How to learn Chinese: what the research says works

The best way to learn Chinese, at a glance

Here are the eight habits with the best evidence behind them, as we rated it. Five come from general learning science, and three rest on evidence specific to Chinese:

Each row has its sources in the sections below. "Strong" means several meta-analyses or large experiments agree; "Moderate" means real support with a lot of variation between studies.

Build your week

We built the planner around these habits. Tell it your goal, level and daily time, and it lays out a typical week. Every activity names the principle behind it and how strong the evidence is, and everything it links to is free.

The default week, for a complete beginner at 15 minutes a day, starts every study day with flashcard review. That order is the first habit in action.

Test yourself, and space the tests

Test yourself instead of rereading, and spread the tests over days. It's the best-supported finding I came across.

In a classic experiment, students learned 40 Swahili–English word pairs. Once they'd recalled a pair, some kept being tested on it and some didn't. Here's how much they remembered a week later:

Source: Karpicke & Roediger 2008, Science. "Kept testing" covers the two conditions that kept testing a pair, with or without more study.

The students couldn't feel the difference: every group predicted it would remember about half. You can't trust the feeling of knowing a word, so test it.

The finding holds up at scale. Across 222 classroom studies, quizzing raised achievement by a medium amount. And recalling an answer beats picking it from options.

Spacing matters as much as testing. A 2022 meta-analysis of second-language experiments found a medium-to-large benefit from spreading practice out, with longer gaps winning on delayed tests.

The best gap between reviews also grows with how long you want to remember, a large study found. You don't need to work it out: a flashcard app with a scheduler does it for you.

What students actually do is the opposite. A review of ten study techniques rated rereading and highlighting "low utility", and practice testing and spaced practice "high utility". Yet most of 177 surveyed students reread their notes rather than testing themselves.

Here's what I'd do:

• Review once a day, before anything new. Say the word out loud, then flip.
• Test both directions: see 汉字(hànzì) and recall "Chinese characters"; see "Chinese characters" and recall 汉字(hànzì).
• Use any spaced-repetition app. SayMei's flashcards are free with an account for 10 words a day, and Anki is free and unlimited.

Train your ear for tones first

Get the tones right from the start, and train your ear before you drill your mouth. Mandarin has four tones plus a neutral tone, and the same syllable with a different tone is a different word: mother, 妈(mā); hemp, 麻(má); horse, 马(mǎ); and to scold, 骂(mà).

Here we found unusually direct evidence. Eight American learners did eight listening sessions over two weeks. They identified tones 21% more accurately, carried the gain to new words and new voices, and still had it six months later.

Their speech improved too, without any speaking practice: native listeners identified their tones 18% more accurately afterwards. And across 25 years of studies on many languages, listening training improved perception more than pronunciation.

So listening isn't enough on its own. In one study of 38 learners, those who practised only listening or only speaking did much worse on the skill they skipped. And across 86 studies, pronunciation teaching worked better when learners got feedback.

Can't people work out the meaning from context? In quiet, mostly: sentences with the pitch flattened stayed about 94% intelligible, one study found.

In background chatter, they fell to 60%, against 80% for normal speech. Real life has background chatter.

Tones also change next to each other. Two third tones in a row turn the first into a rising tone, so hello, 你好(nǐ hǎo), is written nǐ hǎo but said ní hǎo.

My advice: spend five minutes a day on two-syllable tone pairs with the tone-pair trainer. Then record yourself with Tone Trial and compare your pitch shape with the model.

One caveat about our own tool: Tone Trial's accuracy has so far been measured only on synthetic voices, so treat it as a practice aid, not a judge. The tone sandhi checker covers the tone changes, and how to learn Chinese tones as an adult has the full method.

Start with pinyin, then add characters

Start with pinyin for roughly your first two weeks, then learn characters through words you can already say. I haven't found a study that tests this order directly. It's common teaching practice, and it follows from the evidence below.

Pinyin, 拼音(pīnyīn), spells Mandarin sounds with the Roman alphabet. It's how you'll look words up, type messages and see the tones. The characters, 汉字(hànzì), are what you read.

Two studies with adult learners show why you want both:

• Handwriting helps reading. Writing characters by hand improved recognition and the link to meaning, while typing pinyin helped only with the sound, one study found.
• Each kind of practice trains its own skill. In another, writing was best for reproducing characters from memory, and reading was best for recalling their meaning and sound.

So type in pinyin, and handwrite a few characters a day if you want to read.

Characters are less arbitrary than they look. Many combine a meaning part and a sound part: mother, 妈(mā), is woman, 女(nǚ), plus horse, 马(mǎ), for the sound.

