The Case of the Japanese Verb

Just a quick one for today: Description and Explanation in Inflectional Morphophonology: The Case of the Japanese Verb, by Waseda professor Brente de Chene. I quote Professor de Chene's homepage:

Attempting to go beyond the question of what the descriptively adequate (psychologically real) analysis of the system [of Japanese verbal inflection] is to the question of the explanatory principles on the basis of which that analysis has been chosen over observationally adequate alternatives has led me to a reconsideration of the principles of base form (underlying representation) choice in phonology more generally.

Y'see, there's historically been debate over the underlying principles that drive the Japanese verb system. Specifically, when we see patterns like this (cut-down version of table from paper):

Inflection typeC-stem versionV-stem version
(root)mat- ("wait")mi- ("see")

... One obvious hypothesis is that each inflection comes in a "long version" and a "short version":

VersionExamplesBegins withAttaches to
Longru, rare, saseConsonantV-stem verbs
Shortu, are, aseVowelC-stem verbs

... But the question then is, which is basic? Is the short version the basic form, with the initial consonant of the long version added simply to avoid hiatus? Or is the long version the basic form, with the short version created by deleting the consonant, or allowing it to assimilate (leaving no trace) to the consonant at the verb stem? Or is the "basic" form something else altogether, or different for each type of inflection, etc.?

Each point of view has had its champions in the past. What I like about de Chene's paper is that it offers a new approach to the problem that is based on and tested against actual data — the results of nationwide dialectical surveys — rather than the dreaded "armchair theorizing". It's long and quite technical, but very rewarding reading.



Dig, if you will, Nagai Kafū's 1913 translation of Paul Verlaine's piano poem:

Shinayaka naru te ni fururu piano
oboro ni somaru usubarairo no yūbe ni kagayaku.
Kasuka naru tsubasa no hibiki chikara naku shite kokoroyoki
sutareshi uta no hitofushi wa
tayutaitsutsu mo osoru osoru
utsukushiki hito no utsurika komeshi keshō no ma ni samayofu.

Aa yuruyaka ni wagami o yusuru nemuri no uta,
kono yasasiki uta no fushi, nani o ka ware ni omoe to ya.
Hitofushi goto ni kurikaesu kikoenu hodo no REFRAIN wa
nani o ka ware ni motomuru yo.
Kikan to sureba kiku ma mo naku sono utagoe wa koniwa no kata ni kiete yuku,
mosome ni akeshi mado no suki yori.

一節毎に繰返す聞えぬ程の REFRAIN は

One immediate point of interest is the translation of baise ("kiss") as fururu ("touch"). This isn't a watering-down or a cop-out; it's actually a marvelous choice. None of the Japanese words that meant "kiss" unambiguously in 1913 would have worked: kuchizuke and seppun are, quite apart from their status as unappealing translationese, too specifically about mouths to carry the metaphor here. Kisu had been loaned from English and was a little less grossly physical, but I don't think it had become quite Japanese enough yet to work as a translation of a French word.

Fureru, on the other hand, has been in Japanese since people started writing the language down, and it has been usable in a semi-metaphorical, romantic way for just as long. Man'yōshū 2320:

wa ga swode ni/ purituru yuki mo/ nagare yukite/ imo no tamoto ni/ iyuki purenu ka
The snow fallen on my sleeve: might it not flow away, to reach and touch the hand of my beloved?

The next point of interest, for me, is the treatment of Verlaine's capitalized Elle, "Her". Kanojo had already been coined (under the influence of Indo-European gendered pronouns), but I'm not sure if it was considered unmarked enough for the kind of poetry Kafū is writing here. On the other hand, the term he did use, utsukushiki hito ("beautiful person"), doesn't seem to work very well either. Not only does it abandon the subtlety of the original, it doesn't even seem to make sense: how can we tell from fragrance alone whether the perfume came from someone beautiful or not?

The final stanza can't, and doesn't attempt to, reproduce the effect of the French with all those Qs. The appearance of REFRAIN in there is a bit of a shock (the furigana indicate that it is to be pronounced refuren); I suppose that Kafū couldn't think of a good equivalent, and I can't either. Note that he throws in hitofushi goto ni kurikaesu ("repeating once each verse") to explain what a refrain actually is.

