Strange tales and labiovelar transcriptions

« previous post | next post »

East Asians have been addicted to strange stories for millennia.  Many of these fall under the rubric of guài 怪 ("strange"), e.g., zhìguài 志怪 ("records of anomalies"), the name of one of the earliest genres of strange stories in China.

One of the strangest aspects about East Asian strange tales is that perhaps the most famous collection of all was written by a Westerner, Lafcadio Hearn (1850-1904).

As Bob Ramsey relates:

If ever there was a colorful character in Meiji Japan, it was Lafcadio Hearn.

In 1896 Hearn became a legal Japanese citizen (the first Westerner ever to do so), then renamed himself Koizumi Yakumo. He became a Buddhist.

But even before going completely “bamboo” (as old Asia hands would put it), Hearn had led a dizzyingly complex and unusual life. A Greek-Irish writer, he was rootless and impoverished before ending up in America, but once there he found work at a Cincinnati newspaper and started writing lurid accounts of local murders. These stories brought him fame, money—and notoriety. Adding to his scandalous reputation, Hearn married a 20-year-old black woman, a former slave, and was fired for breaking Ohio’s anti-miscegenation law.

Still young and restless, Hearn moved on (after a quick divorce) to a place better suited to his macabre interests, New Orleans, and there he produced a vast number of writings about the city, its Creole population, its cuisine, voodoo, its decay. More books have been written about Lafcadio Hearn than any other New Orleans resident except Louis Armstrong!

But it was in Japan that Lafcadio Hearn produced what we best know him for today. Sent by his newspaper to Japan, he fell deeply in love with that nation’s past and its strange tales of ghosts and demons. By now he had a Japanese wife, and with her help he collected folk literature and traditional tales, adapting them into haunting stories of the supernatural.

Hearn was a captivating writer, and through these ghost stories he became known as a major interpreter of Japan. When his books were translated, they became even more popular in Japan. Their appeal to Japanese readers "lies in the glimpses Hearn offered of an older, more mystical Japan lost during the country’s hectic plunge into Western-style industrialization and nation building. His books are treasured as a trove of legends and folk tales that otherwise might have vanished because no Japanese had bothered to record them."

In 1965, the Japanese director Kobayashi Masaki adapted four Hearn tales into a classic film called Kwaidan (‘ghost stories’). The image shown above is from one of the stories, ‘The Woman of the Snow.’

The Japanese title of Hearn's classic collection is 怪談, which in Modern Standard Japanese is pronounced in kaidan (MSM guàitán), which means, quite simply, "strange tales".  Yet, in transcription, it is usually referred to as kwaidan and rendered into English as "ghost stories" rather than the more accurate "strange tales", which may simply be a matter of translator's preference, but kwaidan (which strikes me as archaic [the rounded medial reminds me of Old and Middle Sinitic]) rather than kaidan, is a matter of phonology, so I set about trying to determine whence the "w" in the former.

Bob Ramsey:

Yes! There were many such archaic spellings back in his day. And you'll notice that Chinese-style spellings such as this one (representing 怪談) were common in the Edo period (1603-1867)–they were written the same way in kana–and apparently, they represented the reality of how people had once pronounced such Sino-Japanese words (but certainly NOT in the late Edo, or the Meij period (1868-1912) when Hearn was writing).

Frank Chance:

Not so much an archaic pronunciation as an archaic transcription into hiragana. 怪談 - くわいだんin the so-called kyuukanazukai (旧仮名遣い) may reflect an earlier pronunciation difference that was already largely lost by the Edo period.   Roy Andrew Miller and other scholars explained it by noting that ancient Japanese had seven vowels, but a couple of them eventually merged into the five vowel sounds of modern Japanese.  Hence the Bodhisattva of Compassion (観音) was written in kana as くわんのん (Ku wan no n), and we find many early transcriptions of that name as Kwannon, though by the 19th century it was pronounced Kannon.  Similarly,  we see many early transcriptions of the Shogunal capital  as Yedo, which was pronounced as Edo by the nineteenth century, because the kana transcription of 江戸 was ゑど and not, as we would do it today, えど. 

