One of the first things you learn in Japanese is that the proper pitch accent of ありがとう has the step-down after り (in standard dialect). But if you listen to the pronunciations on Forvo (and more generally, but Forvo will do), the pitch of が tends to be at the same level as り、if not slightly higher, followed by a drop (which is sometimes not so steep compared to the drop on the long vowel). This is confirmed by analyzing the numerous recordings in Praat. In none of them, regardless of region, does り actually seem to take the accent.

Please note that this is not just my ears saying the pitch is different from what it "should" be - I'm attaching a few of the Praat spectrums for ありがとうございます to show what I mean. In all cases, the highest pitch occurs with が. Even if you look at ありがとう by itself the same pattern emerges. Note that some speakers will also use the pitch pattern 低高高高低 as well, but never 低高低低低 as it is "supposed" to be. So my question is, why is the "correct" pitch accent rarely used, or am I just missing something? It seems more likely to me that rather than every accent dictionary in existence being incorrect, I am mistaken in one thing or another, and I would be happy to find out what that is.

enter image description hereenter image description here

  • The graphs remind me of Linguistics 101.
    – Eddie Kal
    Commented Mar 25, 2021 at 17:00
  • Can you link to the specific audio samples the Praat graphs are for? Commented Mar 25, 2021 at 18:16
  • Yes, they are the top few entries under ありがとうございます on forvo: Here
    – user42345
    Commented Mar 25, 2021 at 18:28

1 Answer 1


This will be a somewhat unrigorous answer (since that’s faster to write), but basically there’s a layer of interpretation that your mind does on the pitch, where it’s actually not just the height of the pitch that matters, but the direction of the pitch change throughout the mora that matters.

For example


or even

/ ̄\_

will generally get interpreted as HLLL

Of course a clear

_ ̄__

sounds like LHLL and entirely wrong.

In other words, when there are slides in the pitch, it causes an “fuzzy” accent kernel, which the listener’s brain will then just place in the right spot.

As far as ありがとうございます on forvo goes, they all unambiguously sound accented on the り to me (aside from a kansaiben recording in there). The が might be high in pitch but as long as it’s sliding downwards there’s no way it will be interpreted as the accent kernel.

  • I have to say I disagree with the interpretation of pitch this way (and I unambiguously hear all of the examples on forvo as NOT having pitch nor stress accent on the り), but perhaps that is because I'm a trained musician. It has been pointed out to me that I'm perhaps too sensitive to the actual pitch levels, though, so this answer does provide some clarification. Do you happen to have a more detailed reference regarding the perception of pitch in relation to the direction of pitch change?
    – user42345
    Commented Mar 25, 2021 at 18:33
  • 3
    I’m also a trained musician and I had to learn how to do such interpretation. btw my pronunciation (including pitch accent) is native-level in Japanese. Here’s a reference: conference.wdc-jp.com/psj/2018/contents/common/doc/A4.pdf Commented Mar 25, 2021 at 18:36
  • Appreciate it, and thanks for the reference! It doesn't affect pronunciation per se, since I can listen and imitate it, but it did make me rather confused due to the "contradiction." I will have to read the paper when I have some more time, and will accept the answer if it answers the question sufficiently edit: took a brief glance and yes, it does :). Thank you!
    – user42345
    Commented Mar 25, 2021 at 18:43
  • 1
    Another one: konan-u.ac.jp/hp/kitlab/pub/psj201706_slide.pdf. This phenomenon is known as 遅下がり, but its mechanism is not fully investigated yet, including whether it is a matter of accent or intonation. I personally suspect that the vocal tension plays some role in accent recognition besides pure pitch height. Commented Mar 26, 2021 at 13:23

You must log in to answer this question.