How to Learn Kanji: Radicals, Readings, and Volume

Ask English-speaking learners what the hardest part of Japanese is and almost all of them say the same thing: kanji. The 46 kana are a two-week problem. Kanji is a two-year one, and it comes with a number that sounds absurd — the government's jōyō list, the characters expected for ordinary literacy, currently runs to 2,136.

The difficulty is real, but it is badly understood. Most wasted effort in kanji study comes from three wrong beliefs: that kanji are pictures, that each one has a single meaning and reading, and that the way to learn them is one at a time from a list.

Kanji are components assembled by rules, their readings only make sense inside words, and they are learned through vocabulary rather than in isolation. Get those three right and the 2,136 stops being a wall.

Kanji are not pictures

The most persistent myth is that each character is a little drawing. A few are: 山 (yama, "mountain") looks like peaks, 川 (kawa, "river") looks like flowing water, 木 (ki, "tree") looks like a tree. Those are the ones textbooks show you, and they are wildly unrepresentative.

The large majority of kanji are phono-semantic compounds: one component hints at the meaning category, and the other carries the sound. This is the most useful single fact in kanji study, because it means a character is not an arbitrary 12-stroke drawing — it is a formula.

Once you can see 氵 (the water radical) you can predict something about 海 (umi, "sea"), 池 (ike, "pond"), 酒 (sake, "alcohol") and 泣く (naku, "to cry") before you learn any of them. That is the payoff: components turn thousands of separate memorisation tasks into a few hundred reusable parts.

Radicals and components: the system underneath

The traditional list has 214 radicals — the dictionary-classification system that dates back centuries. A radical is the component a dictionary files a character under, and knowing the common ones gives you meaning hints and a way to look characters up.

Note the honest limits. The component predicts a category, not a translation: 海 is a body of water, and so is 池, but knowing "water" does not tell you which is the sea and which is the pond. Use components to organise and to remember, not to guess meanings you have not learned.

Two smaller facts worth knowing early. Stroke order follows rules, not taste — top to bottom, left to right, horizontal before vertical — and correct order is what makes a shape reproducible from memory. And rendaku (sequential voicing) voices the second element of many compounds: 山 (san) plus 火 (ka) gives 火山 (kazan, "volcano").

On-yomi and kun-yomi: two readings, and no shortcut

Each kanji usually carries at least one on-yomi — a reading borrowed from Chinese over a thousand years ago — and at least one kun-yomi, a native Japanese reading attached to the character because it meant something similar.

The rough pattern, which is right often enough to be useful and wrong often enough that you should hold it loosely:

• A single kanji standing alone or with hiragana attached is usually kun-yomi: 山 (yama, "mountain"), 高い (takai, "expensive, tall"), 食べる (taberu, "to eat").
• Two or more kanji in a compound are usually on-yomi: 学校 (gakkou, "school"), 旅行 (ryokou, "travel"), 経済 (keizai, "economy").
• The exceptions are numerous and common. 手紙 (tegami, "letter") and 山道 (yamamichi, "mountain path") are both kun plus kun. 今日 (kyou, "today") and 大人 (otona, "adult") are read in ways that no component analysis will predict, and simply have to be learned as words.

The mistake to avoid is treating readings as a list to memorise against a character. A kanji does not have readings; words have readings. 生 (sei, shou, ikiru, umareru, nama and more) is not a character with six facts attached — it is a character that behaves differently in 学生 (gakusei, "student"), 生まれる (umareru, "to be born") and 生ビール (namabiiru, "draft beer"). Learn the words and the readings come along; learn the readings and you still cannot read anything.

Why learning kanji in isolation fails

This is the failure mode that produces the learner who "knows 800 kanji" and cannot read a menu.

A character is rarely a word on its own. Knowing that 上 (jou) means "up" and 手 (shu) means "hand" gives you two facts that do not combine into 上手 (jouzu, "skilled, good at"), and the keyword "cut" does not get you from 切る (kiru, "to cut") to 切手 (kitte, "postage stamp"). A character learned alone is stored as a single pair — shape to keyword — with no usage, no grammar and no sound, and that kind of memory decays fast.

What works instead is the principle that governs vocabulary: meet the character inside a word, inside a sentence, with its reading attached. Context is what makes retrieval possible later.

There is one legitimate exception: if handwriting is a goal, studying shapes as shapes is a real exercise, because writing is genuinely separate from reading. That is the logic behind one of the methods below.

The three approaches

These are the three schools you will encounter, presented as fairly as they deserve. None of them is a scam, and all three produce people who read Japanese.

Remembering the Kanji (RTK, Heisig) — shape-first

Learn to write all ~2,200 general-use kanji from an English keyword, using vivid imaginative stories built from components, with no readings taught at all. Readings are meant to arrive later, from vocabulary study.

