Andrew West recently mentioned the Japanese name 五十鈴 Isuzu which has an unusual spelling.

The Chinese character 五 'five' corresponds to Japanese i- 'five' from previous entries

鈴 'bell' corresponds to suzu 'bell'.

So what is 十 'ten' doing in between? Is it silent?

A similar name is 五十嵐 Igarashi:

Once again, 五 'five' seems to correspond to Japanese i- 'five'.

嵐 'hillside haze' corresponds to Japanese arashi 'storm'.

(This is a case in which the Japanese use a Chinese character to represent a native word with a different meaning.)

One might conclude that 十 'ten' corresponds to Japanese ga, but 十 is never read ga in any other word. What's going on?

For years, I assumed that -ga- was an archaic genitive marker, but Andrew suggested that Igarashi could be ig-arashi. Could ig- or iga be an old word for 'fifty' that only survives in this compound? The independent native Japanese word for 'fifty' is iso (-so is a suffix for 'ten'), not iga. (The common Japanese word gojuu 'fifty' is a borrowing from Chinese.)

Odder still, the archaic word for 'fifty days' is ika (ka is 'day') but 'five days' is itsuka. One might expect ika to be 'five days' and itsuka with the extra syllable to be 'fifty days'. Maybe i 'fifty' was originally [ii] (double-five? but wouldn't that be ten, or five times five = twenty-five?) whereas 'five' was [i].

The Iwanami Dictionary of Archaic Words (rev. ed., 1990: 81) lists two meanings for i:

1. 'five'

2. 'fifty' an initial element in compounds

It explains that

i 'five' and i- 'fifty' were originally the same word [in Japanese]. This is similar to how 'four' and 'forty' are expressed with the same word in Polynesian languages, etc.

Is this true? I've never heard of such homophony and am skeptical. Hawaiian is a Polynesian language with distinct words for 'four' and 'forty':

ʻe-hā 'four' (ʻe- is a prefix in other numerals; see the bottom of this list of numerals in Austronesian languages)

kana-hā 'forty' (kana- is a suffix for tens)

But I don't know if 'forty' is homophonous with 'four' are in other Polynesian languages. DID PROTO-JAPONIC HAVE FRONT AND BACK VOWEL CLASSES?

Proto-Japonic had at least six vowels:

*i *u
*e *o

Martin 1987 reconstructed only *a, *i, *u, *ə. *e and *o are reconstructible on the basis of Ryukyuan. I have provided the philological evidence for *e and *o in my 2003 article "Philological evidence for *e and *o in Pre-Old Japanese". I am not yet convinced that the seventh vowel proposed by Bjarke Frellesvig and John Whitman (2008) is necessary.

The mid vowels *e and *o merged with *i and *u except in final position. (This is a first approximation.)

In my last entry, I found two classes of vowels in the 'single' and 'doubled' Old Japanese numerals:

Nonback (single bases) Back (double bases)
High i u
Nonhigh ə a (IPA [ɑ]?)

Perhaps the Proto-Japonic noncentral mid vowels (in bold) could be added to this classification:

Front (single bases) Back (double bases)
High *i *u
Mid *e (later to merge with *i) *o (later to merge with *u)
Low (later ə?) *a (IPA [ɑ]?)

A shift of to ə occurred in Chinese: e.g., Middle Chinese 庚 *kæŋ became Mandarin geng [kəŋ]. (Incidentally, ə happens to be the symbol for [æ] in the Azerbaijani alphabet.)

Maybe was always central *ə, which is still fronter than back *ɑ.

If Japonic numerals really were derived via vowel alternations, I would expect to see the same alternations in other words involving quantity: e.g.,

pitə 'person' could have had a plural *puta 'people' with back vowels indicating 'more'.

Cf. Arabic 'broken' plurals with different vowels: rajul 'man' : rijaal 'men' (or English man : men!)

But the actual Old Japanese word for 'people', tami, contains both a back vowel and a front vowel.

