Since Georgia has been in the news lately, I've been reading about the Georgian language and alphabets. Although none of the alphabets look like Greek, the order of letters is like Greek with some insertions:

Inserted between the equivalents of ε and ζ: ვ [v]

Inserted between the equivalents of ν and ο: ჲ [j] (now obsolete)

Inserted between the equivalents of π and ρ: ჟ [ʒ]

Inserted between the equivalents of χ and ω:

Numbers indicate the ordering of letters: e.g., 1. ღ [ɣ] is the first letter after the equivalent of χ. The voiceless postalveolars (3-4) and the alveolar affricates have been grouped together (5-7), but the rest seem to be randomly scattered.

The two Greek letters for Cs-clusters have no equivalents: ξ, ψ.

On the basis of the above table, one might guess that Georgian has a three-way opposition between

1. plain

2. ejective

3. voiced

but aspirated stops appear in place of the expected plain stops with the exception of ჴ q (unless it's phonetically [qh] in the dialects that retain it).

(8.24.3:12: ჴ is [qh] in the Svan language.)

One might expect special letters for the ejectives. However, the actual ejective letters correspond to Greek letters for voiceless unaspirated stops:

(8.24.2:32: თ [th] happens to look like the Burmese letter for [t]!)

Note also that Georgian aspirates correspond to Greek aspirates, even though the latter had become fricatives before the first attestation of Georgian writing in the 5th century. The Georgian letters for [v x ɣ] do not correspond to Greek β χ γ. Does this tell us something about the variety of Greek known to the originator of Georgian writing (or at least the originator of the ordering of letters)? (Cf. Cyrillic in which в and х have always corresponded to Greek β and χ. The Belarusian pronounciation of г as [ɣ] postdates the creation of Cyrillic.)

They must be more popular than my Tangut greeting cards.

08.8.22.4:40: 'HAND' AND 'FIVE' IN KRA

In Proto-Austronesian, *(qa)lima 'hand' and *lima 'five' are (nearly) homophonous. The two etyma are still homophones in modern Hawaiian (lima and lima).

In Ostapirat's (2000: 221, 245) Proto-Kra reconstruction, 'hand' is *mja A and 'five' is *r-ma A. Is this (near-)homophony in Kra and Austronesian coincidental, the product of borrowing, or a sign of common ancestry? To approach an answer to this question, I need to closely examine Ostapirat's reconstructions:

Reconstructions in parentheses are mine.

There is no Kra evidence for a liquid in 'hand'.

I presume Ostapirat reconstructed Proto-Kra *-j- on the basis of Pubiao m̥ii B1. No other examples of PK *-ja : Pubiao -ii are available. PK *-j-, the voiceless Pubiao initial and the series 1 tone are reminscent of Sanchong Sui mja A1/A2 'hand', but all other Kra evidence point to a proto-voiced initial that conditioned series 2 tones. The B tone of Pubiao implies an earlier *-h apparently absent in the rest of Kra-Dai.

The rhyme of Proto-Gelao *ml-? 'five' is uncertain. Laozhai has mlen < PG *mlun with an inexplicable final nasal but Qiaoshang mbau and Wanzi mpu imply PG *mlə. A final schwa may be due to the influence of PG *pə 'four'.

Although Paha mhaa A1 'five' points to *r-ma (via *ʁ-ma), there is no other evidence for an earlier *r. I would regard pre-Paha *r-ma as an innovation and reconstruct the Proto-Kra word for 'five' as *l-ma which obviously resembles Proto-Austronesian *lima.

The resemblance of PAN *lima 'hand' to PK *mja is much weaker. Perhaps PK *-j- (and the -j- of Sanchong Sui) is a trace of the *-i- in the lost first syllable:

I wonder if there are any other examples of Kra-Dai etyma with medial glides -j- and -w- corresponding to PAN *i and *u.

*qàlimá > *qalma > *qamla (loss of least stressed vowel?) > *qəmja > *qmja > *xmja > Pubiao m̥ii B1, Sanchong mja A1

but *qəmja > *mja > A(2) m-forms elsewhere in Kra-Dai

08.8.21.1:30: INNATE NUMERALS?

I have two problems with this claim:

Australian Aboriginal children can count even without having words for numbers, according to a study by British and Australian experts released Tuesday ...

The study found that four to seven-year-olds from two Aboriginal communities have an "innate system" to count with, even though their languages only have normal words for one, two, few and many.

First, according to RMW Dixon (1997), only 20 out of the c. 250 Aboriginal languages that existed in the 18th century are still being learned by children. (The number is probably lower now.) Are speakers of only two languages (Warlpiri and Anindilyakwa) an acceptable sample*?