That monster of a last line is mostly weighed down by kikan to sureba kiku ma mo naku ("if one tried to listen to it, it would [vanish] before one could"), which seems to have all come out of tantôt in the original. Finally, we note with interest that while in the original it is the Chant that wants something of the narrator and the refrain that escapes from the window to die in the yard, Kafū reverses the order so that our final image is that of a song — a voice — escaping. We start with graceful hands caressing a piano, survey the room haunted by fragrance and memory, and close with a faint voice slipping away to die in peace.


We'll Shield

Don't worry, this post is about contemporary Japanese! Specifically, the We'll Shield line of products from Taiko Pharmaceutical (大幸楽品).

The brand name is written "WE'LL SHIELD", in Roman characters, and the katakana reading given is ウィルシールド, wirushīrudo. This is a portmanteau word, combining:

  1. English "We'll", rendered in katakana as ウィル, wiru. (You also get English "will" for free, since its katakanafication is identical.)
  2. The Japanese word for "virus", ウイルス, uirusu.
  3. English "shield", rendered in katakana as シールド, shīrudo.

Thus, although the name is composed entirely of loanwords, it is also completely dependent on the phonotactics of Japanese to make sense.

One point of (extremely) mild interest is that the overlap of wiru and uirusu suggests that the difference between ui ウイ and wi ウィ is perceived as quantitative rather than qualitative. Which makes sense: in the context of European borrowings the latter is really just just a hipper version of the former, available only to those with sufficiently cosmopolitan idiolects. And those to whom it is not available would, of course, use ui instead.

Question: The history of the word uirusu is a bit of a mystery to me. Most lexicographers agree that the word for "virus" was originally borrowed from German, and early citations tend to have vīrusu or bīrusu, exactly as you would expect for the German word Virus, but I've yet to find an explanation of where the Latinate form uirusu came from or why it became dominant. Does anyone know?


More from the VLI

As promised, here are some things that we can learn from the Vocabvlario da Lingoa de Iapam even if we don't really speak Portuguese.

Let me note before I begin that this is not intended to be a comprehensive survey of 17th-century Japanese, or even of all the implications of the issues raised in this post. Although I provide some context and supporting information from other sources, what I want to do here is to show what the orthography of the VLI shows, or at least suggests, in and of itself. It's all an excuse to type up some early 17th-century Portuguese as faithfully as possible, is what I'm saying.

1. Word-initial <f>

This is probably the most obvious point of interest in the VLI. First, here's some background:

KanaPre-reform kanaNJMeaningOJMan'yōgana (example)
はなはなhanaflowerpana波奈 (MYS 0816)
ひとひとhitopersonpito比等 (MYS 0808)
あわあはawamilletapa安波 (MYS 3451)
かいかひkaishell(fish)kapi可比 (MYS 3709)

(NJ = "Contemporary Japanese", the Japanese of today as opposed to the "Modern Japanese" that stretches back for centuries. OJ = "Old Japanese" as usual.)

Comparing the NJ to the OJ forms, we see that:

  • Intervocalic /p/ became /w/ before /a/, and disappeared before all other vowels.
  • Word-initial /p/ became /h/.

So what do those words look like in the VLI?

  • Fana. Nariz. (78)
  • Fito. Homem, ou molher. (96)
  • Aua. Painço, ou milho, húa das cincº ſementes, ou legues chamados. (15. This actually appears capitalized as AVA because it is the first entry in its section; I took the liberty of normalizing to lower-case.)
  • Cai. Ameijoa, ou outro mariſco do mar ſeme lhante. (33)

OJ intervocalic /p/ has already vanished, except where it has become /w/ before /a/. This is exactly the same as the contemporary Japanese we speak today. But word-initial /p/ hasn't become /h/. It's written <f>, and there are arguments over how it is to be pronounced. The simplest explanation is that it is an unvoiced bilabial fricative (i.e. a "weakening" of the original unvoiced bilabial plosive), but there is contemporary evidence against this — for example, Diego Collado's 1632 Ars Grammaticae Iaponicae Lingvae notoriously says:

Litera, f, in aliquibus Iaponiæ prouincijs pronunciatur sicut in lingua Latina; in alijs autem ac si esset, h, non perfectum: sed quodam medium inter, f, &, h, os & labia plicando, & claudendo, sed non integrum, quod vsu facilè compertum erit: v.g. fito.