David Boxenhorn proposed a lost prefix that altered the root vowel. I'll write that prefix as *CU- with *U representing an unknown back vowel:

*pitə- 'one' > *CU-pitə- 'double-one' > *CU-puta- > *puta- 'two'

*mi- 'three' > *CU-mi- 'double-three' > *CU-mu- > *mu- 'six'

*yə- 'four' > *CU-yə- 'double-four' > *CU-ya- > *ya- 'eight'

This is reminscent of the low-vowelled prefixes I have reconstructed as sources of 'emphasis' (indicated with underlining) in Old Chinese: e.g.,

六 OC *-ruk > *Cʌ-rouk > borrowed into Proto-Tai as *xrok 'six' (could *x- be a trace of the prefix?); Chinese dialects which once had the prefix may be the source of Sino-Japanese roku and the minor Mandarin reading lu

cf. unprefixed OC *ruk > *luk > *liwk > Sino-Japanese riku, Mandarin liu

Many Old Chinese words can be reconstructed with 'emphasis'-triggering prefixes, whereas there is no reason to reconstruct lost prefixes or vowel alternations in nonnumerical Proto-Japonic vocabulary. JAPANESE NUMERALS: 1 3 4 > 2 6 8?

The first ten Old Japanese numerals can be divided into three sets:

Single bases pitə- 'one', mi- 'three', yə- 'four'
Doubled bases puta- 'two', mu- 'six', ya- 'eight'
Other i- 'five', nana- 'seven', kəkənə- 'nine', təwo (< təwə?) 'ten'

Three of the four numerals in the 'other' category are odd: 'five', 'seven', 'nine'.

The single bases have the nonback vowels i and ə.

The double bases have the vowels u and a instead of i and ə:

pitə- 'one' > puta- 'two'

mi- 'three' > mu- 'six'

yə- 'four' > ya- 'eight'

If a were nonback, then I could regard both u and a as back vowels contrasting with the nonback vowels of the single bases.

Nonback (single bases) Back (double bases)
High i u
Nonhigh ə a (IPA [ɑ]?)

One might even be tempted to rewrite ə as æ or e, the front (rather than central) counterpart of back a. But Old Japanese and pre-Old Japanese already have an e (and an o) absent from this proposed paradigm.

Although 'five' has the single base vowel i, 'ten' is not u (the double base vowel corresponding to i).

No other Japanese numerals have these vowel alternations: e.g., 'twenty' was probably Old Japanese pata- (unattested until Middle Japanese), not the expected tawo or tawa from təwo < *təwə 'ten'.

A handful of Japanese words have what appear to be ə ~ a alternations at first glance: e.g.,

kə 'this' : ka 'that' (distal) (but 'that' [mesial]!)

kətə 'word' : katar- 'speak' (if -r- is a verbal suffix)

(The latter was observed by Leon Serafim. I'm not sure about the former.)

Are these vestiges of earlier, more productive alternations, or is the similarity in these words (and in the numerals) coincidental?

I cannot think of any nonnumeral instances of Old Japanese i ~ u alternation.

Cases like Middle Japanese kami ~ kamu- 'god' are not true cases of i ~ u alternation because their i is from an earlier diphthong:

Modern Japanese kami (kamu-, modern pronunciation of earlier kamu-)
Middle Japanese kamu-*
Old Japanese kamɨ (with barred ɨ; not kami!) kamu-
Pre-Old Japanese *kamu-i or *kamo-i *kamu- or *kamo-

*In MJ kamudakara 'treasures offered to the gods' (Genji, Miotsukushi).

Pre-OJ *-o- in 'god' was suggested as a possibility by Leon Serafim, who suggested that the name Kamo is from the bare stem for 'god'.

There is no evidence for pre-OJ diphthongs in pitə- 'one' or mi- 'three'.

2.4.0:30: There is an early Shilla place name written with both 推 'push' and 三 'three':

推良火 ~ 三良火


(良 represents *ra in both Korea and Japan and 火 'fire' represents a common place name element [nearly?] homophonous with the ancestor of Middle Korean pɯr 'fire'; the similarity to Sanskrit pur 'city' [> -pore in Singapore] is coincidental)

In Middle Korean, the descendant of the Shilla language, 'push' was mir-. The use of 三 'three' to write a syllable homophonous with 'push' suggests that there was once a Shilla word (or a non-Shilla word known to the Shilla) for 'three' like *mi(r). (The native Shilla word for 'three' was probably something like *səkV.) If the reconstruction *mi(r) for 'three' is correct, then a Japonic language - a peninsular relative of Japanese - may have once been spoken in or near Shilla territory. JAPANESE NUMERALS: FIND THE ODD MEN OUT*

Can you find a system of derivation for six of these ten Old Japanese numerals?

pitə- 'one'

puta- 'two'

mi- 'three'

yə- 'four'

i- 'five'

mu- 'six'

nana- 'seven'

ya- 'eight'

kəkənə- 'nine'

təwo (< təwə?) 'ten'

Answer tomorrow.