Second and more importantly, do those children truly lack numerals higher than two?

The zompist.com list of numerals in Australian languages lists Warlpiri numerals for 'one' to 'three', with panu 'many' in the column for 'four', followed by rdaka 'hand' in the column for 'five'. This handout has a different Warlpiri numeral for 'three' followed by numerals for 'four' through 'ten' including rdaka 'five'. Are these numerals for 'three' through 'ten' recently coined translation equivalents? Even if they are, are they being used now? And are the children totally unaware of English numerals? In an article on children learning Warlpiri, Bavin in Slobin (1992: 340) mentioned "pala which is used as a suffix on numerals of 4 and over, borrowed from English ... It appears to be borrowed from the English word fellow". Children "hear [English] from the non-Aboriginal employees in the community" (1992: 344). I would guess that exposure to English is even greater now than it was c. 1990. The children may be speaking 'light Warlpiri' instead of Warlpiri proper.

I can't find any list of Anindilyakwa numerals, but John Harris (1987: 34) mentions a 1982 paper by Stokes on Anindilyakwa

mathematical concepts including cardinal number, ordering, position, time, dimensions such as length and mass

According to Stokes. the Anindilyakwa words for 'five', 'ten', 'fifteen', and 'twenty' are not based on hands (ayarrka) or feet, and they do not even appear to be based on each other:

awilyaba 'one'

ambilyuma 'two'

abiyakarbiya 'three'

abiyarbuwa 'four'

amangbala 'five'

ememberrkwa 'ten'

amaburrkwakbala 'fifteen'

wurrakiriyabulangwa 'twenty'

Even if these numerals were lost during the past two decades, surely they would have been replaced by English loanwords.

Harris' paper debunks myths about numerals in Aboriginal languages as a whole:

Even if ambiguous cases are ignored, there is not such a dearth of information on [Aboriginal] numbers above four that serious scholars could be unacquainted with them. The existence of words for five has been known for at least a century ... In anticipation of the unjustifiable but predictable criticism that Aboriginal higher numbers are "not real numbers" because they are compounded from smaller numbers or "hand", I have chosen the above three examples because they are not obviously compound nor is the term for five related to "hand". There are many other such examples. There are also some refernces to "regular quinary" systems of numerals which seem to extend indefinitely upwards (Wurm 1972: 64; Capell 1958: 27).

If 'five-one' for 'six' is not 'real', then Khmer has no 'real' numerals for 'six' through 'nine' which are 'five-one', 'five-two', etc. (Yet the Khmer were able to build Angkor Wat without so-called 'real' numerals, excluding Indic loans!)

Despite my linguistic objections, I don't object to the notion of counting without numerals. How could humans create words for numerals if they did not already have numerical concepts (which aren't even unique to our species)? Harris (1987: 36) met

a profoundly deaf Groote Eylandt girl in grade two at Angurugu school at the time when the language of instruction was English. She was unable to speak either Anindilyakwa or English. She lip-read a little and communicated by gesture. She was by far the most competent mathematics student in her grade. She was very fast at all counting, grouping and sorting activities with structured materials and readily transferred the concrete tasks into written numerals. However she counted, whatever conceptual tags she used, they were not number words. She lacked what would normally be considered to be the language of counting but she could certainly count. I am not claiming that she had been uninfluenced by Western mathematics through school experiences. Rather, I am using her as a striking example of the fact that words are not necessary for real counting to occur.

*Sampling is on my mind because of my recent series on Proto-Kra. I don't think one should project etyma found only in a single branch of Kra back to Proto-Kra unless those etyma also exist in other branches of Kra-Dai. For example, if an etymon is in Gelao and in Hlai, then it must be a Proto-Kra-Dai etymon that was lost in all of Kra except Gelao. (Geography rules out the possibility of borrowing between Gelao which is far from the shore and Hlai on the island of Hainan.) It is possible that some Proto-Kra or even Proto-Kra-Dai etyma survive only in a single branch, but historical linguists should focus on the probable.

08.8.20.23:35: F-REQUENCY

How many instances of f are in this sentence?

FINISHED FILES ARE THE RESULT OF YEARS OF SCIENTIFIC STUDY COMBINED WITH THE EXPERIENCE OF YEARS.

Answer here.

I knew that the sinograph 她 for 'she' was coined in modern times, but I didn't know the name of its inventor until tonight: 劉半農 Liu Bannong*. 她 is based on 他 'he' with the same phonetic (they are homophones) but with 女 'woman' in place of 亻 'person'.