The letter f is pronounced in various regions of Japan as it is in Latin. In others it is pronounced as if it were an imperfect h. For both pronunciations the lips and the mouth should be nearly, but not completely, closed. (Richard L. Spear's translation)

Since Latin <f> is supposed to have been labiodental, not bilabial (here I would insert a reference to Allen's Vox Latina if I had my copy handy), many interpret Collado's "sicut in lingua Latina" as evidence that <f> may not in fact have been unambiguously bilabial. Or, of course, Collado may have been misunderstanding something, writing as he was in that benighted age before even the first edition of Ladefoged's A Course in Phonetics had been published.

Whatever <f> was, it must have evolved into NJ /h/ after this dictionary was written. The whole story is rather more complicated and obviously can't be deduced from a single 1603 dictionary. Moving on!

2. Y before E

Some more background:

KanaPre-reform kanaNJMeaningOJMan'yōgana (example)
まえまへmaefrontmapye麻敝 (MYS 4129)
えだえだedabranchyeda延太 (MYS 3603)
えてえてeteget(ting)ete愛弖 (MYS 806)
えみゑみemismilewemi恵美 (MYS 4106)

And then look how they appear in the VLI:

  • Maye. Diante, ou em preſença, ou dantes. (154)
  • Yeda. Ramo. (320)
  • Yete. He como aduerbio, modo de fazer algũa couſa bem, ou por cuſtume, &c. (322. This also appears as <Ye, yuru, yeta> on 319, but I wanted a form that matched the MYS citation; and yes, <yuru> instead of <vru> is odd; see Doi et al 814-815 for more on this.)
  • Yemi. Alegria, ou ſorriſo. (320)

In fact, as far as I can tell there are no instances of a mora consisting solely of /e/, with no initial consonant, in the dictionary. There are no words written starting with <e>, for example.

One interpretation of this would be that when the OJ distinction between /ye/ and /e/ was lost, the survivor was /ye/, and this didn't become /e/ until very recently — recently enough for the beer brand "Yebisu" (pronounced /ebisu/) to retain that initial <y> in its official English spelling, for example.

Frellesvig, however, rejects this interpretation. He argues that the /ye/ - /e/ distinction collapsed to /e/, and that what we see in the VLI is an additional rule that inserted an "on-glide" whenever /e/ appeared in a mora with no initial consonant. This sounds a bit fiddlesome, but it is in fact more parsimonious in many ways than the hypothesis of long-lived /ye/. For example, Frellesvig's explanation frees us from the need to devise a spaghetti of historical rules explaining how /pye/, /ye/, /e/, and /we/ all ended up as something spelled <ye>; instead, we can simply hypothesize that they all eroded down to phonemic /e/ which was then expressed with an on-glide (Frellesvig 208-210).

3. W before O

Similarly, there are no words starting with <o> in the dictionary. All the ones you might expect to start with <o> start with <uo> instead, and the same goes for word-internal moraic /o/:

otootosoundVoto. Som, ou ſoido. (285)
wotokwootokomanVotoco. Homē varão, ou macho. (285)
awoaoblue/greenAuoi. Couſa de cor azul eſcuro. Itē, de Verde. Item, Verde. i. Que não he maduro. (16)
kapokaofaceCauo. Roſto. (44)

(Note that <V> here is simply a capital <u>. Also note that I'm using the entry <auoi> instead of <auo> solely because the only definition given for the latter is a color that horses can be. Same morpheme, though.)

Frellesvig considers this basically the same phenomenon as the <ye> issue: a range of morae eroding down to /o/, which is then expressed with an on-glide.

However, one important difference from <ye> is that there is a loophole. The best example is probably:

  • Vô (modern /oː/ "big", OJ /opo/)

Here, the second /o/ (the result of the intervocalic /p/ disappearing as discussed in point 1 of this post) has merged with the first one, producing a single long vowel. The two morae now form one syllable, and, logically enough, there is no need to interrupt the syllable halfway through to add an extra on-glide.

