8 Japanese pronunciation rules every learner should know

Japanese has a small set of sounds compared to English or Korean, which is good news — but the few rules it does have are easy to skip and hard to fix later. This guide covers the 8 rules that, once you actually internalize them, instantly make your Japanese sound natural.

Before the eight rules: Japanese is timed in morae, not syllables

English is stress-timed. Stressed syllables arrive at roughly regular intervals and everything between them is squeezed to fit, which is why "PHO-to-graph" and "pho-TOG-ra-pher" have completely different shapes even though they share a root. Japanese is mora-timed. It is built out of small units called morae (拍, はく) that each take about the same amount of time, and nothing gets squeezed or stretched for emphasis. More than any individual consonant, this is what makes learner Japanese sound foreign: the sounds are right, but the clock is wrong.

A mora is not a syllable, and the difference is easy to count once you know the four counting rules. Every full-size kana is one mora. A small ゃ/ゅ/ょ merges with the kana before it into a single mora (きょ, しゃ, にゅ are one beat each). The small っ is a mora of its own. ん is a mora of its own. And the second half of a long vowel is a mora of its own. Apply those and the famous examples fall out immediately:

Here is the drill that makes this physical. Tap the table with one finger at a steady tempo, about two taps per second, and say one mora per tap — と(tap) う(tap) きょ(tap) う(tap). Do 東京, 学校, 切手 and 新聞. It will feel absurdly slow at first, and that feeling is the point: you are hearing how much time you have been stealing from the second half of every word. Rules 2, 3 and 6 below are really this same rule looked at from three angles — long vowels, っ and ん each own a full beat. If you fix nothing else on this page, fix the clock. You can practise it on any word list; the Basic 1 list is short enough to read aloud in one sitting.

1. Vowels are short, clean, and equal length

Japanese has only 5 vowels: あ (a), い (i), う (u), え (e), お (o). Each one is a single, clean sound — never a diphthong. English speakers tend to lengthen "o" into "ow," and "e" into "ay." Japanese ね sounds like "neh," not "nay." Korean speakers usually do better here, but watch out for stretching え into 에이.

2. Long vowels actually matter

Japanese has long vowels (about twice the duration of a regular vowel), and they distinguish meaning. Some classic minimal pairs:

• おばさん (obasan, "aunt") vs おばあさん (obāsan, "grandmother")
• おじさん (ojisan, "uncle") vs おじいさん (ojīsan, "grandfather")
• ゆき (yuki, "snow") vs ゆうき (yūki, "courage")
• ここ (koko, "here") vs こうこう (kōkō, "high school")

If you treat long vowels as optional, you'll get a lot of confused looks. The fix: when you see two of the same vowel in a row, hold the sound for two beats — count "1, 2" silently.

It is worth seeing how many of these pairs exist, because vowel length is not a matter of accent — it is the difference between two unrelated words. Below, the two columns are not variants of one word; they are separate dictionary entries:

Knowing how long vowels are written removes most of the guesswork. Long あ doubles the kana: おかあさん (mother). Long い doubles too: おいしい (delicious). Long う doubles: ふうとう (封筒, envelope). Long え is normally written え + い — せんせい (teacher), えいが (映画, film) — and long お is normally written お + う: とうきょう, おとうさん (father), がっこう. A smaller closed set writes long お as お + お instead, and those simply have to be learned as spellings: おおきい (大きい, big), とおい (遠い, far), こおり (氷, ice), とお (十, ten). Katakana sidesteps the whole problem with the length bar ー: コーヒー, ビール, スーパー, ケーキ. The bar is a gift — the beat is drawn for you, so katakana words are the easiest place to start getting length right. If katakana itself still slows you down, fixing that first will make every long vowel easier to see.

One trap sits next to this one. A small ゃゅょ and a full-size やゆよ look almost identical but differ by a whole beat: 病院 (びょういん, hospital) is four morae — びょ・う・い・ん — while 美容院 (びよういん, hair salon) is five: び・よ・う・い・ん. In the same way 旅行 (りょこう, travel) is three morae, not four, and 手術 (しゅじゅつ, surgery) is three. Read the small kana as one unit with the consonant before it, and the count comes out right.

3. Double consonants (sokuon っ) need a real pause

The small っ (or ッ in Katakana) doubles the next consonant, and that doubling is a real, audible pause. Compare:

• さか (saka, "slope") vs さっか (sakka, "writer")
• ねこ (neko, "cat") vs ねっこ (nekko, "root")
• かた (kata, "shoulder") vs かった (katta, "won, bought")

Treat the small っ like a full beat of silence. Don't smush it.