I wonder who invented 牠 'it' which has 牛 'ox' instead of 亻 'person' or 女 'woman'. (The choice of 'ox' was probably influenced by 物 'thing' which has 牛 'ox' on its left.)

In speech, there is no way to distinguish 他, 她, or 牠. They are all pronounced ta in standard Mandarin. The notion of a three-way gender distinction in pronouns came to East Asia from the West. Japanese kanojo and Korean kUnyO (both lit. 'that woman') were coined as equivalents of 'she'. (Kor kunyO could be a calque of kanojo.)

*The surname Liu is wrtiten with a sinograph 劉 for an obsolete homophone 'kill'. 劉 has been simplified as 刘 = 文 'literary' + 刂 'knife, sword'. 半農 Bannong is 'half' + 'farmer'.

8.19.2:45: This answer to the question "Is the meaning 'she' absent from pure Korean?" credits Korean literature scholar 梁柱東 Yang Chudong** (1903-1977) with the first appearance of kUnyO in print in August 1926, but also notes that it wasn't widely used until 1954 (why that year?). The article also lists other neologisms for 'she': kUhi, kUmE, kUmi, kUne, kUni, kUyO, kUyOja. All contain kU- 'that' and the last two also contain the Sino-Korean morpheme 女 (n)yO 'woman'.

**The graphs for his name are 'roof beam' + 'pillar' + 'east'.

I own his 朝鮮古歌硏究 ChosOn koga yOn'gu (A Study of the Old Songs of Korea).

08.8.19.1:39: "AN EXCELLENT INFORMANT"

Mr. Shwe Waing is an excellent informant. His knowledge of English is limited and his dislike of speaking it is marked. He has no tendency toward philosophizing about language in either Burmese or English, and his ability to explain in Burmese the meaning of a Burmese locution is phenomenal. His patience and co-operation are unlimited.

Until I saw Sydney Lamb's article in this book, I had no idea that Cornyn was also a Russian language instructor involved in a machine translation project with Lamb. A related Chinese MT project involved Samuel E. Martin and Yuen Ren Chao as well as Lamb.

08.8.19.00:43: GEDNEY AT GOLIATH

On Saturday, I was surprised to discover Thomas John Hudak's "William J. Gedney's Elicitation Questionnaire" (2004) at Goliath, a site offering "Business Knowledge on Demand". I have no idea why Goliath datamines The Journal of the American Oriental Society. If an understanding of Tai tones were a key to financial success, I'd be on my way to the top.

Seriously, Willam J. Gedney (1915-1999) was a pioneer of Tai linguistics. I still haven't gotten around to reading all of his 1947 PhD dissertation on Indic loans in Siamese.

08.8.19.00:29: DE(S)CENT EVIDENCE FOR ASPIRATE LENITION IN VIETNAMESE? (PART 1)

While writing "From *m to z", I found this passage by Gregerson (1969: 161):

The witness of doublets such as giuống with xuống 'descend' also group it [Middle Vietnamese x-] with the palatal set of sounds.

At first, I thought that giuống might be an old borrowing of

降 Middle Chinese / Late Old Chinese *kæwŋh < Old Chinese *kruŋ-s 'descend'

However, the sắc tone of giuống implies MC/LOC *-ʔ, not *-h. Moreover, -uông corresponds to MC/LOC *-uoŋ, not MC/LOC *-æwŋ which has no *-u-like vowel. So even though its vietograph is 降, I consider its resemblance to the Chinese etymon (and the Sino-Vietnamese reading giáng < Late Middle Chinese *kjawŋ) as coincidental. This is one of those rare cases in which a non-Chinese word happens to sound like a Sinoxenic reading. (Another is Jpn shin- : 死 SJ shi 'to die'.)

I prefer to view giuống as a prefixed form of xuống < *ch-:

gi- (17th c. *[ɟ] or [dʑ]) < *CV-ʑ- < *CV-ɟ- < *CV-ch-

But I don't know of any other doublets with alternations between lenited and aspirated initials:

v- ~ ph-

d- ~ th-

g- ~ kh-

To complicate matters further, xuống was written with at least five different vietographs* containing the phonetic 竜 long 'dragon', implying that x- is from a *Cl- or *Cr-cluster, even though such clusters should have resulted in later tr- < *Cl- or s- < *Cr-.

Next: External evidence for *-r- in 'descend'.