This kind of transcription error is not unique to Japanese.   Even in English we have words like “often” where the “ft” did not come from a pronunciation but from the misunderstanding of the “thorn” (Þ, þ) letter used to transcribe the “t” sound.   When the thorn fell out of use among printers, it became “ft” or sometimes “st”.  Of course, there are now topolects of (American) English where that word is pronounced “off ten” reflecting the written form translated back into a pronunciation that does not match the original.

Linda Chance:

It is 旧仮名遣い「くわいだん」

I actually use this example to teach this historical spelling (spelled kwai, now pronounced kai).

Other Meiji romanizations such as "yen" and Yebisu preserve historical spellings as well.

John Whitman:

The 合口 labiovelars were borrowed from MC in Kan’on readings (which traditionally are understood to reflect Chinese Tang period pronunciations, and some Go’on readings, which are supposed to have been filtered through Korea, probably Paekche. Some kana glosses reflect <kwV> as early as the beginning of the 13th century.

For example, the 1229 glosses on the 新譯華嚴經意義 by a monk at 高山寺temple have

化 クワ (呉音)

Some earlier glosses, directly on mss of the 新譯華嚴經, have <kw> on kan’on readings as well:

關  クワん(漢音)1233 glosses

縣  クヱン(漢音)1209 glosses

As these examples show, <kw> was borrowed not just with /a/ nuclei but with other vowels (e.g. <e>) as well. My memory is that some of these survive in the Jesuit transcriptions of the late 16th – early 17the century as well, but in those transcriptions kwa is by far the most numerous of this type. <Kwa> in Sino-Japanese words survives in the Kansai and Tokyo standards until the early 20th century. To this day, for example, the English name of 関西学院大学 in Kobe is Kwansei Gakuin Daigaku. Tellingly, the Japanese abbreviated name is Kangaku. 

The labiovelar pronunciation is associated with Buddhism, and more broadly, old/authoritative readings. It would have been pretty current when Hearn wrote his collection. In the 日本方言地図 collection researched in the 1960s, it was still quite widespread for common SH words like 火事, but mainly in rural areas, especially Kyushu. Hearn settled in Shimane Prefecture, so that might have been a factor too.

As for why the labiovelar SJ pronunciations disappeared, the simple answer is nativization. There were never labiovelars in any stratum of the native lexicon.

However the kw- initial of the transcription of 怪 as kwai, rather than modern kai, arose, today it smacks of antiquarianism, if not archaism.

Selected readings


  1. Jim Breen said,

    March 22, 2022 @ 8:32 pm

    Yes, definitely 旧仮名遣い, which was still common in Hearn's day, even if it had totally died out as a pronunciation. It still lives on in dictionaries such as 広辞苑.

  2. Victor Mair said,

    March 22, 2022 @ 9:07 pm

    From David Lurie:

    Yes, the traditional on'yomi (kan'on 漢音 variety) for that character is クヮイ. I'm not an expert but I believe that the disappearance of the distinction between カイ and クヮイ in on'yomi dates only from the establishment of the modern kana spelling system in 1946.

  3. Victor Mair said,

    March 22, 2022 @ 9:09 pm

    From Jim Unger:

    The short answer is: because the word is Sino-Japanese 怪談, which was Middle Japanese /kwaidan/. Most modern dialects have delabialized /kwa/ to /ka/, but, if I recall correctly, there ae some in Kyushu that retain it.

  4. Chris Button said,

    March 22, 2022 @ 10:45 pm

    Similarly, we see many early transcriptions of the Shogunal capital as Yedo, which was pronounced as Edo by the nineteenth century, because the kana transcription of 江戸 was ゑど and not, as we would do it today, えど.

    I’ve always wondered about this. Given that ゑ is “we”, doesn’t its use as “ye” just reflect the merger with え before they both became modern “e”? In that case, and also with “Yebisu”, isn’t it just an archaic variant spelling of an older “ye” sound rather than representative of any “e” vs “ye” distinction?

  5. astrange said,

    March 22, 2022 @ 11:11 pm

    かいだん has several other meanings in modern Japanese. Perhaps kwaidan stays around because it's not scary if everyone thought you said "staircase" (階段) – they have the same pitch accent.

