Kumi data report
Why the easiest kanji have the most readings
The 500 most common kanji carry an average of 3.19 distinct readings. Kanji ranked 1001 and beyond average 2.65. The characters you meet in your first month are the ones with the most readings to tell apart, which is the opposite of how a curriculum is usually described.
How many readings does a common kanji actually have?
The median is 2. Across all 2,501 frequency-ranked kanji the mean is 2.77, and 334 of them have exactly one reading. So far this sounds manageable.
The average is not what hurts you. 269 kanji carry five or more distinct readings and 15 carry eight or more. Those are not obscure characters filed away for later, which is the part worth knowing.
| Frequency band | Mean readings | Kanji |
|---|---|---|
| The 500 most common | 3.19 | 500 |
| Ranks 501 to 1000 | 2.68 | 500 |
| Ranks 1001 and beyond | 2.65 | 1,501 |
| Kanji | Rank | Total |
|---|
Why do the most common kanji have the most readings?
Because they are old and they are useful. A character that turns up in everything gets borrowed into more compounds, picks up more Chinese readings across the centuries in which they were imported, and attaches to more native Japanese words that already existed. Rare kanji never had the opportunity. They were imported once, for one purpose, and kept one reading.
Frequency and simplicity are not the same axis. A curriculum ordered by frequency, which is nearly all of them, front-loads the characters with the most reading ambiguity while a learner has the least context to resolve it with.
What this means in practice
Only 47 of the 500 most common kanji have a single reading. For the other 453, "knowing the kanji" is not a thing you can finish. The reading is a property of the word, not of the character. That is why Kumi teaches readings through vocabulary rather than asking you to recite a character's reading list, and why a kanji you have "learned" keeps coming back attached to words you have not.
Method
- Scope is the 2,501 kanji carrying a frequency rank. Ranking all 13,108 in the dictionary would put characters nobody reads at the top.
- Okurigana is normalised away. 生 is stored as い.きる, い.かす, い.ける and so on, but only the part before the dot is the kanji's reading, so those three count once. That takes 生 from 18 raw entries to 10 distinct kun stems.
- A stem is counted once per kanji, so this measures distinct reading stems rather than distinct words. Some stems are okurigana variants of a single verb (生 records う, うま and うまれ for 生まれる), and collapsing those automatically merges genuinely different readings elsewhere, so they are left as recorded.
- Name-only readings (nanori) are excluded. They are real, but they would inflate the difficulty of characters that are not actually hard to read.
- Numbers are computed live from Kumi's dictionary each day, so this page cannot drift from the data behind it. Last generated 2026-08-19.
Reading and frequency data derive from KANJIDIC2, property of the Electronic Dictionary Research and Development Group, used under CC BY-SA 4.0. The counts, normalisation and rankings on this page are Kumi's own. See attributions.
Kumi teaches kanji readings through the words that carry them, with a knowledge graph connecting radicals, kanji, vocabulary and grammar.