The sound part gives a hint, not the tone. Please, 请(qǐng); clear, 清(qīng); sunny, 晴(qíng); and feeling, 情(qíng), all share 青(qīng).

One catch: among the 2,570 characters taught in Chinese primary schools, the most common were the least likely to have a reliable sound or meaning part. Components pay off more as you go.

You also need fewer characters than you might fear. In Chinese film and TV subtitles, the 356 most frequent characters make up 80% of the characters on screen, by our count.

Learn 1,111 and you're at 95%.

Learn the most frequent words first

Learn the most frequent words first, and let the HSK word lists stand in for frequency. To see how much real Chinese each level covers, we matched the HSK 3.0 lists against film and TV subtitles and against news articles:

Source: our count on SUBTLEX-CH (Cai & Brysbaert 2010, CC BY) and the Chinese Wikinews dump of 1 October 2026 (CC BY 2.5), against the HSK 3.0 syllabus. Full method and data in the characters guide.

In our count, the first few hundred words do a lot of work in conversation. News is another matter, and it stays hard for a long time.

Covering words isn't the same as understanding. For English, readers need to know about 98% of the words in a text to follow it unaided.

Reaching that takes 8,000–9,000 word families, and a later replication has retested the 98% line itself.

My advice: learn words in the list's order, a few a day, in short phrases rather than alone. The HSK word lists have audio.

The new HSK's first regular test is on 13 December 2026. HSK 3.0 explained covers what changed, and the HSK exam guide covers the test.

Read and listen to what you mostly understand

Read and listen as much as you can fit in, at a level where you already know nearly every word. Input builds comprehension well, but on its own it's a slow way to learn new words.

Extensive reading programmes show medium effects across 34 studies, and small-to-medium effects on reading proficiency across 49.

For vocabulary, the numbers are humbling. Here's how many new words learners picked up, by how they met them:

Source: Webb, Uchihara & Yanagisawa 2023, 24 studies of reading and listening; Webb, Yanagisawa & Uchihara 2020, 22 studies of deliberate study. The two rows come from different sets of studies, so treat the gap as rough.

Meeting a word more often helps, though the link is moderate. So I'd use input alongside deliberate study, not instead of it.

Listening has a bright spot. In 32 studies of learning words from speech, gains were large. Interactive tasks beat passive ones, and adults learned more than children.

Here's what I'd do:

• Start with material made for learners: free listening dialogues with transcripts, and graded stories such as SayMei Stories, where chapters 1–2 of every book and 8 full stories are free. Our graded readers comparison covers others, including ones with human narration.
• Replay hard lines instead of pushing on.
• Move to dramas when you can follow most of an episode. Before that, a show is mostly noise.
• Songs add repetition. The 175 free HSK 1 songs are one way in, though the evidence on songs specifically is thin.

Speak early, and get corrected

Speak from the first week, and get corrected. Correction is one of the better-tested ideas in language teaching.

Here's what the research on speaking practice found:

Source: the studies linked in each row; effect sizes are in the Sources list.

Correction is the step people most often skip when they study alone, so I'd build it in on purpose. Your options, roughly from most to least human:

• A tutor gives the richest correction and accountability, at a price per hour.
• A language-exchange partner is free, but correction depends on the partner.
• An AI tutor is available at any hour. Treat the chatbot evidence as promising, not settled: most of those studies involved learners of English, not Mandarin.

If you have no one to talk to, how to practise speaking Chinese without a partner covers the options. The best AI apps for learning Chinese compares the AI tutors, SayMei included.

Try it: a free 10-minute speaking lesson with Mei Lin. No account, card or download. You'll say Chinese out loud in the first minute, learn one sentence pattern, and learn to put a question back to her: "and you?", usually 你呢(nǐ ne).

Short grammar explanations help

A short explanation followed by practice usually beats working a pattern out on your own. Two meta-analyses, of 49 and 41 studies, found explicit teaching more effective than implicit teaching.

A later meta-analysis of 54 studies found instruction worked well overall, but only a minor difference between explicit and implicit. The size of the edge is debated; the value of focused practice isn't.

Many beginner patterns are short. "My name is…", 我叫(wǒ jiào)…, is usable in your first lesson.

My suggestion: learn one pattern at a time from the grammar lessons. Say five examples aloud, then write three of your own and check them with Sentence Lab.

Study most days, even briefly

Chinese takes longer than Spanish or French, and how regularly you study matters more than how long each session lasts.

The US Foreign Service Institute gives its diplomats 88 weeks, or 2,200 class hours, to reach professional speaking and listening in Mandarin. Spanish and French take about a third of that.