*nomfoundation.org lists 12 different vietographs for xuống. Five of the vietographs contain shapes like 冘 resembling 龙, one of the many simplifications of 龍 long 'dragon'. Only two have no phonetic:

宀 (roof) + 下 'down' (not even in Unicode CJK Unified Ideographs Extension B)

𨑜 < 辶 (forward motion) + 下 'down'

08.8.18.4:31: FROM *M TO Z IN OLD CHINESE AND VIETNAMESE

In "Rhyming Rising Roots?", I mentioned two words with Old Chinese *ml-clusters:

塍 893n-o *mləŋ 'raised path between fields'

乘 895a *mləŋ 'mount, ride, ascend'

One might expect these words to have initial m- in Middle Chinese and modern Chinese languages, but in fact they have sibilant initials:

塍乘 Middle Chinese *ʑɨŋ, Md cheng

For a long time, I was troubled by the *ml- to *ʑ- change despite the internal and external evidence in Sagart (1999: 79). I could not think of a similar change in another language until Saturday night when I realized that I had been overlooking the obvious for years.

In Vietnamese, borrowings with palatalized *mʲ- in Late Middle Chinese have initial d- ([z] in Hanoi): e.g.,

民 LMC *mʲin > Middle Vietnamese ?*[dən] > Sino-Vietnamese dân [zən]

The start and end points of the change are clear, but the middle is uncertain. Here's how I account for Sino-Vietnamese reflexes of all Late Middle Chinese palatalized labials and other sources of Vietnamese [z]:

I assume that LMC had already undergone obstruent devoicing: e.g., *pʲ- < *pʲ- and *bʲ- (but with the earlier voicing still reflected in different tones).

'Non-LMC' refers to earlier strata of borrowing from Chinese as well as native Vietnamese syllables.

'Old Vietnamese' has no alphabetic records unlike Middle and Modern Vietnamese.

Comments on rows (with names based on Vietnamese spelling):

T1: I was tempted to reconstruct *bj- in Stage 1 to match *gj- in GI2, but *gj- has a modern voiced reflex whereas *bj- (my current *pj-) wouldn't, and I didn't want to create an arbitrary devoicing rule that affected *bj- but not *gj-.

T2: It is unclear whether Vietnamese borrowed Chinese alveolar affricates as fricatives or affricates. In any case, Chinese *s- was borrowed as Old Vietnamese *s-.

TH1-TH3: An aspirated *sh- is bizarre but exists in Burmese and Korean. (Korean ㅅ s is [sh].)

TH3: It is unclear whether Vietnamese borrowed Chinese *ɕ- as *ɕ- or as the exotic *sh-.

D3: The intervocalic lenition of *-t- followed by presyllabic loss has parallels in Kra (Qiaoshang and Paha).

GI1: The palatal counterpart of D3. In central and southern Vietnamese, GI1 has become j-. Cf. how Paha medial *-tʃ- became j-.

GI2: I would have expected

*kj- > *kɕ- > *ks- > *ts- > *s- > [t]

by analogy with T1. Why is GI2 different? Is the fact that velar *k- is closer to palatal *j- than labial *p- relevant? Cf. Burmese, in which *kj- fused into c-, unlike *pj-. (But affricate reflexes of palatalized labials are certainly possible: e.g., Portuguese chama < *fj- < *flamma 'flame' and chuva < *pj- < *pluvia 'rain' [Parkinson in Comrie 1987: 268].)

Comments on stages of Vietnamese:

Stage 1: I did not want to give *gj- its own column, since the voicing of its initial element seems to have been a sui generis change in Vietnamese.

Stage 2: Shift of *-j- to a palatal fricative whose voicing was determined by the preceding consonant. Intervocalic voicing after presyllabic vowels. (This could have occurred long before stage 2, and is merely put in the same column for convenience.)

Stage 3: Nonmedial palatal fricatives became alveolars after nonpalatals (zero, labials, and dentals). Hence Stage 2 *ʑ generally became *z, but further medial lenition after presyllabic vowels results in a new Stage 3 *ʑ.

Stage 4: Heterorganic clusters became homorganic: *ps- > *ts-, etc. Presyllables were lost, so lenited medials became new initials.

Stage 5: Affricates became fricatives.

Middle Vietnamese: Fricatives hardened to stops (though gi- could have been an affricate as reconstructed by Gregerson [1969: 161]).

Modern Vietnamese: d- and gi- have lenited to [z] or even to [j]. r- is in various states of lenition: [r] (none), [ʐ,], and [z].

Compare row D2 in the Vietnamese table above to the *ml-row below. OC stage 3 *mʑ- is identical to OV stage 2 *mʑ-.

(This table does not incorporate the tristratal hypothesis and its stage numbers have nothing to do with the stage numbers in this table.)