  6. W. said,

    March 23, 2022 @ 1:36 am

    It seems that some Japanese dialects retained this kw- well into 20th century. Here is a map based on a 1905 study:

    In red areas, the ka-kwa distinction was still present, in green areas it was lost, and in orange areas it was mixed.

  7. GH said,

    March 23, 2022 @ 3:20 am

    Even in English we have words like “often” where the “ft” did not come from a pronunciation but from the misunderstanding of the “thorn” (Þ, þ) letter used to transcribe the “t” sound. When the thorn fell out of use among printers, it became “ft” or sometimes “st”.

    Am I reading this wrong, or is Frank Chance saying that the word was originally oþen ("othen", or apparently here supposedly "otten"), with the "f" introduced by mistake first in print and then in speech? Given all the cognates in other German languages ("oft", "ofte", etc.), that strikes me as an improbable story on its face.

    Wiktionary states that the original pronunciation was /ft/, but that the /t/ was later lost in speech to give the standard /ˈɒfən/ ("offen"), only to reemerge as a spelling-influenced, sometimes deprecated pronunciation. This corresponds with what I have always heard.

  8. VVOV said,

    March 23, 2022 @ 7:08 am

    The 1965 film adaptation is outstanding, easily one of my favorite films. It's a lovely example of the creative power of global cultural exchange: a really unique artwork is created through the iterative process of traditional Japanese tales being compiled/adapted by a Western author (Hearn), and then reclaimed/re-adapted into a film by a Japanese director (Kobayashi).

    Btw considering Dr. Mair's interest in Sinographs, I am sure that he would be captivated by this screenshot of an iconic scene from the film:

  9. David Marjanović said,

    March 23, 2022 @ 8:31 am

    What GH said.

  10. Frank L Chance said,

    March 23, 2022 @ 9:48 am

    I stand corrected.

  11. Chris Button said,

    March 23, 2022 @ 11:09 am

    In addition to labiovelars coming from Sino-Japanese, I find the evidence for stop codas in Middle Japanese to be really interesting.

  12. Chris Button said,

    March 23, 2022 @ 11:22 am

    @ Frank Chance

    Regarding your comment about Roy Andrew Miller, presumably you’re referring to the “i” 1/2, “e” 1/2, and “o” 1/2 mergers to take 8 “vowels” down to 5 (I’m going with Miyake’s 8 here). I’m curious what the connection is that you’re drawing with the discussion here? Or is it just an aside about the evolution of Japanese phonology?

  13. MMcM said,

    March 23, 2022 @ 1:31 pm

    Supposedly it was Percival Lowell's The Soul of the Far East that drove Hearn to Japan.

  14. Frank Clements said,

    March 23, 2022 @ 3:52 pm

    You see the same archaic transcription preserved in the name of the X-Men character Kwannon/Psylocke, an assassin named after the bodhisattva. The character (who is two characters who at one point switched minds and much later switched them back) has a convoluted history with some racial issues that writers have recently tried to correct. Still, it's another example of the old transcription surviving in popular culture, presumably because the creator based the character's name on a source that used the older rendition of the deity's name.

  15. Terpomo said,

    March 23, 2022 @ 6:37 pm

    As regards Unger's remark I was going to say- I thought there was some dialect that still retained it.

  16. Jonathan Smith said,

    March 24, 2022 @ 7:34 am

    Interesting… these were (are?) considered real labiovelar segments [kʷ]? The hiragana representations seem ambiguous… even if just sequences [kɯa], they would be phonotactically exceptional and thus still candidates for "nativization" I suppose…

  17. David Marjanović said,

    March 24, 2022 @ 10:55 am

    I wonder if there's a correlation with those unspecified western Japanese accents (unfortunately I have no idea which ones those are exactly) where u is still [u] and not this unrounded "compressed" [ɯ]-like thing.

  18. Chris Button said,

    March 24, 2022 @ 12:08 pm

    I don’t think the hiragana representations are ambiguous. A Kanwa dictionary will usually include the old labiovelar spellings among other changes. A good dictionary like Todo Akiyasu’s one will also include the -m coda as -mu (I appreciate some people think that isn’t reliable evidence). Evidence for a stopped coda like -t won’t be noted of course.

    It makes sense why people have sometimes viewed Sino-Japanese as a sort of dialect of Chinese.

