Here's a spectogram of a speaker's rendition of “プロトタイプ” from English “prototype”, collected from Forvo


It should be clear from this spectogram that the pu-mora is pronounced in half a mora; the mora-length in this segment is about 0.16 ms and pu takes about 0.08 ms; the entire tai is also pronounced in about 1.5 moræ.

I've definitely encountered this multiple times in speech, and it wasn't hard to find examples on Forvo at all. I couldn't find any published research on the matter; is this normal? I've heard “ナイス” from English “nice” pronounced with nai as seemingly one mora, and it again wasn't hard to find examples of this on Forvo and confirm it with acoustic analysis.

  • +1 for an interesting question. And that "ae" ligature is interesting. Did you intend to write it like that?
    – rebuuilt
    Aug 5, 2020 at 21:09
  • @rebuuilt, it is an automatism by now, but the interesting thing is that it's actually faster to type than <ae> separately, on my keyboard layout, so that's a nice bonus.
    – Zorf
    Aug 5, 2020 at 21:15
  • "on my keyboard layout" as expected, it's not an English keyboard layout. No English keyboard user would go through the trouble of typing æ. Okay, I digress. Have you checked if other English loanwords or any loanword for that matter exhibits this deviation from a normal moraic rhythm?
    – rebuuilt
    Aug 5, 2020 at 22:45
  • @rebuuilt I believe that the “U.S. International” layout has it mapped to [AltGr+z], which is faster than [ae] in a row. And yes, I have, I cited some other examples in my post. I find it most very common in “ナイス” and things that involve it, and it's certainly not a universal thing. I also often hear it in “シャイ” but that's hard to find a simple case of online without the “〜な” after it to verify that “シャイ” is indeed as long as “〜な”
    – Zorf
    Aug 5, 2020 at 23:35
  • 2
    It's perfectly fine to read プロトタイプ with the same mora lengths, too. Possibly Related: What are the rules regarding “mute vowels” (“u” after “s” and “i” after “sh”)?
    – naruto
    Aug 6, 2020 at 3:28

1 Answer 1


Not an expert, so I cannot cite any research, but I think what you are observing is the vowels not getting pronounced. English does this all the time, and "prototype" is a case in point where "p" is not followed by a vowel, but there are plenty of examples of this in Japanese native words. See 母音の無声化 in Wikipedia that describes some rules along with examples. Consider くつ vs くうき.

As @naruto points out in a comment, What are the rules regarding "mute vowels" ("u" after "s" and "i" after "sh")? covers this in much more depth.

  • That has nothing to do with the moraic rhythm of native Japanese words. In native Japanese words, reduced vowels still take up an entire mora such as that for instance in /doosite/ even though the /i/ is reduced to “formant-less palatalized noise” the entire /si/ mora still takes the same time to pronounce as the /do, o, te/ moræ.
    – Zorf
    Dec 31, 2020 at 17:04
  • @Zorf: A mora is a phonological unit: aren't there still differences in the phonetic length of different parts of Japanese words that have the same moraic length? There are a number of studies that I can find that discuss durational differences between Japanese vowels with the same phonological length, e.g. frontiersin.org/articles/10.3389/fpsyg.2019.00821/full (effects of vowel height and consonant voicedness) and core.ac.uk/download/pdf/213387004.pdf (effects of accent)
    – solute
    Jan 1, 2021 at 1:34
  • As far as I understand the first study though, it is not the mora that is shortened, but the only the vowel in it. I am aware that devoiced moræ in Japanese have been found in some studies to statistically comprise about 95% of the length of voiced moræ, but that's not enough to explain how the pu in プロトタイプ could be 50% of other moræ, especially since it canonically should be a voiced mora, not an unvoiced once since ro starts with a voiced sound.
    – Zorf
    Jan 1, 2021 at 3:17

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .