Kawai, kawaii, and Kaua'i

« previous post | next post »

What's your instinct?  How many syllables do you think there are in the following words?

kawaii かわいい (Japanese for "cute")

Kawai カワイ (the Japanese piano manufacturer)

Kauaʻi (name of one of the Hawaiian islands)

In English, it's our habit to treat diphthongs consisting of two vowels as one syllable, but that's not the way they do it in Japanese, which has no diphthongs.

I asked several of my Japanese Studies colleagues their opinions about the syllabicity of these three words.  Here are three replies:

Linda Chance:

To my mind, kawaii かわいい has four onsetsu 音節 (whether those are syllables or not is another debate).
Kawai カワイ has three.

Kaua'i カウアイ has four, and would not be confused with the first four-onsetsu word. It could be an alternate writing for the second.

Frank Chance:

Kawaii ("cute") has four syllables:  ka wa i and i.

Kawai (the piano manufacturer) has three syllables:  Ka wa and i.

Kaua'i — Not an expert on Hawai’ian but the sense from online is that it is kau-a-i.

Ka-ua-ʻi is wrong
Kau-a-ʻi is correct, but incomplete
There is a spoken feature of the word Kauaʻi that is missing in the depiction above and it is known as a w-glide. W-glides are natural occurrences of the spoken Hawai'ian language. W-glides can follow u vowels. In the case of the word Kauaʻi, a w-glide follows the first syllable and is added to the front end of the second syllable.
Kauaʻi is pronounced KAU-WAH-ee with stress on the capitalized syllables

David Spafford:

Kawaii has four, as far as I can tell. Even n ん counts as a syllable, by the way, at least in music. 

Kawai has three.

My understanding is that the number of syllables is the same, but the Japanese does not have the glottal stop found in Kaua'i.

Part of the problem, at least for me (VHM), is the difference between onsetsu and syllables, also moras, for that matter.  See, for example:

Seiichiro Inaba, "Moras, Syllables, and Feet in Japanese", Language, Information and Computation (PACLIC12), 18-20 Feb, 1998, 106-117.

Han (1994) argues that moras are isochronous units, but no inference is drawn for isochrony regarding the length of the syllable in Japanese. As the tempo of speech increases, however, the phonetic reality of moras seems to become less obvious (Beckman 1982 and Larish 1989). This paper provides a new insight on the apparent gap between phonology and phonetics, which comes from the distinction between Initial Foot Parsing (IFP) and Surface Foot Parsing (SFP). Moreover, it emphasizes an important consideration of timing units larger than moras.



Moras are traditionally considered to be abstract units, but ones that play a crucial role in Japanese phonology (Bloch 1950, Hockett 1955, and Ladefoged 1982). Japanese poetry is often cited as proof that the mora is a psychological unit. Haiku, consisting of three lines of five, seven, and five onsetsu, is one form of Japanese poetry. Onsetsu is often translated as 'syllables' in some of the Japanese literature, but this is a mistranslation. If the syllables are all light, as in (la), the syllable count and the mora count agree as to the onsetsu. But in (1b), the onsetsu of the poetry coincides with the mora count, but not the syllable count.

(1)a. Onsetsu = Mora = Syllable

fu ru i ke ya                    'an old pond'

ka wa zu to bi ko mu      'a frog hopped into'

mi zu no o to                    'the sound of water'


b. Onsetsu = Mora ≠ Syllable

ka ki ku e ba                      'eating persimmon'

ka ne ga na ru na ru           'the bell rings'

hoo ryuu fi                          'at the Horyuji temple'


Selected readings


  1. Frank L Chance said,

    August 1, 2020 @ 3:41 pm

    So what is the point of your question, Victor? In the two Japanese cases that you asked about – kawaii and Kawai – the syllable count and the mor count are identical, four in the first case and three in the second. If you had asked about the syllable counts of 探幽 タンユウ Tan'yū and 他乳 タニュウ tanyū there are in fact differences between the syllables and the mora, but not in the ones you asked about…
    (Tan'yū is the name of a very famous painter, tanyū refers to being fed by a wetnurse.)

  2. Jim Breen said,

    August 1, 2020 @ 3:45 pm

    I avoid using 音節 in a Japanese context, and stick to モーラ (mora/morae).