• Strengths. Fast, systematic coverage of the writing system, with components introduced in a careful order. It produces handwriting most other methods never do, costs a book rather than a subscription, and removes the "I cannot write" anxiety early.
• Weaknesses. You finish able to write thousands of characters and read almost nothing — no readings, no vocabulary, no sentences. Keywords are English approximations you must eventually unlearn, and the system needs a separate vocabulary programme alongside it. The workload is heavily front-loaded, and many people abandon it in the 600–1,000 range.

WaniKani — integrated, paced, subscription

A spaced-repetition service that teaches radicals, then kanji, then vocabulary, in a fixed order with mnemonics written for you. Roughly 2,000 kanji and several thousand vocabulary items across 60 levels.

• Strengths. Everything is decided for you: what to learn, in what order, with a mnemonic and a schedule. Readings and vocabulary are built in, so you never have the RTK gap, and the pacing is designed to keep the review queue survivable.
• Weaknesses. It is a paid subscription and the pace is fixed — slow learners feel pushed, fast ones held back. Its "radicals" are invented names rather than the traditional 214, so its vocabulary for talking about kanji is not the vocabulary anyone else uses. It teaches recognition, not handwriting.

Context-based — vocabulary and reading first

Learn characters as they appear in the words and sentences you are actually reading: a textbook sequence, graded readers, or a sentence-based deck of real example sentences.

• Strengths. Everything you learn is immediately usable, with readings, usage and grammar attached, and there is no separate "kanji phase" delaying your ability to read. It scales naturally into reading practice.
• Weaknesses. Coverage is slower and less systematic, and you need reading material at your level — a real constraint early on. Handwriting is neglected unless you add it deliberately, and component knowledge arrives incidentally.

How to choose. If handwriting matters, take the RTK-style component training and pair it with vocabulary. If you want to read soon and will pay for structure, an integrated SRS service removes the most common cause of quitting, which is having to make decisions. If you already have reading material you enjoy, context-based study with an example-sentence deck is the lightest option that still works. What none of them survive is a learner who stops showing up: the method matters far less than the streak.

A realistic daily volume

The chemistry here is unforgiving, and most published advice understates it.

Two things about that table. "Time to cover" counts only the introduction of new items — the review load continues long after, and it is the reviews, not the new cards, that decide whether you can keep going. And a rough rule of thumb is that steady-state review demand runs several times the daily new-item count; treat any specific multiplier you see online as an estimate.

If you want one honest recommendation: 10 new characters a day, in words, with reviews cleared before anything new is added. If that is too much, drop to 5 rather than skipping reviews — an unpayable backlog, not the size of the language, is why people quit kanji.

You also do not need 2,136 characters before you can read something. A few hundred of the most frequent characters account for the majority of ordinary text (the exact percentages quoted online vary by corpus), so the goal is a few hundred well learned, plus a willingness to keep going.

A routine that works

• Learn each character inside a word. 学 (gaku) arrives as 学生 (gakusei, "student"), 学校 (gakkou, "school"), 大学 (daigaku, "university") and 学ぶ (manabu, "to learn") — one character, both readings, four uses.
• Read the word out loud. A kanji with no sound attached is a picture; pairing the shape with its reading is where pronunciation and kanji study meet.
• Write it once, with correct stroke order. Not twenty times. One careful production beats a page of copying.
• Review before adding, and keep the character in play. Seeing 学 again in a sentence next week is worth more than three more reviews of the isolated card.

Common mistakes for English speakers

1. Treating kanji as pictures. A handful are; the rest are component formulas. Learn the components and the pictures stop mattering.

2. Wanting one reading per character. There is no such thing. Readings belong to words.

3. Studying kanji as a separate subject from vocabulary. This is the single most common cause of the "I know 800 kanji and cannot read" outcome. Characters live in words.

4. Learning English keywords and mistaking them for meanings. 上手 (jouzu, "skilled") is not "up hand", and 大丈夫 (daijoubu, "all right, no problem") is not a large husband. Keywords are scaffolding for the shape.

5. Trying to write before you can read. Reading and writing are separate skills, and reading is the one that unlocks everything else. If handwriting is not your goal, recognition alone is a legitimate destination.

6. Adding 30 new cards a day and blaming your memory for the backlog. The queue, not your ability, is what breaks. Cap the intake.

7. Comparing your count to someone else's. People who "know 2,000 kanji" are usually counting recognition, which is a different inventory from the one used to write a postcard. Comparing across the two is how normal progress starts to feel like failure.

Summary

Kanji are components in a system, readings belong to words rather than characters, and progress comes from daily volume in context. Learn radicals to organise the shapes, learn on-yomi and kun-yomi through real vocabulary, choose a method that matches whether you care about handwriting, and cap your daily intake at something whose reviews you will actually clear. The characters are not the hard part; the consistency is.

Romaji.live puts kanji in the place where they stick: inside real sentences. Every practice item shows the sentence with its kana reading, and you type it word by word, so the character, its reading and its meaning get bound together in one pass — the word-family learning described above, done by default rather than as an extra exercise. Start with the kana table if the readings are not yet automatic, then try a practice session. For the vocabulary side of the same problem, see how to remember words; for the reading that makes them permanent, see reading practice.