Pitch, not stress
In English, the word record changes meaning depending on which syllable you stress: REcord versus reCORD. Japanese has no stress. Every mora is the same length and roughly the same loudness. What changes is pitch: each mora is either high or low, and the pattern of highs and lows belongs to the word.
The key event in a Japanese word is the drop, the point where pitch falls from high to low. Words are classified by where that drop happens, or whether it happens at all.
The famous pairs
These pairs sound identical in romaji and are distinguished by pitch alone. They are worth learning because they make the idea concrete.
| Word | Pitch pattern | Meaning |
|---|---|---|
| 雨 (あめ) | A↘me: high then low | rain |
| 飴 (あめ) | a↗me: low then high, no drop | candy |
| 箸 (はし) | HA↘shi: high then low | chopsticks |
| 橋 (はし) | ha↗SHI↘: low then high, drops after the word | bridge |
| 端 (はし) | ha↗shi: low then high, no drop | edge |
| 神 (かみ) | KA↘mi | god |
| 紙 (かみ) | ka↗MI↘ | paper |
| 髪 (かみ) | ka↗mi | hair |
The four Tokyo patterns
- Heiban (平板), flat: starts low, rises on the second mora, and never drops, even on the following particle. Examples: sakura (桜), ame meaning candy, kami meaning hair.
- Atamadaka (頭高), head-high: starts high on the first mora and drops immediately. Examples: ame meaning rain, hashi meaning chopsticks, inochi (命), and many katakana loanwords such as terebi (テレビ).
- Nakadaka (中高), middle-high: starts low, rises, and drops somewhere inside the word. Examples: kokoro (心) with the drop after ko-KO, tamago (卵).
- Odaka (尾高), tail-high: starts low, rises, stays high through the last mora, and drops on the following particle. Examples: hashi meaning bridge, otoko (男), hana meaning flower.
Heiban and odaka sound identical in isolation. The difference appears only when a particle follows: sakura ga stays high through ga, while hana ga drops on ga. This is why you should always learn nouns with a particle attached.
How much it matters for beginners
Honestly, less than mora timing and vowel length. A learner who holds long vowels and doubled consonants correctly but has flat pitch is easy to understand. A learner with perfect pitch and short vowels is not. So the priority order is: mora timing first, vowel length and small tsu second, pitch accent third.
That said, pitch is what separates understandable from natural, and it is far easier to train early than to fix later. Bad pitch habits on common words such as arigatō, sumimasen, and watashi get repeated thousands of times and become stubborn. Getting the fifty most common words right is a small investment with a large return.
Practical steps that work
- Learn the four patterns until you can name them. That takes an afternoon.
- Listen for the drop. When you hear a new word, ask: where does it fall? Hum the word without consonants to isolate the pitch.
- Mark new words with their pattern in your flashcards, and learn nouns with ga or wa attached so heiban and odaka stay distinct.
- Shadow native audio daily, copying pitch rather than thinking about it. Ten minutes on a short passage is enough.
- Record yourself and compare, or use pronunciation scoring, because your own ear cannot judge your pitch reliably at first.
- Fix the top fifty words first: greetings, pronouns, numbers, and the words you use in every conversation.
Common patterns for beginners to notice
Some patterns are frequent enough to internalise as defaults. Many i-adjectives drop on the mora before i: taKAi becomes ta-KA-i with the drop before the final i, and the same goes for atsui, samui, and ōkii. Verbs in dictionary form are usually either heiban, such as kiku (listen) and tsukau (use), or drop on the second-to-last mora, such as taBEru and kaEru (return). Recent loanwords in katakana very often take atamadaka, like TE-rebi and KO-ohii. These are tendencies, not laws, but they give you a sensible first guess.
Getting feedback on your pitch
The hard part of pitch accent is not understanding it but hearing your own mistakes. Unihongo's pronunciation lab gives you native model lines, records your attempt, and scores the result so you can see which words drifted. Combine that with regular spoken conversations, where your AI Sensei speaks in natural Tokyo-standard Japanese and you copy it constantly without noticing. Over a few months, the patterns move from something you know to something you do.