4. An extra vowel

Speaking of <Vô>, check these dictionary entries out. All of them, in modern Japanese, would be pronounced /koː/.

  • . Obediencia de filho pera o pay, ou mãy. (52) [孝]
  • . Adu. Aſsi, deſta maneira. (340) [斯う]
  • . Hum certo paſsaro grande. (52) [鴻]
  • . Exercicio, ou habito. (52) [功]

There's a systematic distinction being made between <ǒ> and <ô>. How it works becomes clearer when you look at the original (pre-reform) kana spellings of these words:

  • 孝, 斯う = かう
  • 鴻, 功 = こう

/au/ becomes <ǒ>, /ou/ becomes <ô>. What about words with the NJ pronunciation /Cyoː/?

  • Qiǒ. Miyako. Cidade principal, ou corte. (198) [京, きやう]
  • Qiǒ. Liuro. (198) [経, きやう]
  • Qeô. [ve]l, qiô. Oje. (193) [今日, けふ]
  • Qeô. l, qiô. Prazer, graça, &c. (193) [興, けう]

The rules appear to be:

  • /au/ → <ǒ>
  • /ou/ → <ô>
  • /eu/ → <:iô>

Or we could skip the science and just read the VLI's front matter:

Nos vocabulos que tem acento longo como, Fiǒrǒ, Meǒji, &c. Eſcreuemos a primeira ſylaba ora por, E, ora por, I: & da meſma maneira os que tem o acento breue como Fiô, Qiô, &c. Por que poſto que na letra Cana, ſe eſcreuão huns, Fiau, & outros Feu: todauia na pronunciação não pedem mais, E, que I [...](ii)

In the words which have a long accent like "Fiǒrǒ", "Meǒji", &c., we write the first syllable now as "E", now as "I": & in the same manner those [words] which have a short accent like "Fiô", "Qiô", &c. This is because in the kana letters, [the former] ones are written "Fiau", & the others "Feu": however, the pronunciations do not call more for "E" than "I" [...] (My translation, after Doi et al 5)

That is, the acento longo of <iǒ> comes from kana spellings corresponding to /yau/, while the acento breve of <eô> comes from spellings corresponding to /eu/. Note the denial of any difference in pronunciation between <i> or <e> — it all comes down to <iǒ> vs <eô>.

The current thinking, by the way, is that <ô> was pronounced [oː] (you could see this as simple extension of the original /o/, at least in the cases where there was one), while <ǒ> was [ɔː].


  • Collado, Diego. Ars Grammaticae Iaponicae Linguae. Rome, 1632. Project Gutenberg. Web. 12 March 2012. <http://www.gutenberg.org/ebooks/17713>
  • Doi, Tadao 土井忠生, Morita Takeshi 森田武, and Chōnan Minoru 長南実. Hōyaku Nippo Jisho 邦訳日葡辞書. Tokyo: Iwanami Shoten 岩波書店, 1980. Print.
  • Frellesvig, Bjarke. A History of the Japanese Language. Cambridge: Cambridge University Press, 2010. Print.
  • Man'yōshū Kensaku Sisutemu Ver 2.2.0 万葉集検索システム. Web. 12 March 2012. <http://infux03.inf.edu.yamaguchi-u.ac.jp/~manyou/ver2_2/manyou.php>
  • Spear, Richard L. Diego Collado's Grammar of the Japanese Language. University of Kansas, Center for East Asian Studies: 1975. Project Gutenberg. Web. 12 March 2012. <http://www.gutenberg.org/ebooks/21197>
  • Vocabvlario da Lingoa de Iapam. Nagasaki, 1603. Tokyo: Benseisha 勉誠社, 1978. Print.

Other works cited

  • Allen, W. Sidney. Vox Latina: The Pronunciation of Classical Latin. Cambridge: Cambridge University Press, 1965.
  • Ladefoged, Peter. A Course in Phonetics. Orlando: Harcourt Brace, 1975.



The Vocabvlario da Lingoa de Iapam, a.k.a. Nippo Jisho 日葡辞書 ("Japanese-Portuguese Dictionary") is, like most dictionaries, great fun to browse through. It's kind of baffling that (as far as I know) there's no modern typeset version of any kind available, so we have to make do with facsimiles of varying quality and an admittedly marvelous translated version (Doi, Morita, and Chōnan). Or is there a samizdat e-text floating around somewhere? Someone hook me up!

