Japanese runs three vocabularies at once and you can see which is which on the page. Native words are written with kanji plus hiragana endings and they are the everyday register. Words built from Chinese readings are written as kanji compounds and they are the formal and technical register. Borrowings are in katakana and they are the modern one. Two words can mean nearly the same thing and be completely different sentences, and the script is telling you which.
So the useful question about a new word is not only what it means. It is which stratum it belongs to, because that decides where it can be said.
Six facts about the words themselves
- loanwords are written in katakana, which means a large slice of modern vocabulary announces itself visually and is often guessable from English.
- アルバイト came through German, from Arbeit, and means a part time job. Not every katakana word is from English, which is why guessing works less often than it looks.
- 同音異義語, homophones, of which Japanese has an enormous number because Chinese readings collapsed into a small sound inventory. Context and kanji separate them.
- onomatopoeia is ordinary grammar, not childish. Thousands of these words describe manner, texture and emotional state, and they appear in newspapers and novels.
- ぴかぴか, sparkling or spotless, is a standard adult word and the source of a famous character name.
- the counter changes the word. One, two and three take different readings depending on what is being counted, so numbers have to be learned per counter rather than once.
Three layers, three registers, three behaviours
The native layer carries the verbs and adjectives that do the work of daily life, and it inflects: the endings change and the hiragana attached to the kanji is where that happens. The Sino-Japanese layer is largely nouns, assembled from two characters the way English assembles from Latin roots, and it is the vocabulary of newspapers, contracts, science and anything official. The borrowed layer is mostly nouns too, it arrives already reshaped by Japanese phonology, and it carries a flavour of the modern, the technical or the commercial.
The parallel with English is close enough to use and it runs in the same direction. An English speaker reaching for a formal word reaches for the Latin one; a Japanese speaker reaching for a formal word reaches for the Chinese-derived one. Which means an English speaker's instinct about register transfers, once they know which layer they are looking at, and the script tells them that before they have read the word.
The complication is that the layers overlap in meaning. There is frequently a native word and a Sino-Japanese word and a borrowing for the same idea, all three current, and choosing among them is a register decision rather than a vocabulary one. That is a real cost and it is not a difficulty of memory.
Assuming the borrowings are free
Here is the stall, and it catches English speakers specifically. The katakana layer looks like a discount: thousands of words that came from English, sitting in a script you can decode. It is not a discount and it is a trap in three separate ways.
The phonology reshapes the word until it is unrecognisable by ear, so recognising it on the page and catching it in speech are different skills. The meaning has often narrowed, so a borrowed word covers one sense of the English original and not the others. And a large number of them were assembled in Japanese out of English parts and are not English words at all, which means confidently using one in English is a mistake in the other direction. A learner who treats this layer as free will read it slowly, mishear it constantly, and be wrong about a fixed share of it without ever finding out.
Same idea, three words
Where each layer sits, from a kitchen to a contract
An English speaker's instinct about formality transfers here, because the two languages built their formal registers the same way.
Placed by where each layer is normally heard, from casual speech to formal writing. The placings are this page's judgement.
- Native everyday wordsThe verbs and adjectives, and the default in conversation
- Borrowings, in casual useModern, light, and sometimes deliberately fashionable
- Borrowings, technicalThe same layer, doing a specialist job
- Chinese derived compoundsNewspapers, official notices, anything written to be filed
What the rail cannot show is that the borrowed layer sits in two places at once, which is why it is drawn twice. A word from that layer can be the casual option or the technical one, and nothing about the script distinguishes them.
Two ways to decide what to learn
A frequency ordered core deck is the efficient starting point, and the ones built from subtitle and newspaper corpora have been the standard for years. It replaced the textbook chapter list, which is ordered by the topic of the lesson and has no relationship to what turns up in a sentence. Use it for the first couple of thousand words and stop: past that, frequency lists get thin and what you need diverges by what you read.
Tatoeba is the other half, free, and it answers the question a dictionary cannot, which is where a word can be said. Search a word and read twenty real sentences with their translations, and the register becomes visible in the company it keeps. It replaced asking somebody, which returns one person's usage and no sense of the range, and it matters more here than in most languages because three near synonyms with different registers is the normal case rather than the exception.
The mistake: learning one of the three and using it everywhere
A learner who has met the Sino-Japanese word for something will use it in a conversation with a friend, which is correct and sounds like a news bulletin. The mirror error is using a casual native word in something written, which is correct and sounds careless. Neither is a grammar mistake and neither will be corrected.
The fix is to store the register with the word, in the same movement as the meaning, and the script makes that nearly free: a card carrying the word in its own script has the register on it already, provided you know what the scripts mean. A learner who romanises their cards has thrown that information away, which is one more reason the romanisation has to go early, as argued in ninety two symbols, and the fortnight it takes.
What the vocabulary connects to
Which layer a word belongs to also decides how its characters are read, since the Chinese readings surface in compounds and the native ones surface with hiragana attached, and that is readings, radicals, and an order that makes them stick. The register of the word interacts with the register of the verb ending, which is a separate axis and is chosen at the end of the sentence, in plain, polite and humble, and choosing between them. The words are held together by particles rather than by word order, in particles, and a verb that waits at the end. And whether a given word earns a card at all is the general question in making cards that teach rather than cards that pile up.