  19. Philip Anderson said,

    March 24, 2022 @ 1:16 pm

    When printers wanted to replace thorn (for ‘th’), they sometimes used Y, hence phrases such as “Ye Olde Tea Shoppe”.

  20. Jonathan Smith said,

    March 24, 2022 @ 1:19 pm

    ^ The smaller "yōon" are a modern convention; for the potential ambiguity given historical spellings in this case see e.g. Whitman's representations vs. Lurie's representations above.

    Yes I was thinking along the lines of David Marjanović above…

    speaking of which modern standard Japanese [wa] is a bit of a sore thumb in the system no? I feel like tongue is distinctly forward during glide here but lips certainly seem more rounded than is the case for [ɯ]…

  21. Chris Button said,

    March 24, 2022 @ 1:59 pm

    But what's ambiguous about クワ being the one syllable reading for 化 as "kwa" ? Sure, writing クヮ with the smaller "wa" makes it clear that two syllables aren't intended as "kuwa", but I don't think anyone is suggesting that these sino-Japanese pronunciations represent two syllables.

  22. Rodger C said,

    March 24, 2022 @ 3:21 pm

    I've seen an edition of Hokusai's sketchbooks in which the editor (James Michener?) asserted that the correct Japanese title was Hokusai Manga, not the "incorrect" Mangwa sometimes seen.

  23. Dara Connolly said,

    March 24, 2022 @ 4:28 pm

    As an Irish person with Japanese connections, I am very familiar with the life story of Lafcadio Hearn. I've visited his former home and museum in Matsue.

    By an extraordinary coincidence, just minutes before opening this article, I saw a short piece on TV about the Lafcadio Hearn Japanese garden in Waterford, and thought to myself that it was some years since I had heard that name.

    The Wikipedia article on historical kana orthography has a good overview of the pre-reform spellings including kwa, we, etc.

  24. Dara Connolly said,

    March 24, 2022 @ 4:31 pm

    For those interested in an entertaining and irreverent retelling of Japanese "strange stories", see Linfamy's videos on YouTube.
    Example: yuki no onna

  25. Chris Button said,

    March 24, 2022 @ 5:07 pm

    I've seen an edition of Hokusai's sketchbooks in which the editor (James Michener?) asserted that the correct Japanese title was Hokusai Manga, not the "incorrect" Mangwa sometimes seen.

    I wonder why? It's グワ gwa /ᵑɡwa/ (go-on) and クワイ "kwai" /kwaj/ (kan-on) in Todo. Also the -w- is clearly present in the Middle Chinese source.

  26. Jonathan Smith said,

    March 24, 2022 @ 5:26 pm

    My thought could definitely be uninteresting, e.g., if there is (practically) never any such minimal contrast… but if you're claiming that orthographical "クワ" for say 化 in early sources is clearly one "syllable" /kwa/ as opposed to something like /kɯa/~/kua/ because duh, it's from Chinese, than no, that's not right.

  27. Philip Taylor said,

    March 24, 2022 @ 6:26 pm

    Dara — yuki no onna, 02:56: "Bish ain't getting my viewers". "Bish" ? I clearly heard "Bitch". Is this Youtube trying to be uber PC, or did it simply guess a non-existent word ?

  28. Chris Button said,

    March 24, 2022 @ 7:10 pm

    @ Jonathan Smith

    You would have to treat きや “kya” (modern きゃ) in the same way. You would also disrupt the notion of long and short syllables, which ky- and kw- did not affect. Also would you really propose to treat something like くわい “kwai” for 画 at face value as as “ku.wa.i”?

    And of all the things to object to terms of phonotactics, wouldn’t the evidence for a final -t coda be more objectionable?

  29. Jonathan Smith said,

    March 24, 2022 @ 8:21 pm

    kya ≠ ki.ya, etc., was exactly my point, thinking about whether or not a similar contrast could have (does?) exist around a glide /-w-/… the question may be (is probably?) academic.

    SJ has affected Japanese phonotactics in many ways; the /-j-/ "glide" is one obvious one and so my question again is what might have happened re: /-w-/ such that it remained regional / disappeared.

    Yes, given historical orthographical "くわい" of course you try to figure out whether or not this = ku.wa.i in phonological esp. prosodic terms — whether/when/to what degree Japanese has or had /-w-/ or "labiovelars" proper is to be demonstrated, not assumed.