More precisely: the small っ is a full mora, and in most cases that mora is silence. Your mouth moves into position for the next consonant, closes, and waits one beat before releasing. This is why "just say the consonant twice, quickly" still sounds wrong — the doubled consonant is not long enough. 切手 is き・っ・て, three beats, with the middle one soundless. Tap it out at the same steady tempo you used for 東京 and you will hear the difference immediately. These pairs are worth saying out loud back to back:

Three details finish the picture. First, before さ/し/す/せ/そ the っ is not silence but a stretched hiss — 雑誌 (ざっし, magazine) and 一緒 (いっしょ, together) hold the s for its own beat rather than stopping. Second, in native Japanese vocabulary っ appears only before k, s, t and p sounds; loanwords broke that restriction, which is why ベッド (bed), バッグ (bag) and ホットドッグ (hot dog) exist with voiced consonants after the pause. Third, っ also turns up at the end of an exclamation — あっ！ or ちょっと待って！ — where it is a sharp glottal stop that cuts the vowel off. Verb conjugation produces っ constantly, so this sound is unavoidable: 買う becomes 買った, 行く becomes 行った, 待つ becomes 待った. If those forms feel unfamiliar, the verb conjugation guide covers where they come from.

4. ふ is not "fu" or "hu" — it's somewhere in between

The Japanese ふ sound is made by blowing air through almost-closed lips, with no real "f" or "h." It's softer than English "fu." A common trick: pretend you're blowing out a candle from a meter away. That airy quality is ふ.

The technical name for it is a bilabial fricative: both lips do the work, and the upper teeth never touch the lower lip — which is precisely what an English "f" requires. Test yourself physically. Rest a fingertip on your lower lip and say the English word "food": you will feel the teeth press down. Now say 富士山 (ふじさん), 冬 (ふゆ, winter), 二人 (ふたり, two people) and 服 (ふく, clothes) keeping the teeth completely out of the way. The result sounds lighter and slightly breathier than English, and that is correct. Note also that ふ is the only kana in the は row with this sound — は, ひ, へ, ほ are ordinary h sounds. Modern loanwords extend the lip-only friction to new combinations with small vowels: ファイル (file), フィルム (film), フェリー (ferry), フォーク (fork). Katakana is full of them, and the katakana loanwords guide collects the patterns.

5. つ is "tsu," not "two"

Many beginners pronounce つ like "two" or "tu," but it should sound like the "ts" in English "cats" + "u." The "ts" is a single sound. Korean speakers often replace this with 츠 — same idea, but be careful not to make it 쯔 or 츠우.

6. The "n" sound (ん) changes shape

The Japanese ん is a single mora (a beat of its own), and its actual sound depends on what follows it:

• Before m, b, p → it sounds like "m": さんぽ → sampo (walk).
• Before t, d, n, r → it sounds like "n": はんとう → hantō (peninsula).
• Before k, g → it sounds like English "ng": にんき → ninki (popularity).
• At the end of a word or before a vowel → a nasalized "n": ほん → hon (book).

You don't need to consciously switch — if you simply hold the nasal sound for one beat and let the next consonant pull it, it usually comes out right.

Three facts separate ん from an English "n" and are worth stating plainly. It is a whole mora: みんな is three beats (み・ん・な), 神田 (かんだ) is three, 全部 (ぜんぶ, all) is three. It never starts a word — no Japanese word begins with ん, which is why it sits alone at the end of the kana chart. And before a vowel or before the さ/は rows it is not a contact consonant at all: the tongue touches nothing and the air simply passes through the nose for one beat. Say 電話 (でんわ, telephone), 本屋 (ほんや, bookshop), 恋愛 (れんあい, romance) and 全員 (ぜんいん, everyone) that way — hold the hum, then continue. English speakers usually stamp a hard "n" down and then rush the rest, which shortens the word by a beat and is one of the most recognizable learner tells.

This also explains a spelling oddity you will meet in older romanization: さんぽ written "sampo" and しんぶん written "shimbun". Those spellings follow the sound rather than the kana. Do not let it confuse you when you read the kana back — it is still ん, still one beat, still the same letter. Drilling the chart itself helps here; the 50-Sound Chart puts ん in its proper place at the very end, outside the main grid, precisely because it behaves unlike every other kana.

7. Devoiced "i" and "u"

Between voiceless consonants (k, s, sh, t, ch, h, p), the vowels い and う often go almost silent. For example:

• すき (suki, "to like") often sounds like "ski."
• です (desu, "is") usually sounds like "des."
• します (shimasu, "to do") often sounds like "shimas."

This is normal native Japanese. If you fully voice every "u," your Japanese sounds robotic — like reading aloud word by word.