Much of the fun comes from charming definitions for terms you already know. Take akanu naka 飽かぬ仲. Brinkley et al define it simply as "an inseparable and delightful relation" (16). The Nippo Jisho sez (with my translation):

Acanunaca. Amizade, ou liança como de cazados, ou amigos bē vnidos, & que não ha cauſa por via de deſamor pera ſe apartarē. (2)

Friendship, or alliance like that of the married, or friends well united, & which has not cause by way of disaffection to be ended [separated].

At least, I'm pretty sure that's what it means. This is where Doi et al's Japanese translation comes in handy, although then of course you have the problem of deciding whether to trust them when they say that "amigo" here means "lover" 愛人 rather than "friend" (10) — the Nihon Kokugo Daijiten, for one, disagrees, glossing it 親友 "close friend", and I have no idea what usage was more likely for a Portuguese Jesuit in 1603.

Sadly, IE-based bluffing can only get you so far. And that's why "learn Galician-Portuguese and cautiously move forward to early Modern Portuguese" is on my to-do list. (Well, that and cantigas d'amigo.)

Fortunately, even for those of us who can't read the dictionary properly, it can still teach us all kinds of interesting things about Japanese phonology in the Late Middle/Early Modern period. I plan to post about these a bit next week.


  • Brinkley, Frank, Nanjō Bunyū 南條文雄, Iwasaki Yukichika 岩崎行親, Mitsukuri Kakichi 箕作佳吉, and Matsumura Jinzō 松村任三. An Unabridged Japanese-English Dictionary. Tokyo: Sanseido 三省堂, 1896. Archive.org. Web. 8 March 2012.
  • Doi, Tadao 土井忠生, Morita Takeshi 森田武, and Chōnan Minoru 長南実. Hōyaku Nippo Jisho 邦訳日葡辞書. Tokyo: Iwanami Shoten 岩波書店, 1980. Print.
  • Nihon Kokugo Daijiten 日本国語大辞典. Shōgakukan 小学館. JapanKnowledge. Web. 5 March 2012.
  • Vocabvlario da Lingoa de Iapam. Nagasaki, 1603. Tokyo: Benseisha 勉誠社, 1978. Print.


Did mata mean "head"?

The Yamata no Orochi, as any fule kno, was a giant serpent slain in Japan's Heroic Age by Susonoo. "Orochi", though slightly etymologically murky, apparently means "[some kind of] serpent," and "Yamata" is generally understood (and, in the original sources, written) as "eight-forked" (八俣 or 八岐).

In the Kojiki, the Yamata no Orochi is described as follows.

(Incidentally, I'm following Saigō for Kojiki quotes throughout this post, w/r/t both kanji [which is all the original text has] and native Japanese "reading" [which is disputable and indeed disputed]. Readings are romanized a la Frellesvig. Translations are mine, indebted to Saigō's commentary.)

毎年來喫。 [...] 彼目如赤加賀智而、身一有八頭八尾。亦其身生蘿及檜榲、其長度谿八谷峽八尾而、見其腹者、悉常血爛也。

tosi goto ni kite kuraperi [...] sono me pa akakagati no gotoku site, mi pitotu ni ya kasira ya wo ari. mata sono mi ni koke to pisugi to opwi, sono take pa tani ya tani wo ya wo ni watarite, sono para wo mireba, kotogotoku ni tune ni ti ni tadaretari. (Saigō 2:191-193)
Every year it comes to feed. ... Its eyes are like winter cherries; its body has eight heads and eight tails, and is covered in moss, Japanese cypresses and Japanese cedars. Its length extends across eight valleys and eight hills, and if you look at its belly, it is always covered in blood.

In the Nihon Shoki (Sakamoto et al 1:92, 1:447):

Its head and tail both have eight forks. Its eyes are like winter cherries. Pines and Japanese cedars grow on its back, and it extends across eight hills and eight valleys.

