Pitch accent

Japanese is a tonal language, and mastering its tonality is key to pulling off an authentic native-like accent. Unfortunately, the system is complex, and resources for studying it are very limited.

In this article, I plan to share my own understanding based on studying Dogen's phonetics course, referring to the NHK Pitch Accent dictionary appendix and the primary data of the Online Japanese Accent Dictionary, and reading academic reviews such as Kawahara's The Phonology of Japanese Accent.

Nouns

Non-compound nouns

… with three syllables

Most likely, there will be an accent on the first syllable. However, there are some exceptions, particularly in words of native Japanese origin:

… that are loanwords with four morae

Four-mora loanwords where the last two syllables are both free of epenthetic vowels and light are likely to be unaccented. To explain:

• Take the last two syllables of the word.

• Does either syllable include an epenthetic vowel?

  const epentheticVowels = /[くすつぬむるずづぐクスツヌムルズヅグ]/g;
  
  // Each of these words include an epenthetic vowel in their last two syllables:
  // - "idol": ア・イ・ド・ル
  // - "process": プ・ロ・セ・ス
  epentheticVowels.test("ドル"); // true
  epentheticVowels.test("セス"); // true
  
  // Each of these words are free of epenthetic vowels in their last two syllables:
  // - "America": ア・メ・リ・カ
  // - "antenna": ア・ン・テ・ナ
  epentheticVowels.test("リカ"); // false
  epentheticVowels.test("テナ"); // false

  • Yes: The word will generally have an accent somewhere.

  • No: Is either syllable "heavy"?

    const isHeavy = (morae: string[]) => morae.length > 1;
    
    // Each of these words include a heavy syllable in their last two syllables:
    // - "Parade": パ・レー・ド：HL
    // - "London": ロン・ドン：HH
    [["レ", "ー"], ["ド"]].some(isHeavy); // true
    [["ロ", "ン"], ["ド", "ン"]].some(isHeavy); // true
    
    // Each of these words are free of heavy syllables in their last two syllables:
    // - "America": ア・メ・リ・カ： LL
    // - "antenna": ア・ン・テ・ナ： LL
    [["リ"], ["カ"]].some(isHeavy); // false
    [["テ"], ["ナ"]].some(isHeavy); // false

    • Yes: The word will generally have an accent somewhere.
    • No: The word is likely to be unaccented! 🎉

Compound nouns

… formed from two nouns, N1 and N2

… where N2 has one or two morae

TODO: attack Kawahara's Section 4.1 with a strong coffee

… where N2 has three to four morae

If N2 is unaccented, or has accent on the final syllable, then the accent generally falls on the initial syllable of N2.

• si'n+yokohama → sin+yo'kohama "Shin-Yokohama"

Source: The Phonology of Japanese Accent, section 4.2:23

Otherwise, the accent of N2 is retained:

• si'n+tamane'gi → sin+tamane'gi "new onion"

Source: The Phonology of Japanese Accent, section 4.2:24

… where N2 has five or more morae

The compound noun tends to retain the accent of the N2.

Source: The Phonology of Japanese Accent, section 4.2:26.