The rule in one line: い and う lose their voicing when they sit between two voiceless consonants (k, s, sh, t, ch, ts, h, f, p), and often at the end of a word after one. Crucially, the vowel is not deleted — its beat is still there — but the vocal cords stop vibrating, so what you hear is a whispered puff of air. Here is the effect across common words:

Two cautions, because this rule is easy to over-apply. Devoicing is a strong tendency rather than a law: it is most consistent in Tokyo speech, noticeably weaker in the Kansai region, and it tends to be suppressed when that particular mora carries the word's pitch accent — so you will hear native speakers differ. And you must not remove the beat along with the voice. 靴 is still two morae, 明日 is still three, です is still two. Aim for "whispered", never "deleted"; learners who delete the beat end up with the same timing problem the mora section warned about, just arriving from the opposite direction.

8. Pitch accent — yes, it exists

Japanese is not a tonal language like Mandarin, but it does have pitch accent: words can have a high pitch on a particular syllable, and switching it changes meaning. Classic example:

• 箸 (はし, chopsticks) — high then low. The drop lands inside the word itself.
• 橋 (はし, bridge) — low then high, but the drop lands on whatever follows: 橋が sounds は-し-が with が falling.
• 端 (はし, edge) — low then high with no drop at all: 端が keeps が high.

Notice that 橋 and 端 sound identical in isolation — the difference only surfaces on the next word, which is why it's called the particle test. You don't need to memorize pitch for every word as a beginner; context carries almost all of it. But once you reach intermediate level, taking pitch from native audio rather than from tables makes a real difference in how natural you sound. The "Pitch accent, honestly" section below covers the four patterns and how much to invest.

The Japanese r: a flap, not an R and not an L

ら・り・る・れ・ろ are written with an r in romaji, and that letter has misled generations of learners. The sound is neither the English r nor the English l. It is an alveolar tap: the tip of the tongue makes one extremely quick contact with the ridge just behind your upper front teeth and leaves again immediately. Nothing is held, nothing is rolled, and the lips do absolutely nothing.

Most English speakers already own this sound without knowing it. In North American English, the "tt" in butter, the "dd" in ladder, and the "t" in a lot of said quickly are all the same quick tap. Say "a lotta water" at conversational speed, notice what your tongue tip does, then keep that motion and say ら・り・る・れ・ろ. If that does not click, start from だ・で・ど and make the contact progressively lighter and faster until the consonant stops feeling like a full stop and starts feeling like a flick.

What to avoid is just as useful. The English r bunches the tongue back, rounds the lips, and never touches the roof of the mouth — used in ありがとう it is instantly recognizable as foreign. A full English l is the opposite error: the tongue tip stays pressed against the ridge while air escapes around the sides. Japanese has no l/r distinction at all, which is why English pairs collapse when they are borrowed: ライト is both "light" and "right", グラス is both "glass" and "grass", and only context tells them apart. Korean speakers start with an advantage here — the ㄹ between two vowels is very close to the Japanese tap.

Drill it inside real words rather than in isolation, because the tap is easiest when it is surrounded by vowels: これ (this), それ (that), ありがとう (thank you), 車 (くるま, car), りんご (apple), 面白い (おもしろい, interesting), 疲れた (つかれた, tired), 帰る (かえる, to go home), 冷蔵庫 (れいぞうこ, refrigerator). Then try 料理 (りょうり, cooking), which has two taps in three morae and is a good self-test. Almost every one of these words is in the beginner core, so you can drill sound and vocabulary at once from the Basic Vocabulary module.

Pitch accent, honestly: the four patterns and how much to invest

Rule 8 introduced はし. Here is the fuller picture, because pitch accent is the one topic learners either ignore entirely or overinvest in wildly. In standard Tokyo Japanese, every mora is pronounced either high or low. A word has at most one place where the pitch drops from high to low, and once it has dropped it never rises again inside that word. There is also a rhythmic constraint: the first two morae always differ — if the first is high, the second is low, and vice versa. Those two rules leave exactly four possible shapes for any word.

Notice the awkward pair: flat and tail-high words sound identical on their own. Both start low and rise, and neither falls. The difference only appears on the next word, which is why it is called the particle test. 花が (はなが, "the flower [subject]") drops on が; 鼻が (はなが, "the nose") keeps が high. That single beat of pitch is the entire difference between flower and nose. The same test separates 橋 (bridge, tail-high) from 端 (edge, flat), while 箸 (chopsticks) is obvious from the first mora because it falls immediately.

Now the honest part, because the internet is full of people telling beginners that pitch accent is make-or-break. It is real, but it is not where your first year of minutes pays off best, for three concrete reasons. One: the set of true minimal pairs is small and context resolves nearly all of them — nobody hears 雨が降る and pictures falling candy. Two: accent is regional. Kansai, Kyushu and other areas use systems that differ from Tokyo's, sometimes inverting it, and tens of millions of native speakers get along fine; Tokyo accent is a standard, not a requirement for being understood. Three: accent shifts in compounds and conjugation, so memorizing isolated words has poor transfer. Compare that with vowel length and っ, where an error produces a genuinely different word: saying おばあさん for おばさん is a misunderstanding, whereas the wrong pitch on 橋 is an accent.