Alexander Vovin's "Pre-Hankul Materials, Koreo-Japonic, and Altaic" raises an interesting point about these descriptions:

At first sight, there is an obvious contradiction: while the Kojiki describes the serpent as having eight heads and tails, the Nihonshoki tells us about eight forks: that is, the serpent should have nine heads and nine tails. More careful scrutiny of both texts, however, reveals that the serpent had eight and not nine heads: Susanowo tricks him by placing eight big jars with sake for each head, which the serpent drinks, then gets drunk, and is subsequently killed. [...] The only way out of this confusion is to assume that -mata in Ya-mata really means "head" and not "fork." (144)

This -mata, Professor Vovin argues, would in turn be related to *matay or *matæ, Vovin's reconstruction of a twelfth-century Korean word for "head" recorded in Chinese characters as 麻帝. The use of 俣 and 岐 (both meaning "fork" in some sense) in the Japanese texts instead of a character meaning "head" is "best explained as ateji," Vovin proposes. "It is also likely that by the early eighth century the real meaning of the word *-mata "head" in Ya-mata was already forgotten, and only the context of the myth in combination with simple arithmetic allows us to reconstruct it" (145).

This is an interesting and bold hypothesis, and Professor Vovin seems quite confident that it is true as he includes -mata in lists of Old Japanese loans from Korea in both Vovin 2005 and Vovin 2010 (in both cases with the note that yamata no orochi is the only example of this mata.)

However, with all due respect to Professor Vovin, I do not feel that this theory is supported by the evidence. X-mata does not mean "forked X times (into X+1 child nodes)." It means "forked into X child nodes" — and mutatis mutandis for related constructions like the Nihon shoki's 有八岐 "had eight forks". There is therefore no conflict between the Yamata no Orochi's name and head-valency, and no need to invoke a nonce-borrowing from Korea.

For example, consider the word putamata. This word is morphologically identical to yamata except with "two" (puta) instead of "eight" (ya). Conveniently, it also appears in the Kojiki, in the Emperor Suinin chapter:

[...] 二俣榲。作二俣小舟而。持上來以。浮倭之市師池。輕池。[...]

[...] putamata-sugi wo putamata-wobune ni tukurite, moti noborikite, Yamato no Itisi no Ike, Karu no Ike ni ukabete [...] (Saigō 5:316-317)
[... They] made the two-forked Japanese Cedar into a two-forked boat, carrying it up to Yamato and floating it on Ichishi Pond and Karu Pond there [...]

This episode is no doubt related to similar stories in the Nihon Shoki: in the Emperor Nintoku chapter, a tree described as "本一以末両" ("one trunk [forked] double") floats down the river and is made into a boat; in year 3 of Emperor Richū's reign, the Emperor launches a "double-forked boat" (両枝船) on Ichishi Pond (市磯池) in Iware (磐余), with the note that he sat on one side and his consort on the other (與皇妃各分乘而遊宴) (Saigō 5:317-318).

There is no argument, as far as I am aware, that these trees and boats have three or four ends (lobes, hulls, whatever) protruding from two separate instances of forking. They are understood to be an object that is united at its base but forked, once, into two.

(Sakamoto et al connect the "double-forked boat" motif of the Nihon Shoki to Southeast Asian/Polynesian catamaran technology (Sakamoto et al 2:449). There is also an interesting discussion of myths and taboos surrounding forked trees in Tsunemitsu, including the legend, new to me, that Kakinomoto no Hitomaru was born from a fork in a persimmon tree (Tsunemitsu 442); the inevitable connection of the forked tree with the manner in which humans are born (431, quoting Hotta); and the observation that in the Kojiki itself Ōkuninushi escaped/was reborn via the fork of a tree after being killed by his eighty brothers for the second time (441).)

We can find more examples by going forward in time. The 1603 Vocabvlario da lingoa de Iapam offers us:

  • Futamata. Forquilha de duas pontas. (Vocabvlario 112)
  • Mitçumata. Forquilha de tres pontas. (363)

In the Sino-Japanese department of the lexicon, we have the morphologically equivalent sansa 三叉, the Nihon kokugo daijiten's earliest citation for which is from the 1508 Rokumotsu zushō 六物図抄 and clearly describes a trident-like, or rather forward-facing-part-of-a-chicken-claw like (same diff, sorry Poseidon) structure: "鳥の爪を前へ三つかけてつかむが如し". Similarly, trigeminal nerve is known as the 三叉神経, with the 1872 Igo ruiji 医語類聚 ("Anthology of medical terminology") the NKD's earliest citation. (I bring this one up because it illustrates that Indo-Europe can use "fork"-like morphemes in the same way: as I understand it, etymologically, "trigeminal" doesn't mean "thrice-twinned [for a total of four]", it means "'twinned' into three.")

