How common is heiban in Japanese?

Heiban, where the pitch never drops, accounts for 47.7% of Japanese words (38,770 of 81,212 measured). Atamadaka accounts for 14%, and the remaining 38.3% drop somewhere after the first mora.

Can a Japanese word have more than one correct pitch accent?

Yes. 2,663 words, 3.3% of those measured, carry more than one accepted accent. Both are correct, and usage varies by region, generation and register.

Do you need to learn pitch accent for every Japanese word?

The largest single pattern is heiban at 47.7%, where the pitch does not drop at all, so assuming heiban is right for close to half of words. The remaining 31,111 drop somewhere after the first mora and are not predictable from spelling or meaning, so those are learned per word.

Kumi data report

Nearly half of Japanese words have no pitch drop

Across 81,212 Japanese words that carry a recorded pitch accent, 47.7% are heiban: the pitch rises and then simply never comes back down. Pitch accent is usually described as a per-word memorisation tax, and the largest single pattern is the one where nothing happens.

What is heiban, and how common is it?

Heiban (平板) is the pattern where the pitch steps up after the first mora and never drops again, including onto whatever particle follows. It accounts for 38,770 of the 81,212 words measured here, or 47.7%.

Atamadaka (頭高), where the pitch drops immediately after the first mora, is the next largest single pattern at 14% (11,331 words). Everything else, 38.3%, drops somewhere further in.

Where the pitch drops, across 81,212 words

Pattern	Words	Share
No drop (heiban)	38,770	47.7%
After mora 1 (atamadaka)	11,331	14%
After mora 2	5,721	7%
After mora 3	11,747	14.5%
After mora 4	6,458	8%
After mora 5	5,151	6.3%
After mora 6	1,008

Do some words have more than one correct pitch accent?

Yes. 2,663 words, 3.3% of this set, carry more than one accepted accent. Both are correct, and which one a speaker uses varies by region, generation and register.

It is a small share, but it matters for how you treat the topic: a resource that gives every word exactly one pitch is simplifying, not being precise. If you have ever been corrected by two native speakers in opposite directions, this is sometimes why.

Does this mean you can skip pitch accent?

No, and the numbers do not say that. 31,111 words drop somewhere after the first mora, and the pattern is not predictable from the spelling or the meaning. What the distribution does say is that the work is smaller and less uniform than it looks: the largest single class is the one requiring no drop at all, so the useful default is to assume heiban and learn the exceptions as they arrive attached to words you are already studying.

That is the same reason Kumi attaches pitch to vocabulary rather than teaching it as a separate subject. A pattern is a property of a word, and it sticks when it is learned with the word rather than as a rule to apply afterwards.

Method

Scope is every vocabulary entry in Kumi's dictionary carrying a recorded pitch accent: 81,212 words.
Each word is counted once. 2,663 words have more than one accepted accent, and counting each accepted accent separately would weight those words twice, so the primary accent is used.
Only heiban and atamadaka are named. The remaining patterns, odaka and nakadaka, are separated by whether the drop falls on the final mora, which needs a mora count Kumi does not store, so those words are reported by drop position instead of guessed at.
Numbers are computed live from Kumi's dictionary each day. Last generated 2026-08-19.

Pitch-accent data derives from the Kanjium project, used under CC BY-SA 4.0. The distribution and analysis on this page are Kumi's own. See attributions.

Kumi teaches pitch alongside the words that carry it, with audio on every vocabulary entry.