    What would be great is if a speaker of one of the purported conservative varieties that preserves a "-w-" in such SJ items would chime in on the question of likeness to loan word forms like グアテマラ, etc. :D

  30. Terpomo said,

    March 24, 2022 @ 9:56 pm

    Phillip, that's not YouTube's auto-captions, it's the uploader's intentional mistranscription, probably to avoid the video being demonetized or age-restricted.

  31. Chris Button said,

    March 24, 2022 @ 10:07 pm

    @ Jonathan Smith

    Presumably it was a result of the broad distribution of Cy- as opposed to the highly restricted distribution of kw-

  32. Rodger C said,

    March 25, 2022 @ 9:59 am

    Chris Button: I imagine Michener's knowledge of Japanese wasn't historical.

  33. Jonathan Smith said,

    March 26, 2022 @ 5:51 pm

    No update? :(
    I did find an online resource featuring a staggering amount of info on Kyoto dialect(s) at
    by a Mr/Ms 藤月 Fujitsuki; this page
    discusses the issue in detail and does state, at least, that the sounds at issue are monomoraic as suggested by the miniscule "wa" representations (「ク-ワ」「グ-ワ」と2拍で発音するのではなく、[kwa] [gwa] と1拍で発音されます。")
    The author is uncertain whether contrasts kwa vs. ka survive in any modern dialects… a shame as I was really hoping for audio…

  34. Chris Button said,

    March 27, 2022 @ 7:12 am

    @ Jonathan Smith

    Basically “kwa” just persisted longer. Things like “kwi” etc used to exist but disappeared earlier. Presumably that’s connected with the loss of w- onsets before certain vowels—hence no ゐ “wi” or ゑ “we” anymore (see my earlier post).

  35. Chris Button said,

    March 27, 2022 @ 7:15 am

    It might be worth adding that I think diphthongs are reconstructed for proto-Japonic. There you get -ua- giving “o”, etc.

  36. Jonathan Smith said,

    March 27, 2022 @ 11:40 am

    …my interest was in changes to Japanese phonology occasioned by the influx of Chinese loans… my understanding (right/wrong?) is that /kwa/, etc., were exclusively SJ (and exclusively associated with the so-called "Go-on" stratum if that matters.) The point is that /Kwa/ (and many other Chinese syllable types) were not to begin with commensurate with Japanese phonotactics and would have been nativized in some way; incorporation of such syllables per se (which apparently did happen with /Kwa/) would only have been possible under a strong and sustained influence. Note monomoraic /kwa/ has apparently not been reintroduced in modern Japanese even given numerous English, etc., loan words with this shape; see above. Thus the need for careful consideration in interpreting earlier spellings…

  37. Chris Button said,

    March 27, 2022 @ 5:03 pm

    @ Jonathan Smith

    I’m no expert in Old Japanese. I will just add three more comments to what is have added above.

    1. “kwa” occurs in go-on and kan-on.
    2. If you think “kwa” is more than one syllable, then I think you are disrupting old Japanese poetic rhythm
    3. Miyake notes that 悔い is transcribed with two phonograms in the 8th century rather than one “kwi”

  38. Moon Moth said,

    March 29, 2022 @ 1:25 pm

    Ah, this explains why James Clavell's "Shogun" uses "Kwanto" and "Yedo" instead of "Kanto" and "Edo", being set around the year 1600. The last time I read it was before the Internet became easily searchable. I eventually guessed that it was just something the author did to indicate that he was fictionalizing them, the way "Tokugawa Ieyasu" became "Toranaga Yoshi". But it now seems like it was an accurate representation of the period Romanizations, and likely of the pronunciations, too (give or take regional/status variations). Thanks!

  39. Chris Davis said,

    April 5, 2022 @ 10:49 am

    [kwa] is found throughout Ryukyuan, and also in Japanese as spoken by the speakers of Ryukyuan. My consultants of Yaeyaman (Southern Ryukyuan) regularly say things like “roojinkwai” for 老人会 (standard Japanese “kai”). These same speakers pronounce words like 英語 (eigo, “English”) as “yeigo”.

RSS feed for comments on this post