So what should you actually do? Take pitch from audio, not from tables. When you learn a new word, imitate the recording's melody as one shape — the way you would copy a two-note tune — instead of assembling it mora by mora. Mark accent explicitly only for words people have actually misheard from you. And know that the information exists when you want it: Japanese dictionaries print accent as a number in brackets, where [0] means flat and [1], [2], [3] mean the drop comes after that mora, so 卵 appears as [2] and 雨 as [1]. Around intermediate level — roughly when you are working through the N3 word list — one focused hour spent listening for the drop in words you already know beats a week of memorizing numbers.

Intonation: how a whole Japanese sentence rises and falls

Pitch accent lives inside a word. Intonation is what happens across the whole sentence, and unlike accent it is quick to learn and immediately noticeable to listeners.

• Statements drift downward. A Japanese sentence starts relatively high and steps down phrase by phrase, each new phrase resetting a little lower than the last. Landing softly on です or ます, without an English-style final flourish, is what makes a sentence sound finished and calm.
• Plain-form questions rise. With no か to mark them, casual questions are signalled by intonation alone. 食べる? with a rise means "are you going to eat?", while the identical word with a fall means "I'll eat." Same with 行く?, これ?, わかる?. Getting this wrong is one of the few pronunciation errors that changes the grammar of what you said.
• Questions with か often barely rise. In polite or business speech, いかがですか and よろしいでしょうか are commonly said with a gentle fall — the か has already marked the question, so a steep rise on top of it can sound pushy or overly casual. A small rise is perfectly normal in everyday conversation; a big one is not.
• ね changes meaning with its shape. Rising ね asks for agreement or confirmation ("right?"). Level or falling ね expresses shared feeling: 暑いですね said flat is a sympathetic "hot, isn't it" rather than a real question.
• よ delivers new information and normally takes a short fall. Stack the two and you get ですよね, offering something you assume the listener also knows.
• Unfinished clauses step up slightly. In 朝ごはんを食べて、コーヒーを飲んで… each て-form phrase ends a touch higher, signalling "still going". Dropping your pitch at a comma tells the listener you have finished when you have not.
• Emphasis is not stress. English emphasizes by making one syllable louder and longer. Japanese widens the pitch range of the whole phrase, slows slightly, or reaches for a particle such as こそ or は. Hammering a single mora louder does not read as emphasis — it reads as strange. If particles are the part you are unsure about, the particles guide covers the grammar side of this.

Errors that mark you as an English speaker — and the fix for each

Almost every learner problem on this page comes from importing English habits wholesale. Here they are side by side, so you can diagnose your own recording in a couple of minutes:

Shadowing: the drill that fixes rhythm fastest

Shadowing means speaking along with a recording about half a second behind it, continuously, without pausing to think. It works better than isolated repetition for one reason: you cannot slow down. The recording sets the clock, and the clock is exactly what mora-timed Japanese demands of you. Fifteen minutes looks like this:

1. Choose 30–60 seconds of audio you already understand — a textbook dialogue, an NHK Easy News item, one anime scene. Comprehension is not the point, so reuse the same clip all week. Novelty is the enemy here.
2. Listen twice with the transcript in front of you and physically mark every long vowel, every っ and every ん. These are the beats learners drop, and marking them turns a vague "sound natural" goal into a checklist.
3. Read aloud with the audio playing, three times. This is parallel reading, and it is much easier than true shadowing because your eyes carry the words.
4. Put the transcript away and shadow four to six times, staying about half a second behind. You will lose your place; keep going and rejoin at the next phrase rather than restarting.
5. Record the final attempt and listen back once, then pick exactly one thing to fix tomorrow. "The っ in 学校" is a usable target. "My accent" is not.

Ten to fifteen minutes of this four times a week produces audible change within a month, far faster than reading about phonetics. It also doubles as vocabulary work, since the words you shadow are the words you keep — pair the clip with the day's list from JLPT N5 vocabulary and push anything that trips you up into your review deck. When a reading is unclear, confirm it on the card before you drill it, because a mispronunciation practised twenty times is much harder to remove than to prevent.

How to practice all of this

• Open Mirai Voca's 50-Sound Chart and tap each kana to hear it. Imitate exactly.
• Record yourself reading 5 simple sentences out loud once a day. Compare to a native sample.
• Shadow short clips: play, pause, repeat. Even 5 minutes a day builds rhythm.
• Don't ignore "boring" rules like long vowels — those are exactly the ones that build a native-like sound.