  3. John Rohsenow said,

    August 1, 2020 @ 4:12 pm

    The right pronunciation of the Hawai'ian of Kaua`i is similar to pronouncing the sounds 'cow-wa-ee' with the 'cow' and the 'wa' flowing together and a glottal stop before pronouncing the 'ee,' which is why the okina* is included. [Some islanders pronounce Kaua`i as 'cow-eye' with an accent on the second syllable.]
    *The 'okina is a glottal stop, similar to the sound between the syllables of "oh-oh." In print, the correct mark for designating an 'okina is the single open quote mark. The kahako is a macron, which lengthens and adds stress to the marked vowel.
    [Hawaiian Language Online – University of Hawaii]

  4. Victor Mair said,

    August 1, 2020 @ 5:04 pm

    The point of my question in the o.p. was directed at those who are not specialists in Japanese, to elicit the difference between English diphthongs and lack of diphthongs in Japanese. That's why I called in the specialists to set the record straight for the Japanese side.

  5. Bruce said,

    August 1, 2020 @ 6:10 pm

    Don't forget current NBA superstar Kawhi Leonard!

  6. PeterL said,

    August 1, 2020 @ 7:46 pm

    If Japanese doesn't have a glottal stop, what is the small "tsu" (っ or ッ) at the end of some words? (Not to be confused with the small "tsu" that's romanized as a double consonant, such in キップ ("kippu" or ticket)).
    E.g. "超かわいいっ!" (https://woman.mynavi.jp/article/161223-65/)
    or "アッー!" (https://dic.pixiv.net/a/%E3%82%A2%E3%83%83%E3%83%BC%21)

    And as for syllable/mora/onsetsu length: what about the "〜" that often shows in speech
    E.g. "は~い!" (https://www.amazon.co.jp/%E3%81%AF-%E3%81%84-%E3%82%A2%E3%83%B3%E3%83%91%E3%83%B3%E3%83%9E%E3%83%B3%E3%81%AE%E3%81%84%E3%81%AA%E3%81%84%E3%81%84%E3%81%AA%E3%81%84%E3%81%B0%E3%81%81-2-%E3%82%84%E3%81%AA%E3%81%9B-%E3%81%9F%E3%81%8B%E3%81%97/dp/4577030721)
    "えぇ〜!?" (https://twitter.com/masuo_mousou)

  7. Noel Hunt said,

    August 1, 2020 @ 8:10 pm

    You will find a definitive history of the terminology in 'An Introduction fo Japanese Phonology', Timothy Vance.

  8. Eric Sadoyama said,

    August 2, 2020 @ 1:46 am

    John Rohsenow: the word Hawaiian is not spelled with a ʻokina. It’s an English word that does not use accent marks, just as Spanish is not spelled Spañish and Irish is not spelled Éirish.

  9. Philip Taylor said,

    August 2, 2020 @ 4:46 am

    Two comments on the immediately preceding comment by Eric Sadoyama :

    1) Why was it addressed to John Rosenhow ? The spelling was used in the original post by Frank Chance, in his embedded reply to Victor's question.

    2) "the word Hawaiian is not spelled with a ʻokina" — vacuously true. But the word Hawaiʻian is spelled with an ʻokina (equally vacuously true), so it is unclear — to me, at least — what point you are seeking to make.

  10. James said,

    August 2, 2020 @ 5:04 am

    Can someone explain what onsetsu means? Victor says: "Part of the problem, at least for me (VHM), is the difference between onsetsu and syllables, also moras, for that matter." However, all the examples written below that seem to have onsetsu equal to mora.

  11. Philip Spaelti said,

    August 2, 2020 @ 8:48 am

    Who ever gave you the idea that Japanese has no diphthongs?

  12. David Marjanović said,

    August 2, 2020 @ 10:04 am

    A very short review of syllables and moras in Japanese (shorter still: both exist, but moras are much more important) is on pp. 21–22 of this paper.

    Inside a big book on the history of Japanese tone phenomena (it's on academia.edu) there's a description of diphthongs in Japanese: vowels that follow another vowel can (but need not) behave like the moraic nasal in that they're skipped for purposes of tone/accent/stress, which means that they form a (long/bimoraic) syllable together with the preceding mora, and thus a diphthong together with the preceding vowel. I don't have time to dig this up now, though.

  13. Philip Taylor said,

    August 2, 2020 @ 10:57 am

    Possibly this one (7,2MB, part one only — I've not bothered to search for part two yet).