In Stage 2, *(h)l became lateral fricatives and in Stage 3, these fricatives became sibilants. The unusual clusters of Stage 3 were simplified in Stage 4. Stage 4 ʑ- was the result of a chain shift:

*mʑ- > *ʑ- > *j-

OC *ml- has both fricative and affricate reflexes in later Chinese languages: e.g.,

乘 OC *mləŋ >

MC *ʑɨŋ > Shanghai zəŋ, Cantonese siŋ

but Md cheng implies MC *dʑɨŋ

The affricates may reflect an *m- that denasalized and assimilated to the following alveolar in some Late Old Chinese dialect(s). Some modern fricatives may be from earlier affricates: e.g., some Shanghai z- are from *dz- (which in turn could be from *dʑ-).

08.8.17.5:08: *ST-RATA OF *ST-CLUSTERS IN OLD CHINESE

稱 894g OC *s-təŋ > later OC *thəŋ > MC *tɕhɨŋ 'to lift'

升/昇/陞 897a-d OC ?*s-təŋ > MC ɕɨŋ 'ascend'

Did *s-t- develop at random? No, because my first and second *s-təŋ are from different periods during the long history of Old Chinese:

In this scenario, OC *st-syllables have three possible origins:

1. Original *s(V)t-roots (impossible given Sagart's constraints on OC root structure; no such roots are known - do any Austronesian words with *SVt- correspond to OC *st-?)

2. Roots prefixed with *sV-

3. Roots prefixed with *s- (if there were such a prefix distinct from *sV-)

Stratum 1 syllables were the first to undergo *s-t-fusion. I am hesitant to propose a *ht- stage between *s-t- and *th- because initial preaspirates are rare.

Another possible source of stratum 1 aspiration is *k-: *kt- > *xt- > ?*ht- > *th-. I believe that *k- is also a possible source of aspirates in Tangut words with unaspirated roots. (*s- cannot be a source of Tangut aspirates since it conditioned vowel tension.)

Stratum 2 syllable initials underwent metathesis.

Stratum 3 syllable initials palatalized to *stɕ- which may have briefly become *ɕɕ- (like Russian щ) before simplifying to *ɕ-. Proto-Min may have merged *stɕ- with *tɕ-: e.g.,

升 OC *stəŋ > *stɕɨŋ > Proto-Min ?*tɕiŋ > Xiamen tsin, Fuzhou tsiŋ 'measure of capacity'

升 'ascend' with s- in Min could be later borrowings from Middle Chinese *ɕɨŋ

書 OC *sta > Chaozhou tsɿ, Fuzhou tsy 'book'

The prefixes in Stratum 2 syllables may be newer than those in Stratum 1, and the Stratum 3 prefixes may be the newest of all.

Voiceless sonorants and at least some aspirates were from stratum 1: e.g.,

*sŋ(ʷ)-, *sn-, *sm-, *sr-, *sl-, *sw- > *hŋ(ʷ)-, *hn-, *hm-

?*sj-, *sr-, *sl-, *sw- > ?*hj-, *hr-, *hl-, *hw-

*sk(ʷ), *sts-, *sp- > *kh(ʷ)-, *tsh-, *ph-

Sagart's (1999: 69) table of *s-cluster development covers strata 2-3:

*sŋ-, *sn-, *sm- > *s-

Stratum 2 *s-hn-, *s-hl- > *tsh-

Stratum 3 *s-hl- > *sɕ- > *s-

Stratum 2 *sl- > *zd- (cf. Classical Greek ζ [zd]*) > *dz-

Stratum 3 ?*sj-, *sl- > ?*zj-, *zl- > *z-

But why did *sr- become voiceless *ʂ- instead of voiced *zr- > *ʐ-?

Could MC *ʂ- be a merger of

*sr- > *zr- > *ʐ- and

*rs- > *rʂ- > *ʂ-

stratum 3 *ks- >*kʂ- > *ʂ-: e.g., 殺 *kset > *ʂɛt, 山 *ksen > *ʂɛn

but stratum 2 *ks- > *kʂ- > *tʂh-: e.g., 剎 *ksat > *tʂhæt

Stratum 2 clusters became affricates whereas stratum 3 clusters became fricatives. Transcriptions like 烏弋山離 'Alexandria' (Former Han) and 剎那 Skt kṣaṇa 'moment' (Later Han) may help date the two strata (and these two particular examples imply that my numbers for the strata should be reversed.)

No transcriptions reflect stratum 1, which could even predate the birth of sinography.

*I've long considered [zd] for ζ to be improbable, and I just discovered I'm not alone. How frequent is initial zd- in the world's languages?