Most of us have less time than a diplomat. Thirty minutes a day for six months is about 90 hours: a real beginner foundation, nowhere near fluency.

How long does it take to learn Mandarin? converts the official estimates into months at your pace, and is Chinese hard to learn? explains where the time goes.

Here's the daily goal our new learners picked when they signed up:

Source: SayMei onboarding data, aggregated: about 840 new learners in the two weeks to 6 October 2026, internal accounts excluded.

Any of those works if it happens most days. Fifteen to thirty minutes is a sustainable base, and the planner above shows how to split it.

Make it a habit you keep

Attach Chinese to something you already do every day, and don't panic about a missed day.

In a study of volunteers building a daily habit, it took anywhere from 18 to 254 days for the behaviour to become automatic.

The median was 66 days, and missing a single day didn't materially set people back.

The cheapest tool here is an if-then plan: "After my morning coffee, I'll do ten minutes of Chinese at the kitchen table." Across 94 tests, plans like this improved goal attainment by a medium-to-large amount.

Motivation matters too. Across 75 samples and 10,489 learners, motivation was more closely tied to achievement than attitudes or reasons for learning were.

Keep a reason in view, whether it's a trip, a test date or a relative you want to talk to.

Popular advice that doesn't hold up

Three pieces of common advice I'd drop: matching study to your "learning style", relying on immersion from day one, and rereading your notes. The explorer rates 17 popular practices, with sources.

• Learning styles. A review for Psychological Science in the Public Interest found no adequate evidence that teaching people in their preferred "visual" or "auditory" style helps. Several well-designed studies contradicted it.
• Immersion alone, from zero. Learners keep only 6–17% of the new words they meet incidentally, as the input section showed. Input works best once you understand most of it.
• Rereading and highlighting. Rated low utility next to self-testing, as the study-techniques review above found.
• "Tones don't matter, people use context." Not in a noisy room, as the flattened-pitch study showed.
• "Method X is the fastest proven path." I haven't seen a study that compares whole methods for learning Chinese head to head. What's tested are the ingredients.

Shadowing, repeating a recording right after the speaker, is promising but thinly tested. One four-week study with Mandarin learners found better tones in spontaneous speech, with 14 participants.

Pick one tool for each job

I'd pick one tool for each job: a spaced-repetition app, a source of graded listening and reading, a dictionary, and a way to speak with correction. Tools matter less than whether they cover those jobs.

• Comparisons: the best apps to learn Chinese, the best AI apps, graded readers and our Duolingo review, all linked on this page.
• Free, not from SayMei: Anki for flashcards, a dictionary app such as Pleco, and the official HSK 3.0 sample papers from Chinese Testing International.
• Free from SayMei: the pinyin chart, the worksheet generator and a free 30-day study plan as a PDF. The tone-pair trainer, HSK word lists, Sentence Lab and listening dialogues are linked above.

Start with Mandarin and Simplified characters

I'd choose Mandarin and Simplified characters, unless your family, job or destination says otherwise. China's language law sets Mandarin, 普通话(Pǔtōnghuà), and standardized characters as the standard spoken and written forms of Chinese, China Daily reported. The HSK tests Mandarin.

If your family speaks Cantonese, or you're moving to Hong Kong, read Mandarin vs Cantonese. If you'll read mostly in Taiwan or Hong Kong, read Simplified vs Traditional; SayMei can show Traditional characters throughout the app.

If you grew up hearing Chinese at home but don't speak it, this guide for heritage learners is for you.

Where SayMei fits, and where it doesn't

SayMei is built for the step people most often skip when studying alone: speaking with correction. You talk with Mei Lin, an AI tutor, in 5–10 minute lessons built on one pattern each, and most of the practice tools in the planner are free.

Where it falls short: there's no human teacher, and a good tutor still gives richer feedback and accountability. No study has tested whether SayMei's lessons beat other methods, and Tone Trial's accuracy has only been measured on synthetic voices.

After two free lessons, more speaking time is part of SayMei Premium, at $5.99 a month for 200 minutes. Its seven-day trial needs a card.

Free flashcards stop at 10 words a day, where Anki has no limit. And SayMei doesn't teach Cantonese.

What to do next

• Start a daily review habit. A few minutes of flashcards before anything new, every day.
• Train your ear on tones this week. Short tone-pair drills first, then record yourself.
• Learn words in frequency order. Follow the HSK lists, a few a day, in short phrases.
• Find someone, or something, that corrects you. A tutor, a partner or an AI tutor: pick one and book the first session.
• Tie Chinese to a daily cue. Write your if-then plan today.

Build your own week with the planner above, then take the free speaking lesson to start the speaking part today.