Japanese does not put spaces between its words. It does not need them, because it changes script instead: the content sits in kanji, the grammar wrapped around it is in hiragana, and anything borrowed from outside is in katakana. Before you have read a word of a sentence you can see its joints, which is the opposite of the reputation the three systems have.
That is worth understanding early, because it turns the writing from three obstacles into one instrument. The scripts are not alternatives and you are not choosing between them. They are three parts of a single surface, each carrying a different kind of information, all in the same line.
Three facts that decide the shape of the work
- three scripts used together. Hiragana for grammar, katakana for foreign words, and kanji for the meaning bearing parts, all in the same sentence. This is not three writing systems to choose between; it is one system with three components and all three are needed.
- the verb comes last, after everything else, and the negative and the tense are on the very end of it. A sentence can reverse its meaning in its final syllable, which makes listening a genuinely different skill from reading.
- politeness is grammar, not tone. The verb form itself changes according to who you are speaking to and about, so there is no neutral way to say anything. This is the feature with no European equivalent and it cannot be postponed.
The script change is doing the work spaces do
Take any ordinary sentence apart and the pattern holds. A run of kanji is a noun or the stem of a verb. The hiragana immediately after it is a particle marking what that noun is doing, or an ending marking tense and politeness. A block of katakana is a word that arrived from somewhere else, usually English, usually reshaped enough that recognising it is a skill of its own.
So the first fortnight of this subject buys more than it looks like it does. Ninety two symbols, and the fortnight it takes is not a preliminary to the real writing system. It is the half that shows you where every boundary in the sentence falls, and a learner who is slow at kana is slow at everything downstream of it.
The unit is the word, not the character
A kanji does not have a pronunciation. It has several, and which one applies is decided by the word it is sitting in: readings that came in from Chinese tend to surface in compounds, native readings tend to surface when the character stands alone with hiragana attached. Common characters carry half a dozen readings and a few carry considerably more.
Which makes the standard beginner exercise, learning a character with a list of its readings, close to useless. The list is not learnable and it is not what you will need; what you need is a few hundred words, each of which happens to contain the character, and the readings then arrive attached to something that selects them. Learn kanji inside vocabulary from the first day, and readings, radicals, and an order that makes them stick becomes a question of sequence rather than of memory. Native words, Chinese readings, and English borrowings is the same split seen from the vocabulary side, and it is why two words meaning nearly the same thing can feel completely different in a sentence.
Most of the sentence was never said
Japanese omits anything the listener can reconstruct. The subject goes first and goes most often; the object follows it; whole clauses are carried by a verb form and a particle. What is left in the air can be two syllables, and the rest was supplied by the previous three turns of the conversation and by the form of the verb itself.
This is why listening is reconstruction rather than decoding, and why a learner who is translating what was said has translated the visible part only. Particles, and a verb that waits at the end is the machinery, and the particles are the part to over invest in, since they are the only marks left saying who did what. Pitch accent, and speech that drops what is obvious adds the other half: Japanese distinguishes words by pitch as well, quietly enough that most courses ignore it and audibly enough that natives notice.
An ordinary exchange
What was said, and everything the sentence left for you to supply
Japanese omits whatever the listener can reconstruct, so listening is reconstruction rather than decoding and the audible part is the smaller half.
- The verbFrequently the only word present
- A particle or twoMarking whatever could not be guessed
- Who is doing itDropped whenever the context supplies it, which is most of the time
- Who it is being done toOften carried by the direction built into the verb
- What you are to each otherCarried by the form of the verb rather than by a word
- What was said three turns agoStill live, and still supplying the missing pieces
A learner who translates what was said has translated the part above the line. The proportion is the whole claim of this drawing.
You cannot finish a sentence without deciding what you are to them
Politeness in Japanese is not a set of polite words layered on top. It is in the verb ending, so the choice is made at the moment the sentence closes, and there is no neutral setting. Plain, polite and humble forms exist for the same verb, and choosing among them is a statement about your relationship to the person in front of you and to the person you are talking about.
Beginners are taught the polite form first, which is right, and then discover that everything they hear in a drama or between friends is in the plain form, which they were told was for later. Plain, polite and humble, and choosing between them is worth reading before that discovery rather than after it.
The certificate has no output section
Here is the stall, and it is structural rather than personal. The JLPT tests language knowledge, reading and listening. It does not test speaking and it does not test writing, at any level. So a person can hold N2, which is the band employers usually ask for, and have never once been assessed on producing a sentence.
Because the test defines the market, the textbooks, the schools and the apps all optimise for recognition, and a learner following the obvious path can spend three years getting extremely good at understanding Japanese and remain unable to hold a conversation. Nothing in the system will tell them, because nothing in the system is measuring it. N5 to N1, and what each level assumes you know is worth reading for what each band contains, and the correction is simple to state and easy to postpone: buy output separately, from the first year, because nobody is going to sell it to you as part of the course.
A free dictionary, a paid queue and a popup
Jisho is the dictionary. Search it in English, in kana, by radical, or by drawing the character you cannot name, and it will hand back the readings, the component parts and the words that contain it. It replaced the printed kanji dictionary, where finding an unknown character meant identifying its radical and then counting the strokes that were left.
WaniKani takes the ordering problem seriously and charges for it: radicals first, then the kanji those radicals build, then the vocabulary that uses those kanji, on a spaced schedule that will not let you run ahead. It replaced the notebook of repeated characters, which is the method most people try first and abandon. The first levels are free and the rest is a subscription, and the honest description is that you are paying for the order rather than for the content.
Yomitan, which continues the Yomichan extension that stopped being maintained in 2023, puts a dictionary inside every Japanese page you read: hover a word and the reading and meaning appear, with one keystroke to send it to a card. It replaced the loop of copy, paste, search and return, which is the loop that stops most beginners reading at all.
Whether the characters are worth doing twice
The question after a first Japanese year is whether the kanji work has bought you anything outside Japanese, and the answer is a real one. No conjugation, no plurals, and a writing system to learn instead shares most of the characters and almost none of the readings, so the shapes transfer and the sounds do not, which is a stranger experience than it sounds. An alphabet designed to be learnt in a morning shares the grammar shape and the borrowed vocabulary while dropping the characters, and is the easier second language of the two by a wide margin. Every tongue this atlas speaks has a version of this question and the method, which is the same whichever language you picked is where the answer stops being about the language.