Does pitch accent matter, and when should you learn it?

Pitch accent matters, but less than the minimal pairs used to advertise it suggest, and in a different place than they suggest: not in whether you are understood, but in how easily you hear Japanese and how much relearning you sign yourself up for later. 箸 and 橋 are settled by the sentence around them long before pitch has to settle them, and the same is true of nearly every pair anyone quotes at you.

The timing answer follows from that. Learn it one word at a time, starting with your first words, and never as a course you take separately from vocabulary — because the cost of storing a word's pitch at first contact is close to zero, and the cost of adding it to two thousand words you already say flat is a project.

Japanese pitch is a fall, and a word has at most one

The first thing to unlearn is the word accent. English accent is loudness and length — RE-cord against re-CORD. Japanese accent is a drop in pitch, and everything else about the word's melody follows from where that drop is.

Every mora is either high or low. A mora is a beat, not a syllable: きって is three beats, とうきょう is four, and the small っ and the ん each get one of their own. Two rules then govern standard Tokyo Japanese, and between them they do almost all the work:

1. The first two morae are always different in pitch. If the first is low, the second is high, and the other way round.
2. Once the pitch has fallen inside a word, it never rises again.

Which means a word's entire pattern is fixed by one piece of information — where the fall is, or that there isn't one. That is why dictionaries print a single number rather than a shape. 0 means no fall anywhere. 1 means the fall comes after the first mora, 2 after the second.

はし is the standard demonstration, and it is worth reading with the particle attached, because that is where the third member hides:

word	notation	with が
箸 chopsticks	はし [1]	箸が — high, then low, staying low
橋 bridge	はし [2]	橋が — low, high, and the fall lands on が
端 edge	はし [0]	端が — low, high, and it stays high

橋 and 端 sound identical on their own. The difference exists, but it is not in the word; it is in what happens to the next thing you say.

Four patterns, and most words sit in one of them

The traditional labels name where the fall is. 頭高 (atamadaka) drops after the first mora. 中高 (nakadaka) drops somewhere in the middle. 尾高 (odaka) drops after the last mora, so it only shows itself on a following particle, like 橋 above. 平板 (heiban) never drops at all.

平板 is the largest class, and this is the practically useful part: a flat word stays high through whatever follows it. 鼻 and 花 are both はな, both low-high, and they part company only when something comes after — 鼻が stays up, 花が falls. If you must guess a word's pattern with nothing to go on, flat is the best bet you have.

That is four categories, one number per word, and a rule that the fall never reverses. The system is not large. What is large is the vocabulary you have to attach it to, which is the whole reason the timing question has an answer at all.

The famous minimal pairs are famous because they are rare

箸 and 橋. 雨 and 飴. The same handful appears in every argument for pitch accent, and their reappearance is the evidence: if there were many, nobody would need to keep using these.

More to the point, none of them are ambiguous in use. Nobody at a table hears bridge when you ask for chopsticks. 雨 arrives with 降る and a window; 飴 arrives with a bag and a child. Japanese resolves homophones by context constantly, because it has an enormous number of them — and plenty that pitch does not separate at all.

The proof of that is what native speakers do about it. 科学 and 化学 are both かがく with the same pattern, so when it matters, speakers say ばけがく for the chemistry one, reading the character with its native reading instead. 私立 and 市立 are both しりつ, so they become わたくしりつ and いちりつ in speech. Those workarounds exist because pitch could not do the job. If pitch were the load-bearing distinction people describe, nobody would have needed to invent them.

Vowel length breaks sentences in a way pitch almost never does

If you have limited attention for pronunciation — and everyone does — the mora clock deserves it before pitch does.

おばさん and おばあさん differ by one beat, and they mean aunt and grandmother. ビル is a building and ビール is a beer. きて, きって and きいて are three different requests. 病院 and 美容院 are the hospital and the hair salon, four beats against five.

Get the length wrong and you have said a different word, in a way context often cannot repair, because both candidates fit the sentence. Get the pitch wrong and you have almost always said the right word with an accent. That asymmetry should decide what you work on first, and it usually gets the ordering exactly backwards in online arguments about pitch.

Pitch earns its keep in your ears before your mouth

Here is the case for pitch that rarely gets made, and it is the one that survives scrutiny.