Incidentally, futamata is still a word in modern Japanese, and one of the metaphorical senses it has picked up over the centuries is "cheating [on a partner in a romantic relationship]." The folk etymology for this sense is, ahem, "two-crotched," but of course the real etymologically is "divided into two," what we might call "two-faced" in English.

It seems to me that the general meaning of X-mata and equivalent Sino-Japanese constructions, has been from antiquity and still remains "forked into X child nodes," not "forked X times into X+1 child nodes." I see no reason to assume that yamata works differently, except insofar as the use of "eight" also hints at "countless." Thus, using the word yamata to describe an eight-headed dragon is not problematic, and there is no need to invoke an otherwise unattested loanword from Korea.

Why "forks" and not "heads"?

You might ask, well, if this is the case, why use "fork" at all? Why not just say "eight-headed" and be done with it? This is a reasonable question, and I imagine the answer has to do with the Yamata no Orochi as a symbol of Japan's then-untamable and terrifying rivers (see e.g. Tsugita 1:103-104 for discussion and comparison with other mythologies; see also Juhl for a similar story from eastern Japan in the Enoshima Engi).

Rivers, like roads, were more naturally understood as "forked" than "headed". In the "Dialects" section of its entry for futamata, the NKD lists "point where rivers meet" (川の合流点) as a Nagano-specific meaning, and "valley divided into two" (二つに別れている谷) for Oita prefecture. Note also that the Yamata Orochi has eight heads and eight tails; being "split into eight [=many]" is clearly more basic to its nature than simple eight-headedness. So it doesn't really seem odd to me that a "fork" rather than a "head" morpheme would find its way into the name of a river symbol.


  • Frellesvig, Bjarke. A History of the Japanese Language. Cambridge: Cambridge University Press, 2010. Print.
  • Hotta, Yoshio 堀田吉雄. Yama no Kami Shinkō no Kenkyū 山と神信仰の研究. Kuwana: Ise Minzoku Gakkai 伊勢民族学会, 1966. Print.
  • Juhl, Robert. The Goddess, the Dragon, and the Island. 2003. Web. 5 March 2012.
  • Nihon Kokugo Daijiten 日本国語大辞典. Shōgakukan 小学館. JapanKnowledge. Web. 5 March 2012.
  • Sakamoto, Tarō 坂本太郎, et al, eds. Nihon Shoki 日本書紀. 1994. 5 vols. Tokyo: Iwanami Shoten 岩波書店, 2001. Print.
  • Saigō, Nobutsuna 西郷信綱. Kojiki Chūshaku 古事記注釈. 8 vols. Tokyo: Chikuma Shobō 筑摩書房, 2005. Print.
  • Tsugita, Masaki, ed. Kojiki 古事記. 1977. 3 vols. Tokyo: Kōdansha 講談社, 1993. Print.
  • Tsunemitsu, Tōru 常光徹. "Futamata no Reisei to Kaii Denshō" 二股の霊性と怪異伝承. Nihon kaibutsugaku taizen 日本妖怪学大全. Ed. Komatsu Kazuhiko 小松和彦. Tokyo, Shōgakukan 小学館: 2003. 425-443. Print.
  • Vocabvlario da Lingoa de Iapam. Nagasaki, 1603. Tokyo: Benseisha 勉誠社, 1978. Print.
  • Vovin, Alexander. "Pre-Hankul Materials, Koreo-Japonic, and Altaic." Korean Studies 24 (2000): 142-155. Monumenta Altaica. Web. 5 March 2012.
  • ---. A Descriptive and Comparative Grammar of Western Old Japanese: Part 1: Phonology, Script, Lexicon and Nominals. Folkestone, Kent: Global Oriental, 2005. Print.
  • ---. Koreo-Japonica: a re-evaluation of a common genetic origin. Honolulu: University of Hawaii Press, 2010. Print.