  14. ~flow said,

    August 2, 2020 @ 12:43 pm

    @David Marjanović my thinking exactly. Whether or not Japanese has diphthongs would of course hinge on the precise definition of that term. I perceive J. kawaii as having two syllables (ka, wai:) (in connected speech; can also be ka, wa, i:: when emphasized) and four moras (ka, wa, i, i).

    One could certainly argue that in Japanese each vowel receives 'full attention', as it were, i.e. each vowel represents a Schallgipfel and is, therefore, a syllabic nucleus; from this starting point, it is straightforward to build a nice depiction of Japanese phonology when one observes that the universally preferred CV structure is one of the few syllabic blueprints available in that language (so a,k,a can only be syllabified as a-ka, not as *ak, a).

    So far so good. However there are also some phenomena like the tendency to blur the distinction between kai and kae and the devoicing of vowels, most noticeably in desuka ('des-ka'), which seem to be in favor of a somewhat richer picture, one in which kai, kae are monosyllabic (but still bi-moraic) with a preponderance on a, and where desuka has (except in exaggerating styles of speech) evolved to shed its middle syllable (su) in favor of allowing CVSv, that is syllables that sort-of-end in sibilants, but ones which can still be released in moderately careful speech (akin to Eng. 'hat', where in many dialects the release of the t must be audible). The devoiced vowel v must be either u or i and can be recovered when deemed preferable.

  15. Chris Button said,

    August 2, 2020 @ 12:49 pm

    Personally, I've never understood why moras aren't used more in linguistic analysis. To me, an English world like "real" could be analyzed as having one syllable and two moras. I wonder if it goes back to the discussion we had a while back on LLog regarding a preference for consonant/vowel analyses over far more useful obstruent/sonorant analyses.

    I find that moras in Japanese can often be more clearly discernible in singing than standard speech (ties into the comment about poetry above)

  16. Chris Button said,

    August 2, 2020 @ 4:33 pm

    Actually, take a GenAm pronunciation of "near" on its own and in "nearer". I'd propose treating "near" as one syllable and two moras on its own, but "near" as one syllable and one mora when in combination.

  17. Noel Hunt said,

    August 2, 2020 @ 10:25 pm

    How many syllables does 'o wo oou' (おをおおう, 尾を追おう; "[let's] follow [the] tail?") have? How many syllables does 'aioi' (あいおい,相生・相老; "grow up together/grow old together") have? How many dipthongs does each have respectively?

  18. Noel Hunt said,

    August 2, 2020 @ 11:00 pm

    There is confusion over nomenclature, and this has arisen from the translation of 'onsetsu' (音節) as 'syllable'. In early Japanese linguistics, this term was used synonymously with what is now called 'mora'; to use one of Vance's examples, to a Japanese, 'kekkon', 結婚, 'marriage', has 4 'onsetsu', 'ke-Q-ko-N', but from the point of view of syllables, there are only 2.

  19. Chris Button said,

    August 3, 2020 @ 4:55 am

    Now we're getting into Old Chinese. Mora is syllable weight. It has a lot to do with the bifurcation of Old Chinese into two syllable types. Weight on the first part of the syllable or weight on the second part of the syllable. The lack of attention to moras and how that pertains to tense/lax alternations and contrastive length is what has perplexed people studying old Chinese for a while. It's a shame because the cause is pretty obvious in my opinion. Pulleyblank and then Ferlus are the only two academics (as far as I know) to have properly addressed this in any way.

    Unfortunately, we instead have all manner of theories to account for the distinction in Old Chinese. Theodore Stern actually wrote a description of a Tibeto-Burman language, Sizang Chin, along moraic lines. Naturally everyone ignored the moraic, or syllabic weight, analysis. Then Pulleyblank noticed it, applied it to Old Chinese, and naturally everyone ignored him too! Ironically, others chose to follow Starostin's proposal for a length distinction who, in so far as moraic weight surfaces as length, essentially proposed the opposite of Pulleyblank! What really vexes me is the attempt to statistically correlate length distinctions in languages related to Sizang (Staroatin used Lushai/Mizo) as if that should produce any results. It doesn't correlate because how things surface phonetically when you get into the actual phonology is far more complicated than a simple attempt to match things up one on one.

RSS feed for comments on this post

Leave a Comment