Japanese speech does not put spaces between words, and it does not lean on stress to mark where they start. What it has instead is the pattern above: a phrase begins low and steps up, and once it falls it stays down until the next phrase begins. That rise is a boundary cue. Your ear can use it to cut the stream into pieces.

Learners who have never attended to pitch tend to describe fast Japanese as a wall — not too fast to identify sounds, but impossible to divide into words. Part of what is missing is that cue. You do not need to produce pitch correctly to benefit from this; you need to have noticed that it is there, and to have stored roughly the right shape with each word so that what arrives matches what you hold.

That is a listening skill, and it is acquired the way listening skills are: by hearing many words many times, with attention, not by reciting pattern tables.

A word does not keep its dictionary pitch inside a sentence

This is why isolated drilling disappoints people who commit to it.

Pitch is realised across a phrase, not a word. Words group with their particles and with what follows, and the group takes one fall between them — so a [2] word buried in the middle of a phrase may show no fall at all where you expected one. On top of that, the whole utterance drifts downward as it goes, a question pulls the final mora up regardless of what the word's own pattern says, and fast casual speech flattens plenty of it.

So the citation form you drilled — one word, alone, carefully — is a form you will rarely hear. It is not wrong, and it is what dictionaries record for good reason. But it is a starting position rather than the thing itself, which is another argument for meeting patterns inside real sentences rather than on their own.

Verbs and adjectives are a two-way choice, not a table

The prospect of a pattern per word is what makes people give up before starting. For verbs and adjectives, it isn't one.

They fall into exactly two classes: accented and unaccented. 話す is flat. 食べる has a fall. Once you know which class a verb is in, its conjugated forms follow regular rules rather than needing individual storage — and one of those rules is a gift, because the polite 〜ます form takes its fall on ま for every verb in the language, whichever class it came from. 話します and 食べます behave identically.

One bit of information per verb, then, plus a small set of rules. That is a different proposition from a table of thousands, and it is worth knowing before you decide the whole subject is too expensive.

Not every Japanese speaker has the pattern you would be drilling

The Tokyo system is not Japanese pitch accent. It is one of several.

Kansai runs a different system with its own history, and many words carry a shape there that does not line up with Tokyo's at all. It is not a corruption of the standard; if anything it preserves distinctions Tokyo lost.

More striking for the does it matter question: several regions have no lexical pitch distinctions whatsoever. Across parts of northern Kyushu, and a band running through Tochigi, Ibaraki and southern Tōhoku, 箸 and 橋 are pronounced identically, and speakers there communicate with everyone else in Japan without difficulty.

You can still choose Tokyo as your target — it is what dictionaries, national broadcasting and most teaching materials record, and having one consistent target beats having none. Just hold it as a convention rather than as the definition of correct, and stop worrying that a wrong pattern is a failure of comprehensibility. It is an accent, and everyone has one.

Which is why pitch belongs on the card, not in a syllabus

Everything above points at the same practical conclusion. Pitch is a property of the individual word — like which reading a kanji takes in that particular word, or which counter a noun is counted with, it is not derivable from the parts and it has to be stored with the whole. So store it when the word arrives.

The cost asymmetry is the entire argument. When you meet a word for the first time, you are already storing how it sounds; taking the pattern with it adds nothing to the work. If you skip it for two years, you do not simply have a gap to fill later — you have a motor habit that fires before you can intervene, and unlearning it takes longer than learning it would have. This is the same trap as recognising a word without ever testing whether you can produce it: the deficit is invisible until the moment it costs you.

In practice that means hearing each word rather than only reading it. A card with audio on it does this passively, provided you actually say the word back during the review instead of reading the meaning and moving on. Kikusho puts native audio on every card alongside the word, its reading, romaji, meaning, a register note and an example sentence you can play, and all of it lives on the device, so the audio is there on the train with no signal. The point is not the feature; it is that a Japanese card without sound on it has quietly decided that pitch is somebody else's problem.

So: does it matter? Less than the people drilling patterns will tell you, more than the people ignoring it entirely will admit — and mostly on the listening side, where it decides how much of a sentence you can even divide into words. The mistake is not choosing wrong about whether pitch is important. It is treating it as a subject to study later rather than a field on a card you are making today.