This topic has been in my queue for days now, but I've pushed it ahead of others since Andrew West mentioned it.

*Hint: One of the odd men out is actually even. WHEN FIVE IS FOURTH

The Khitan calendrical system used the Chinese set of twelve earthly branches symbolized by animals. In the Khitan large script, the fourth earthly branch (hare) was written in three different ways (Kane 2009: 176):



3. (< + )

The third spelling is a vertical stacking of the characters in the first spelling and is reminscent of the character combinations of the Khitan small script (see below). (Were such compound large script characters predecessors of the small script, or were they influenced by the small script?)

All three spellings contain a character or a character component resembling Chinese 五 'five'.

is also the Khitan large script character for 'five'.

Why would 'fourth earthly branch' be written with 'five'?

In the Khitan small script, the fourth earthly branch is written completely differently as a combination of three characters:

The first component in isolation means 'five' just like the large script character . Let's assume they're homophonous. We can then assume by process of elimination that

small script = large script or

Hence 'fourth earthly branch' sounded like 'five' plus some additional sounds represented by the other characters.

The large script character was used to write the second half of a Khitanization of Khitan period Chinese 招討 *tɕjewthaw 'suppress'. Thus and, in turn, must have sounded like *thaw. Kane (2009) reconstructed their pronunciation as *tau.

On the basis of other evidence, Kane reconstructed as *li and as *a, so the small script spelling.

is a stack of *tau and *li atop *a.

I presume that the large script characters and corresponding to *li-a were both *lia, and the complex large script character was *taulia.

Janhunen (1994: 14) described the small script spelling of 'hare' as being "of importance for the genetic identification of the Khitan language." In Classical Mongolian, 'hare' is taulai (not taulia, but very close) and 'five' is tabun (not tau). I don't know how to explain the correspondence between CM -ai and Khitan -ia in 'hare', but Khitan, though older than CM, has lost -b- and -n in 'five'.

If one encounters a language in which two basic vocabulary* words A and B are written with the same character X, and there is a known language whose words for A and B share the same sound (sequence), it is possible that the two languages are related (or even the same). For example, if I find a script in which 'hand' and 'five' are written with the same character, that script may represent an Austronesian language since those two words are often homophonous in Austronesian (e.g., Hawaiian lima 'hand/five'). (Obviously 'five' is from the five fingers of a 'hand'.)

*Basic vocabulary is less likely to be borrowed and is therefore a better indicator of a potential genetic relationship. MONGOLIC MATH: TWO TIMES THREE EQUALS ...

... six. Were you expecting five? If so, I do have some confusing Mongolic 'math' in my next post.

Thanks to Guillaume Jacques for leading me to Juha Janhunen's (2003) discussion of Proto-Mongolic numerals. It turns out that Proto-Mongolic *jirgu-ɣa- 'six'

is transparently a compound word, analysable as *jir+guxa/n '2 x 3', with 3 *gu(r)ba/n as the latter component. This, on the other hand, suggests that the Proto-Mongolic numeral for 'six' was also an innovation replacing a more original stem. (Janhunen 2003: 17)

Proto-Tungusic *ɲu-ŋu- and Proto-Japonic *mu- are 'six', not 'two (times) three', so it makes no sense to compare them with Proto-Mongolic *jir-gu- 'two-three'. And PM *ji-r 'two' is definitely not cognate to PJ *puta- 'two' and only shares an initial with Manchu and Jurchen juwe 'two'.

Which other languages have 'X times Y' lower numerals? I can only think of 'X times Y' higher numerals like French quatre-vingts 'four twenties' for 'eighty' or the 'five plus X' numeras of Khmer: e.g., ប្រាំមួយ prammuəy 'five-one' = 'six'.

Tangut fonts by Mojikyo.org
Tangut radical font by Andrew West
All other content copyright © 2002-2010 Amritavision