Characters chosen for sound
Some words are written with kanji picked to represent the sounds rather than the meaning. The characters spell the word phonetically and their individual meanings are irrelevant or only loosely connected.
This was a common way of writing foreign words and place names before katakana took over that role, which is why many survive in older vocabulary.
| Japanese | Romaji | Meaning |
|---|---|---|
| 寿司 | sushi | The characters are chosen for sound |
| 珈琲 | kōhī | Coffee, now usually in katakana |
| 亜米利加 | Amerika | America, now in katakana |
| 出鱈目 | detarame | Nonsense: characters unrelated to meaning |
| 滅茶苦茶 | mechakucha | A mess: sound only |
| 兎に角 | tonikaku | Anyway: written phonetically |
The last three appear in ordinary writing, though often in kana. Recognising them saves puzzling over characters that carry no meaning.
Readings that apply to the whole word
A second category has readings attached to the entire compound rather than to each character. You cannot divide the sound between the characters, because it was never assigned that way.
Several of these are among the first words a learner meets, which is why the phenomenon needs explaining early.
| Japanese | Romaji | Meaning |
|---|---|---|
| 今日 | kyō | Today: the reading covers both |
| 昨日 | kinō | Yesterday |
| 明日 | ashita | Tomorrow |
| 大人 | otona | Adult |
| 一人 | hitori | One person |
| 二十歳 | hatachi | Twenty years old |
Everyday irregulars
Beyond the two named categories, a set of common words simply take readings that cannot be predicted. They are frequent enough that they are learned early and rarely cause trouble afterwards.
The practical approach is the same: learn the word, not the characters.
| Japanese | Romaji | Meaning |
|---|---|---|
| 大丈夫 | daijōbu | All right |
| 下手 | heta | Unskilled |
| 上手 | jōzu | Skilled |
| 田舎 | inaka | Countryside |
| 土産 | miyage | Souvenir |
| 眼鏡 | megane | Glasses |
Several of these are among the most common words in the language, which is why they are usually learned as vocabulary long before the reading question arises.
Why this exists
Japanese adopted a writing system built for a different language, and the fit was never exact. Characters were pressed into service for sound where meaning did not fit, readings attached to words rather than characters, and conventions accumulated over centuries.
Knowing that these are historical accidents rather than rules you have failed to learn is genuinely useful, because it stops you looking for a pattern that is not there.
| Category | Characters chosen for | How to learn it |
|---|---|---|
| Phonetic use | Sound | As vocabulary |
| Whole-word reading | Meaning | As vocabulary |
| Everyday irregulars | Mixed | As vocabulary |
| Regular compounds | Meaning and sound | Decodable |
Three of the four rows have the same answer. The category matters less than recognising that the word is not decodable.
Recognising the category quickly
The useful skill is noticing early that a word does not decode, so you stop trying. A few signals make that quick.
Furigana is the strongest one: if a text supplies the reading for a word you would expect to be readable, it is often because the reading is irregular.
- The reading has fewer syllables than the characters suggest.
- The characters' meanings have no relationship to the word's meaning.
- The word is very common but the characters are not.
- A dictionary marks the reading as applying to the whole word.
- Furigana appears in text where it otherwise would not.
What to do with them
Treat them as vocabulary. Learn the word, its reading and its meaning as a unit, and do not attempt to attach parts of the reading to individual characters.
This is also a good reason to learn kanji through words rather than in isolation, since a word-first approach handles irregular readings without any special treatment: they are simply words, learned the same way as any other.
The consolation is that the set is small and mostly common, so it is encountered early and then largely done.
If a word will not decode, stop decoding it. Recognising that quickly is worth more than any amount of reasoning about characters that were never meant to be reasoned about.