in each level's books, in two cases: knowing only the HSK words up to that level, and knowing all 2,000 HSK 1–4 words.

Source: our count over all 166 books on 6 October 2026: the median share of running words, names excluded, against the HSK 3.0 word lists of 2025. Method and data.

No SayMei book reaches 95% for a reader who knows only the HSK words of its own level. Even for a reader who knows all of HSK 1–4, the median beginner book comes out at 93%.

The gap comes partly from the stories themselves. Legends need words like "to weave", 织布(zhībù), and "the Jade Emperor", 玉帝(Yùdì), that no HSK list contains. Partly, the HSK lists are just short at the bottom: HSK 1 is only 300 words.

So we want to be clear about what our stories are for. They're built for assisted reading, with pinyin over every character, an English line under every sentence, audio for every line and a dictionary under every word.

Used that way, they suit the "listen and read with support" stage. They aren't unassisted extensive reading at the 98% level for a reader who knows only those words. If that's what you want, choose a series with a strict cap, or check your coverage before you start.

A simple plan for reading Chinese stories

Here's the plan we'd follow, based on the research above:

1. Pick for coverage, not for level labels. Aim for 95% or better with support, and 98% for fast "just reading". Check with a tool like the one above, or an app that shows your known-word percentage.
2. Read, listen, reread. On the first pass, turn audio on at 0.8×, keep pinyin on, and hide the translation unless you're lost. On the second pass, read the characters only, at normal speed.
3. Go for volume. Aim for a short graded book, or about five chapters, a week, as Nation and Wang suggest. Twenty minutes a day beats an hour on Sunday.
4. Add a little accountability. Answer the questions at the end, or retell the story in two sentences. Effects were larger when learners had something to answer for, in the 2025 review.
5. Move the best words to flashcards. Pick five or so per story, the ones you've now seen several times.
6. Climb when it feels easy. When you rarely tap a word, move up a level, or switch pinyin to new words only.

Where SayMei Stories fits, and where it doesn't

SayMei Stories is our library of graded stories. Here it is at a glance:

Source: SayMei's live site and code, checked 6 October 2026. You can also switch the stories to Traditional characters.

Here's where it loses, plainly:

• The library is small next to Du Chinese's 3,000+ readings.
• It stops around HSK 4.
• It runs only in a browser, with no offline mode.
• By the coverage measure above, its levels are generous.

We compare it with the alternatives in our guide to the best apps to learn Chinese through stories.

What to do next

• Check one chapter's coverage. Use the tool above with your level before you commit to a series.
• Read each story twice. Listen with pinyin first, then read the characters alone.
• Read a little every day. Volume, not any single story, is what makes words stick.
• Learn the sounds and check your level. The pinyin chart and the free level test are good places to start.

Then read chapter 1 of any of our stories. It's free, with no sign-up.